Split Air Flow Vertical or Horizontal 40mm Case

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Fri Apr 21, 2017 12:01 pm

-
Passive vs Active Comparison
I know a lot of people are running semi-passive or fully passive systems. For comparison, I ran the P6 cu shim/perch configuration passively on the same OS/kernel-setup as the previous actively cooled tests. The Split Airflow Case was in the vertical position for all tests and the ambient temperature was between 69(20.5) to 70 (21.11). The vertical orientation really seemed to work well in a passive configuration and performed likewise. While running, the heatsink was hot with lots of warm air noticeably exiting the top of the case.
Passive-Active_sysbench100k_chart_s.png
Passive-Active_sysbench100k_chart_s.png (61.1 KiB) Viewed 7022 times
Passive-Active_sysbench100k_chart.png
Passive-Active_sysbench100k_chart.png (40.17 KiB) Viewed 7022 times
Included in the test was the actively cooled P0 Aluminum heatsink and the actively cooled P6 Cu shim/perch results for comparison. I was impressed that the passive P6 cu shim/perch system thermally outperformed the actively cooled Aluminum heatsink and gave a similar time result with little to no performance loss.
Passive_sysbench300k_chart_s.png
Passive_sysbench300k_chart_s.png (47.63 KiB) Viewed 7022 times
Passive_sysbench300k_chart.png
Passive_sysbench300k_chart.png (31.68 KiB) Viewed 7022 times
There was clearly a thermal and performance difference for the 1.8ghz sysbench 300k test.

PorterDon
Posts: 15
Joined: Sat Apr 01, 2017 8:56 pm
languages_spoken: english
ODROIDs: C1, C2(2), XU4, XU4Q, HC1
Location: Thailand
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by PorterDon » Wed Apr 26, 2017 1:57 pm

Hugely interesting for me to follow along and observe this development! :geek: Some great problem-solving taking place here.

@Hominoid, I am ready to order one. Is your design final yet? Any reason for me to wait a bit?

@Ameridroid, are you ready to fulfill orders once the design is final?

Love this case, and love the XU-4! I'm done with all the other inferior SBCs. :D

Don (IT dinosaur retired in Thailand)

phaseshifter
Posts: 3520
Joined: Fri May 08, 2015 9:12 am
languages_spoken: english
ODROIDs: U-2,U3+,,XU-3,,XU3-LITE,,XU-4
C1+,,C-2,,,
N-1,,N-2,...other odroid acc`s as well
Has thanked: 41 times
Been thanked: 43 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by phaseshifter » Wed Apr 26, 2017 2:38 pm

ameridroid are advertising them 4 sale

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Thu Apr 27, 2017 12:25 am

PorterDon wrote:Hugely interesting for me to follow along and observe this development! :geek: Some great problem-solving taking place here.

@Hominoid, I am ready to order one. Is your design final yet? Any reason for me to wait a bit?

@Ameridroid, are you ready to fulfill orders once the design is final?

Love this case, and love the XU-4! I'm done with all the other inferior SBCs. :D

Don (IT dinosaur retired in Thailand)
No reason to wait, the final release is posted and I believe Ameridroid is taking orders at their web site. Thanks for following along and the encouragement! Hope to get some operational feedback once your up and running. :)

Ameridroid
Posts: 256
Joined: Thu Mar 17, 2016 4:19 am
languages_spoken: english
Has thanked: 32 times
Been thanked: 27 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by Ameridroid » Thu Apr 27, 2017 6:02 am

PorterDon wrote: @Ameridroid, are you ready to fulfill orders once the design is final?
Yes, we've been printing and shipping already. When we get updates from hominoid, we implement them ASAP.

Vort
Posts: 153
Joined: Fri Nov 22, 2013 10:04 pm
languages_spoken: english
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by Vort » Sat May 06, 2017 12:22 am

Great work, bravo!

I am wondering about the temperature spikes on the hominoid's graphs after the test load is finished (the cooling down part).
What causes them?

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Sat May 06, 2017 11:40 pm

Vort wrote:Great work, bravo!

I am wondering about the temperature spikes on the hominoid's graphs after the test load is finished (the cooling down part).
What causes them?
I believe it is just one or more system housekeeping processes that run normally. Notice they are relatively evenly spaced. It is not unusual for the temperature to increase by 10 degrees or more when a process starts. In all tests, I was watching the watchtemp.sh script waiting for the system to cool down to the target temperature (or for a fixed time depending of the test). No other user tasks were running during this period. In general I have also noticed that if you raise the line weight above 1 in GNUPlot or other charting software, you start to loose the detail in the plot. There is a smoothing affect that happens as the line weight increases. I specifically used a thin line (line weight 1) in order to capture the nuances of the tests.

Vort
Posts: 153
Joined: Fri Nov 22, 2013 10:04 pm
languages_spoken: english
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by Vort » Sun May 07, 2017 7:20 am

hominoid wrote:
Vort wrote:Great work, bravo!

I am wondering about the temperature spikes on the hominoid's graphs after the test load is finished (the cooling down part).
What causes them?
I believe it is just one or more system housekeeping processes that run normally.
Thank you hominoid.
Two years ago I made similar stress tests on my U3 and didn't notice these spikes.
May be it's because I didn't monitor long enough after stopping the test load. :)

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Sun May 07, 2017 11:21 am

Update
After running for several weeks on the final released version of the Split Airflow Case with the P6 copper shim/perch, everything is running cooool. I have been operating in the vertical position with the same fan profile as in the testing. The fan runs at 25% all the time, rarely faster. I would run passively but it is so quiet I don't hear it at all. Low to mid 30's seems to be general running temps with an occasional burst into the 40's in ambient temperatures of 20-21.5c

Important Note
I run sysbench prime 100k anytime I remount the heatsink as a standard practice to verify proper installation, here's why. A few times, during testing the many heatsink-shim mountings, the heatsink was mounted slightly cocked. A couple times it was obvious, when sysbench was run, the temperatures would start to jump by 4, 5 or 6 degrees at a time, immediately. Another time the temperature overall was just higher, faster than it should have been. And in one instance, everything looked good until 80c before thermal runaway occurred. It started to jump many degrees at a time and hit 104c before I was able to shut it down.

The few times there was an issue, simply loosening and redoing the heatsink tightening process fixed the poor mount. It doesn't have to be off very much for a problem to develop. For those who are going to try different heatsink-shim configurations, I would strongly recommend adopting this practice and run sysbench 100k to verify a proper installation before general operation. And, pay close attention when your tightening the heatsink down, keep it level, don't over tighten it. I use the edge of the 30 pin connector as a reference to check if the heatsink is level when complete. If something seems off, redo it.

I also wanted to share a GNUPlot script to chart multiple test files for those who are interested. Originally I was doing the test charts in a spreadsheet program because it was the fastest and the easiest way. I needed to combine the data anyway. But I wanted another way to create the same chart from the command line from the unmodified data files. I eventually circled back and wrote two GNUPlot scripts, multifile_temp.gpl and multicore_temp.gpl

GNUPlot version 4.6 is the current stable distribution but does not have the command line argument feature of the latest version 5.x. I was able to write the scripts so they can be used on either version by simply editing the file and enabling some text variables for the file names; see the comments in the script file.

multifile_temp.gpl

Code: Select all

#!/usr/bin/gnuplot --persist
#   
#
# Decription: GNUPLOT script file that plots temperature and 
#             frequency from multiple watchtemp.sh data files
#             formated as "count,temp,freq1,freq2,freq3,freq4"
#
# Version 5.x
# Use: gnuplot -c multifile_temp.gpl outputchart.png file1 file2...file8
#
# Version 4.x
# Use: gnuplot multifile_temp.gpl
#
# Author: @hominoid www.forum.odroid.com
# Date: 4/30/2017
# Released to the public domain with no warranty of any kind
#

# for version 4.x only, uncomment and enter file names instead of on the commandline, see use above
#ARG1 = "multifile_tempchart.png"
#ARG2 = "test1.txt"
#ARG3 = "test2.txt"
#ARG4 = "test3.txt"
#ARG5 = "test4.txt"
#ARG6 = "test5.txt"
#ARG7 = "test6.txt"
#ARG8 = "test7.txt"
#ARG9 = "test8.txt"
#ARGC = 9

set term pngcairo size 1024,768 enhanced font 'Verdana,10'
set output ARG1
set datafile separator ","

# title and line style definitions

data_title1 = "Custom Titles or File Names"
if (strlen(data_title1) == 0 ) data_title1 = ARG2
set style line 1 lc rgb "blue" lw 1

data_title2 = ""
if (strlen(data_title2) == 0 ) data_title2 = ARG3
set style line 2 lc rgb "forest-green" lw 1

data_title3 = ""
if (strlen(data_title3) == 0 ) data_title3 = ARG4
set style line 3 lc rgb "magenta" lw 1

data_title4 = ""
if (strlen(data_title4) == 0 ) data_title4 = ARG5
set style line 4 lc rgb "cyan" lw 1

data_title5 = ""
if (strlen(data_title5) == 0 ) data_title5 = ARG6
set style line 5 lc rgb "dark-violet" lw 1

data_title6 = ""
if (strlen(data_title6) == 0 ) data_title6 = ARG7
set style line 6 lc rgb "orange" lw 1

data_title7 = ""
if (strlen(data_title7) == 0 ) data_title7 = ARG8
set style line 7 lc rgb "light-green" lw 1

data_title8 = ""
if (strlen(data_title8) == 0 ) data_title8 = ARG9
set style line 8 lc rgb "bisque" lw 1

# border and grid line styles
set style line 9 lc rgb "black" lt 1 lw 1
set style line 10 lc rgb "black" lt 0 lw 1
 
# graph style
set border 3 front ls 9
set tics nomirror front
set grid back ls 10
set grid ytics ls 10

# legend top right corner
set key noauto at 690,95 font 'Verdana,13'

set multiplot layout 2,1 title "Temperature vs. Frequency" font 'Verdana,16'

# temperature plot
set size ratio 0 1,.8
set origin 0,0.2
set lmargin 11
# y axis
set ylabel 'Temperature (c)' font 'Verdana,12'
set yrange [30:100]
set ytics 30,2 border nomirror out
set mytics
# x axis
set xlabel 'Time (seconds)'
set xtics 0,30 border nomirror out
set mxtics

if ( ARGC == 2 ) { plot ARG2 using 1:2 with lines ls 1 axes x1y1 title data_title1 }
if ( ARGC == 3 ) { plot ARG2 using 1:2 with lines ls 1 axes x1y1 title data_title1, \
					ARG3 using 1:2 with lines ls 2 axes x1y1 title data_title2 }
if ( ARGC == 4 ) { plot ARG2 using 1:2 with lines ls 1 axes x1y1 title data_title1, \
					ARG3 using 1:2 with lines ls 2 axes x1y1 title data_title2, \
					ARG4 using 1:2 with lines ls 3 axes x1y1 title data_title3 }
if ( ARGC == 5 ) { plot ARG2 using 1:2 with lines ls 1 axes x1y1 title data_title1, \
					ARG3 using 1:2 with lines ls 2 axes x1y1 title data_title2, \
					ARG4 using 1:2 with lines ls 3 axes x1y1 title data_title3, \
					ARG5 using 1:2 with lines ls 4 axes x1y1 title data_title4 }
if ( ARGC == 6 ) { plot ARG2 using 1:2 with lines ls 1 axes x1y1 title data_title1, \
					ARG3 using 1:2 with lines ls 2 axes x1y1 title data_title2, \
					ARG4 using 1:2 with lines ls 3 axes x1y1 title data_title3, \
					ARG5 using 1:2 with lines ls 4 axes x1y1 title data_title4, \
					ARG6 using 1:2 with lines ls 5 axes x1y1 title data_title5 }
if ( ARGC == 7 ) { plot ARG2 using 1:2 with lines ls 1 axes x1y1 title data_title1, \
					ARG3 using 1:2 with lines ls 2 axes x1y1 title data_title2, \
					ARG4 using 1:2 with lines ls 3 axes x1y1 title data_title3, \
					ARG5 using 1:2 with lines ls 4 axes x1y1 title data_title4, \
					ARG6 using 1:2 with lines ls 5 axes x1y1 title data_title5, \
					ARG7 using 1:2 with lines ls 6 axes x1y1 title data_title6 }
if ( ARGC == 8 ) { plot ARG2 using 1:2 with lines ls 1 axes x1y1 title data_title1, \
					ARG3 using 1:2 with lines ls 2 axes x1y1 title data_title2, \
					ARG4 using 1:2 with lines ls 3 axes x1y1 title data_title3, \
					ARG5 using 1:2 with lines ls 4 axes x1y1 title data_title4, \
					ARG6 using 1:2 with lines ls 5 axes x1y1 title data_title5, \
					ARG7 using 1:2 with lines ls 6 axes x1y1 title data_title6, \
					ARG8 using 1:2 with lines ls 7 axes x1y1 title data_title7 }
if ( ARGC == 9 ) { plot ARG2 using 1:2 with lines ls 1 axes x1y1 title data_title1, \
					ARG3 using 1:2 with lines ls 2 axes x1y1 title data_title2, \
					ARG4 using 1:2 with lines ls 3 axes x1y1 title data_title3, \
					ARG5 using 1:2 with lines ls 4 axes x1y1 title data_title4, \
					ARG6 using 1:2 with lines ls 5 axes x1y1 title data_title5, \
					ARG7 using 1:2 with lines ls 6 axes x1y1 title data_title6, \
					ARG8 using 1:2 with lines ls 7 axes x1y1 title data_title7, \
					ARG9 using 1:2 with lines ls 8 axes x1y1 title data_title8 }

# frequency plot
set size ratio 0 1,.2
set origin 0,0
set lmargin 11
# y axis
set ylabel 'Frequency (ghz)' font 'Verdana,12'
set yrange [0:2]
set ytics 0,.5 border nomirror out
set format y '%.1f'
# x axis
set noxlabel

if ( ARGC == 2 ) { plot ARG2 using 1:($3/1000.0) with lines ls 1 axes x1y2 notitle, ARG2 using 1:4 with lines ls 1 axes \
					x1y2, ARG2 using 1:5 with lines ls 1 axes x1y2, ARG2 using 1:6 with lines ls 1 axes x1y2 }
if ( ARGC == 3 ) { plot ARG2 using 1:($3/1000.0) with lines ls 1 axes x1y2 notitle, ARG2 using 1:4 with lines ls 1 axes \
					x1y2, ARG2 using 1:5 with lines ls 1 axes x1y2, ARG2 using 1:6 with lines ls 1 axes x1y2, \
					ARG3 using 1:3 with lines ls 2 axes x1y2 notitle, ARG3 using 1:4 with lines ls 2 axes x1y2, ARG3 using \
					1:5 with lines ls 2 axes x1y2, ARG3 using 1:6 with lines ls 2 axes x1y2 }
if ( ARGC == 4 ) { plot ARG2 using 1:($3/1000.0) with lines ls 1 axes x1y2 notitle, ARG2 using 1:4 with lines ls 1 axes \
					x1y2, ARG2 using 1:5 with lines ls 1 axes x1y2, ARG2 using 1:6 with lines ls 1 axes x1y2, \
					ARG3 using 1:3 with lines ls 2 axes x1y2 notitle, ARG3 using 1:4 with lines ls 2 axes x1y2, ARG3 using 1:5 with lines ls 2 axes x1y2, ARG3 using 1:6 with lines ls 2 axes x1y2, \
					ARG4 using 1:3 with lines ls 3 axes x1y2 notitle, ARG4 using 1:4 with lines ls 3 axes x1y2, ARG4 using 1:5 with lines ls 3 axes x1y2, ARG4 using 1:6 with lines ls 3 axes x1y2 }
if ( ARGC == 5 ) { plot ARG2 using 1:($3/1000.0) with lines ls 1 axes x1y2 notitle, ARG2 using 1:4 with lines ls 1 axes \
					x1y2, ARG2 using 1:5 with lines ls 1 axes x1y2, ARG2 using 1:6 with lines ls 1 axes x1y2, \
					ARG3 using 1:3 with lines ls 2 axes x1y2 notitle, ARG3 using 1:4 with lines ls 2 axes x1y2, ARG3 using 1:5 with lines ls 2 axes x1y2, ARG3 using 1:6 with lines ls 2 axes x1y2, \
					ARG4 using 1:3 with lines ls 3 axes x1y2 notitle, ARG4 using 1:4 with lines ls 3 axes x1y2, ARG4 using 1:5 with lines ls 3 axes x1y2, ARG4 using 1:6 with lines ls 3 axes x1y2, \
					ARG5 using 1:3 with lines ls 4 axes x1y2 notitle, ARG5 using 1:4 with lines ls 4 axes x1y2, ARG5 using 1:5 with lines ls 4 axes x1y2, ARG5 using 1:6 with lines ls 4 axes x1y2 }
if ( ARGC == 6 ) { plot ARG2 using 1:($3/1000.0) with lines ls 1 axes x1y2 notitle, ARG2 using 1:4 with lines ls 1 axes \
					x1y2, ARG2 using 1:5 with lines ls 1 axes x1y2, ARG2 using 1:6 with lines ls 1 axes x1y2, \
					ARG3 using 1:3 with lines ls 2 axes x1y2 notitle, ARG3 using 1:4 with lines ls 2 axes x1y2, ARG3 using 1:5 with lines ls 2 axes x1y2, ARG3 using 1:6 with lines ls 2 axes x1y2, \
					ARG4 using 1:3 with lines ls 3 axes x1y2 notitle, ARG4 using 1:4 with lines ls 3 axes x1y2, ARG4 using 1:5 with lines ls 3 axes x1y2, ARG4 using 1:6 with lines ls 3 axes x1y2, \
					ARG5 using 1:3 with lines ls 4 axes x1y2 notitle, ARG5 using 1:4 with lines ls 4 axes x1y2, ARG5 using 1:5 with lines ls 4 axes x1y2, ARG5 using 1:6 with lines ls 4 axes x1y2, \
					ARG6 using 1:3 with lines ls 5 axes x1y2 notitle, ARG6 using 1:4 with lines ls 5 axes x1y2, ARG6 using 1:5 with lines ls 5 axes x1y2, ARG6 using 1:6 with lines ls 5 axes x1y2 }
if ( ARGC == 7 ) { plot ARG2 using 1:($3/1000.0) with lines ls 1 axes x1y2 notitle, ARG2 using 1:4 with lines ls 1 axes \
					x1y2, ARG2 using 1:5 with lines ls 1 axes x1y2, ARG2 using 1:6 with lines ls 1 axes x1y2, \
					ARG3 using 1:3 with lines ls 2 axes x1y2 notitle, ARG3 using 1:4 with lines ls 2 axes x1y2, ARG3 using 1:5 with lines ls 2 axes x1y2, ARG3 using 1:6 with lines ls 2 axes x1y2, \
					ARG4 using 1:3 with lines ls 3 axes x1y2 notitle, ARG4 using 1:4 with lines ls 3 axes x1y2, ARG4 using 1:5 with lines ls 3 axes x1y2, ARG4 using 1:6 with lines ls 3 axes x1y2, \
					ARG5 using 1:3 with lines ls 4 axes x1y2 notitle, ARG5 using 1:4 with lines ls 4 axes x1y2, ARG5 using 1:5 with lines ls 4 axes x1y2, ARG5 using 1:6 with lines ls 4 axes x1y2, \
					ARG6 using 1:3 with lines ls 5 axes x1y2 notitle, ARG6 using 1:4 with lines ls 5 axes x1y2, ARG6 using 1:5 with lines ls 5 axes x1y2, ARG6 using 1:6 with lines ls 5 axes x1y2, \
					ARG7 using 1:3 with lines ls 6 axes x1y2 notitle, ARG7 using 1:4 with lines ls 6 axes x1y2, ARG7 using 1:5 with lines ls 6 axes x1y2, ARG7 using 1:6 with lines ls 6 axes x1y2 }
if ( ARGC == 8 ) { plot ARG2 using 1:($3/1000.0) with lines ls 1 axes x1y2 notitle, ARG2 using 1:4 with lines ls 1 axes \
					x1y2, ARG2 using 1:5 with lines ls 1 axes x1y2, ARG2 using 1:6 with lines ls 1 axes x1y2, \
					ARG3 using 1:3 with lines ls 2 axes x1y2 notitle, ARG3 using 1:4 with lines ls 2 axes x1y2, ARG3 using 1:5 with lines ls 2 axes x1y2, ARG3 using 1:6 with lines ls 2 axes x1y2, \
					ARG4 using 1:3 with lines ls 3 axes x1y2 notitle, ARG4 using 1:4 with lines ls 3 axes x1y2, ARG4 using 1:5 with lines ls 3 axes x1y2, ARG4 using 1:6 with lines ls 3 axes x1y2, \
					ARG5 using 1:3 with lines ls 4 axes x1y2 notitle, ARG5 using 1:4 with lines ls 4 axes x1y2, ARG5 using 1:5 with lines ls 4 axes x1y2, ARG5 using 1:6 with lines ls 4 axes x1y2, \
					ARG6 using 1:3 with lines ls 5 axes x1y2 notitle, ARG6 using 1:4 with lines ls 5 axes x1y2, ARG6 using 1:5 with lines ls 5 axes x1y2, ARG6 using 1:6 with lines ls 5 axes x1y2, \
					ARG7 using 1:3 with lines ls 6 axes x1y2 notitle, ARG7 using 1:4 with lines ls 6 axes x1y2, ARG7 using 1:5 with lines ls 6 axes x1y2, ARG7 using 1:6 with lines ls 6 axes x1y2, \
					ARG8 using 1:3 with lines ls 7 axes x1y2 notitle, ARG8 using 1:4 with lines ls 7 axes x1y2, ARG8 using 1:5 with lines ls 7 axes x1y2, ARG8 using 1:6 with lines ls 7 axes x1y2 }
if ( ARGC == 9 ) { plot ARG2 using 1:($3/1000.0) with lines ls 1 axes x1y2 notitle, ARG2 using 1:4 with lines ls 1 axes \
					x1y2, ARG2 using 1:5 with lines ls 1 axes x1y2, ARG2 using 1:6 with lines ls 1 axes x1y2, \
					ARG3 using 1:3 with lines ls 2 axes x1y2 notitle, ARG3 using 1:4 with lines ls 2 axes x1y2, ARG3 using 1:5 with lines ls 2 axes x1y2, ARG3 using 1:6 with lines ls 2 axes x1y2, \
					ARG4 using 1:3 with lines ls 3 axes x1y2 notitle, ARG4 using 1:4 with lines ls 3 axes x1y2, ARG4 using 1:5 with lines ls 3 axes x1y2, ARG4 using 1:6 with lines ls 3 axes x1y2, \
					ARG5 using 1:3 with lines ls 4 axes x1y2 notitle, ARG5 using 1:4 with lines ls 4 axes x1y2, ARG5 using 1:5 with lines ls 4 axes x1y2, ARG5 using 1:6 with lines ls 4 axes x1y2, \
					ARG6 using 1:3 with lines ls 5 axes x1y2 notitle, ARG6 using 1:4 with lines ls 5 axes x1y2, ARG6 using 1:5 with lines ls 5 axes x1y2, ARG6 using 1:6 with lines ls 5 axes x1y2, \
					ARG7 using 1:3 with lines ls 6 axes x1y2 notitle, ARG7 using 1:4 with lines ls 6 axes x1y2, ARG7 using 1:5 with lines ls 6 axes x1y2, ARG7 using 1:6 with lines ls 6 axes x1y2, \
					ARG8 using 1:3 with lines ls 7 axes x1y2 notitle, ARG8 using 1:4 with lines ls 7 axes x1y2, ARG8 using 1:5 with lines ls 7 axes x1y2, ARG8 using 1:6 with lines ls 7 axes x1y2, \
					ARG9 using 1:3 with lines ls 8 axes x1y2 notitle, ARG9 using 1:4 with lines ls 8 axes x1y2, ARG9 using 1:5 with lines ls 8 axes x1y2, ARG9 using 1:6 with lines ls 8 axes x1y2 }
unset multiplot
multifile_temp.gpl will plot up to 8 test files from watchtemp.sh or a similar data source on the same chart with the expected data formatted as "count,temp,freq1,freq2,freq3,freq4". If a legend title is not specified the name of the file is used by default.

GNUPlot version 5.x use: gnuplot -c multifile_temp.gpl chart.png testfile1...testfile8
GNUPlot version 4.x use: gnuplot multifile_temp.gpl
multifile_tempchart_s.png
multifile_tempchart_s.png (78.33 KiB) Viewed 6995 times
multicore_temp.gpl

Code: Select all

#!/usr/bin/gnuplot --persist
#   
#
# Decription: GNUPLOT script file that plots temperature and 
#             frequency for all cores from a watchtemp_all.sh data file
#             formated as "count,temp,freq1,freq2,freq3,freq4,core1,core2,core3,core4,gpu"
#
# Version 5.x
# Use: gnuplot multicore_temp.gpl outputchart.png file1
#
# Version 4.x
# Use: gnuplot multicore_temp.gpl
#
# Author: @hominoid www.forum.odroid.com
# Date: 5/3/2017
# Released to the public domain with no warranty of any kind
#

# for version 4.x only, uncomment and enter file names instead of on the commandline, see use above
#ARG1 = "multicore_tempchart.png"
#ARG2 = "test1.txt"

set term pngcairo size 1024,768 enhanced font 'Verdana,10'
set output ARG1
set datafile separator ","

# title and line style definitions

data_title1 = "System Temp"
set style line 1 lc rgb "red" lw 1

data_title2 = "Core 4 Temp"
set style line 2 lc rgb "blue" lw 1

data_title3 = "Core 5 Temp"
set style line 3 lc rgb "forest-green" lw 1

data_title4 = "Core 6 Temp"
set style line 4 lc rgb "cyan" lw 1

data_title5 = "Core 7 Temp"
set style line 5 lc rgb "dark-violet" lw 1

data_title6 = "GPU Temp"
set style line 6 lc rgb "magenta" lw 1

# border and grid line styles
set style line 9 lc rgb "black" lt 1 lw 1
set style line 10 lc rgb "black" lt 0 lw 1
 
# graph style
set border 3 front ls 9
set tics nomirror
set grid back ls 10
set grid ytics ls 10

# legend top right corner
set key noauto at 690,95 font 'Verdana,13'

set multiplot layout 2,1 title "CPU Core Temperatures vs. Frequency" font 'Verdana,16'

# temperature plot
set size ratio 0 1,.8
set origin 0,0.2
set lmargin 11
# y axis
set ylabel 'Temperature (c)' font 'Verdana,12'
set yrange [30:100]
set ytics 30,2 border nomirror out
set mytics
# x axis
set xlabel 'Time (seconds)'
set xtics 0,30 border nomirror out
set mxtics

plot ARG2 using 1:2 with lines ls 1 axes x1y1 title data_title1, ARG2 using 1:7 with lines ls 2 axes x1y1 title data_title2, ARG2 using 1:8 with lines ls 3 axes x1y1 title data_title3, ARG2 using 1:9 with lines ls 4 axes x1y1 title data_title4, ARG2 using 1:10 with lines ls 5 axes x1y1 title data_title5, ARG2 using 1:11 with lines ls 6 axes x1y1 title data_title6

# frequency plot
set size ratio 0 1,.2
set origin 0,0
set lmargin 11
# y axis
set ylabel 'Frequency (ghz)' font 'Verdana,12'
set yrange [0:2]
set ytics 0,.5 border nomirror out
set format y '%.1f'
# x axis
set noxlabel

plot ARG2 using 1:($3/1000.0) with lines ls 2 axes x1y2 notitle, ARG2 using 1:4 with lines ls 3 axes \
					x1y2, ARG2 using 1:5 with lines ls 4 axes x1y2, ARG2 using 1:6 with lines ls 5 axes x1y2

unset multiplot
multicore_temp.gpl will plot system temperature, individual A15 core temperatures, gpu temperature and frequency for a single test file from watchtemp_all.sh or a similiar data source with the expected data formatted as "count,temp,freq4,freq5,freq6,freq7,core4,core5,core6,core7,gpu".

GNUPlot version 5.x use: gnuplot -c multicore_temp.gpl chart.png testfile
GNUPlot version 4.x use: gnuplot multicore_temp.gpl
multicore_tempchart_s.png
multicore_tempchart_s.png (72.4 KiB) Viewed 6995 times
As stated previously, for testing I use "./watchtemp.sh | tee testfile.txt" in one shell and the "sysbench --test=cpu --cpu-max-prime=100000 --num-threads=8 run" in another.

Just so everything is available in one location these are the versions of the scripts I have used.
watchtemp.sh

Code: Select all

#!/bin/bash
z=0
echo "Time, Temp, FreqCPU4, FreqCPU5, FreqCPU6, FreqCPU7 "
while true :
do
     t=`cat /sys/devices/virtual/thermal/thermal_zone0/temp`
     t1=$(( $t/1000))
     fa=`cat /sys/devices/system/cpu/cpu4/cpufreq/scaling_cur_freq`
     fb=`cat /sys/devices/system/cpu/cpu5/cpufreq/scaling_cur_freq`
     fc=`cat /sys/devices/system/cpu/cpu6/cpufreq/scaling_cur_freq`
     fd=`cat /sys/devices/system/cpu/cpu7/cpufreq/scaling_cur_freq`
     echo $z, $t1, $fa, $fb, $fc, $fd
     sleep 2
     (( z += 2 ))
done
watchtemp_all.sh

Code: Select all

#!/bin/bash
z=0
echo "Time, Temp, FreqCPU4, FreqCPU5, FreqCPU6, FreqCPU7, sen0, sen1, sen2, sen3, sen4 "
while true :
do
     t=`cat /sys/devices/virtual/thermal/thermal_zone0/temp`
     t1=$(( $t/1000))
     fa=`cat /sys/devices/system/cpu/cpu4/cpufreq/scaling_cur_freq`
     fb=`cat /sys/devices/system/cpu/cpu5/cpufreq/scaling_cur_freq`
     fc=`cat /sys/devices/system/cpu/cpu6/cpufreq/scaling_cur_freq`
     fd=`cat /sys/devices/system/cpu/cpu7/cpufreq/scaling_cur_freq`
     s0=`cat /sys/devices/10060000.tmu/temp | grep "sensor0 : " | awk '{print $3}'`
     s0t=$(( $s0/1000))
     s1=`cat /sys/devices/10060000.tmu/temp | grep "sensor1 : " | awk '{print $3}'`
     s1t=$(( $s1/1000))
     s2=`cat /sys/devices/10060000.tmu/temp | grep "sensor2 : " | awk '{print $3}'`
     s2t=$(( $s2/1000))
     s3=`cat /sys/devices/10060000.tmu/temp | grep "sensor3 : " | awk '{print $3}'`
     s3t=$(( $s3/1000))
     s4=`cat /sys/devices/10060000.tmu/temp | grep "sensor4 : " | awk '{print $3}'`
     s4t=$(( $s4/1000))
     echo $z, $t1, $fa, $fb, $fc, $fd, $s0t, $s1t, $s2t, $s3t, $s4t
     sleep 2
     (( z += 2 ))
done
I hope the GNUPlot scripts are useful and become part of a standardize way of collecting and displaying test results. I believe there still are some issues on the collection side to be worked out. The script latency in recorded time and it's skew in relationship to the sysbench results can be improved. And, a standard means to deal with the difference between operating systems and kernels needs to be considered.

Vort
Posts: 153
Joined: Fri Nov 22, 2013 10:04 pm
languages_spoken: english
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by Vort » Sun May 07, 2017 4:00 pm

Just for fun :D I ran watchtemp.sh on my U3 under Ubuntu Server OS and tor running in the background with 5% CPU load. Did that in the RAM e.g. in the /var/run/shm directory.
I modified the script to the right CPU numbers and for reading temperature every second, and renamed it.
Passive cooling here.

It works ;)

Code: Select all

$ uname -a
Linux "hostname" 3.8.13.30 #1 SMP PREEMPT Fri Sep 4 23:45:57 BRT 2015 armv7l armv7l armv7l GNU/Linux

$ ./Wt.sh | tee testfile.txt   
Time, Temp, FreqCPU0, FreqCPU1, FreqCPU2, FreqCPU3 
0, 35, 1704000, 1704000, 1704000, 1704000
1, 34, 1704000, 1704000, 1704000, 1704000
2, 35, 1704000, 1704000, 1704000, 1704000
3, 34, 1704000, 1704000, 1704000, 1704000
4, 35, 1704000, 1704000, 1704000, 1704000
5, 35, 1704000, 1704000, 1704000, 1704000
6, 34, 1704000, 1704000, 1704000, 1704000
7, 35, 1704000, 1704000, 1704000, 1704000
8, 34, 1704000, 1704000, 1704000, 1704000
9, 35, 1704000, 1704000, 1704000, 1704000
10, 35, 1704000, 1704000, 1704000, 1704000
11, 35, 1704000, 1704000, 1704000, 1704000
12, 35, 1704000, 1704000, 1704000, 1704000
13, 35, 1704000, 1704000, 1704000, 1704000
14, 35, 1704000, 1704000, 1704000, 1704000
15, 35, 1704000, 1704000, 1704000, 1704000
16, 35, 1704000, 1704000, 1704000, 1704000
17, 35, 1704000, 1704000, 1704000, 1704000
18, 35, 1704000, 1704000, 1704000, 1704000
19, 35, 1704000, 1704000, 1704000, 1704000
20, 35, 1704000, 1704000, 1704000, 1704000
21, 35, 1704000, 1704000, 1704000, 1704000
22, 35, 1704000, 1704000, 1704000, 1704000
23, 35, 1704000, 1704000, 1704000, 1704000
24, 35, 1704000, 1704000, 1704000, 1704000
25, 35, 1704000, 1704000, 1704000, 1704000
26, 35, 1704000, 1704000, 1704000, 1704000
27, 35, 1704000, 1704000, 1704000, 1704000
28, 35, 1704000, 1704000, 1704000, 1704000
29, 35, 1704000, 1704000, 1704000, 1704000
30, 34, 1704000, 1704000, 1704000, 1704000
31, 34, 1704000, 1704000, 1704000, 1704000
32, 35, 1704000, 1704000, 1704000, 1704000
33, 35, 1704000, 1704000, 1704000, 1704000
34, 35, 1704000, 1704000, 1704000, 1704000
35, 35, 1704000, 1704000, 1704000, 1704000
36, 35, 1704000, 1704000, 1704000, 1704000
37, 35, 1704000, 1704000, 1704000, 1704000
38, 35, 1704000, 1704000, 1704000, 1704000
39, 35, 1704000, 1704000, 1704000, 1704000
40, 35, 1704000, 1704000, 1704000, 1704000
^C

$ cat testfile.txt
Time, Temp, FreqCPU0, FreqCPU1, FreqCPU2, FreqCPU3 
0, 35, 1704000, 1704000, 1704000, 1704000
1, 34, 1704000, 1704000, 1704000, 1704000
2, 35, 1704000, 1704000, 1704000, 1704000
3, 34, 1704000, 1704000, 1704000, 1704000
4, 35, 1704000, 1704000, 1704000, 1704000
5, 35, 1704000, 1704000, 1704000, 1704000
6, 34, 1704000, 1704000, 1704000, 1704000
7, 35, 1704000, 1704000, 1704000, 1704000
8, 34, 1704000, 1704000, 1704000, 1704000
9, 35, 1704000, 1704000, 1704000, 1704000
10, 35, 1704000, 1704000, 1704000, 1704000
11, 35, 1704000, 1704000, 1704000, 1704000
12, 35, 1704000, 1704000, 1704000, 1704000
13, 35, 1704000, 1704000, 1704000, 1704000
14, 35, 1704000, 1704000, 1704000, 1704000
15, 35, 1704000, 1704000, 1704000, 1704000
16, 35, 1704000, 1704000, 1704000, 1704000
17, 35, 1704000, 1704000, 1704000, 1704000
18, 35, 1704000, 1704000, 1704000, 1704000
19, 35, 1704000, 1704000, 1704000, 1704000
20, 35, 1704000, 1704000, 1704000, 1704000
21, 35, 1704000, 1704000, 1704000, 1704000
22, 35, 1704000, 1704000, 1704000, 1704000
23, 35, 1704000, 1704000, 1704000, 1704000
24, 35, 1704000, 1704000, 1704000, 1704000
25, 35, 1704000, 1704000, 1704000, 1704000
26, 35, 1704000, 1704000, 1704000, 1704000
27, 35, 1704000, 1704000, 1704000, 1704000
28, 35, 1704000, 1704000, 1704000, 1704000
29, 35, 1704000, 1704000, 1704000, 1704000
30, 34, 1704000, 1704000, 1704000, 1704000
31, 34, 1704000, 1704000, 1704000, 1704000
32, 35, 1704000, 1704000, 1704000, 1704000
33, 35, 1704000, 1704000, 1704000, 1704000
34, 35, 1704000, 1704000, 1704000, 1704000
35, 35, 1704000, 1704000, 1704000, 1704000
36, 35, 1704000, 1704000, 1704000, 1704000
37, 35, 1704000, 1704000, 1704000, 1704000
38, 35, 1704000, 1704000, 1704000, 1704000
39, 35, 1704000, 1704000, 1704000, 1704000
40, 35, 1704000, 1704000, 1704000, 1704000

DarkBahamut
Posts: 332
Joined: Tue Jan 19, 2016 10:19 am
languages_spoken: english
ODROIDs: XU4, N1
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by DarkBahamut » Mon May 08, 2017 9:41 pm

I'm envious of the amazing temps. I need this setup for my XU4 :lol:

Given the passive results, on 3.10 I bet it could manage 2.1 or maybe 2.2GHz on 1-2 core loads, then settle at about 1.9GHz at 90C for long loads. Quite amazing for passive or very low noise cooling. Should keep us going until HK release a new shiny new high performance arm64 board ;)

User avatar
rooted
Posts: 7336
Joined: Fri Dec 19, 2014 9:12 am
languages_spoken: english
Location: Gulf of Mexico, US
Has thanked: 495 times
Been thanked: 137 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by rooted » Mon May 08, 2017 11:35 pm

DarkBahamut wrote:I'm envious of the amazing temps. I need this setup for my XU4 Image

Given the passive results, on 3.10 I bet it could manage 2.1 or maybe 2.2GHz on 1-2 core loads, then settle at about 1.9GHz at 90C for long loads. Quite amazing for passive or very low noise cooling. Should keep us going until HK release a new shiny new high performance arm64 board ;)
Have you tried exceeding 2ghz and it actually increasing single threaded performance?

DarkBahamut
Posts: 332
Joined: Tue Jan 19, 2016 10:19 am
languages_spoken: english
ODROIDs: XU4, N1
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by DarkBahamut » Tue May 09, 2017 5:11 am

Yeah. There's not much headroom in the SoC, but some. Performance scales as expected though. 2.1GHz is very usable on the stock cooler for low thread workloads.

sysbench --test=cpu --cpu-max-prime=20000

A15 2.0GHz = 231.90
A15 2.1GHz = 220.97 (4.7% faster)

sysbench --test=cpu --cpu-max-prime=20000 --num-threads=2

A15 2.0GHz = 116.13
A15 2.1GHz = 110.58 (4.8% faster)

I haven't tried 2.2GHz. It has limited use on the stock cooler but under single thread it might be usable, 2 threads would likely need to drop back to 2.1GHz sometimes. Voltages would be right up against the max limit so stability would be questionable on some boards.

User avatar
rooted
Posts: 7336
Joined: Fri Dec 19, 2014 9:12 am
languages_spoken: english
Location: Gulf of Mexico, US
Has thanked: 495 times
Been thanked: 137 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by rooted » Tue May 09, 2017 7:59 am

DarkBahamut wrote:Yeah. There's not much headroom in the SoC, but some. Performance scales as expected though. 2.1GHz is very usable on the stock cooler for low thread workloads.

sysbench --test=cpu --cpu-max-prime=20000

A15 2.0GHz = 231.90
A15 2.1GHz = 220.97 (4.7% faster)

sysbench --test=cpu --cpu-max-prime=20000 --num-threads=2

A15 2.0GHz = 116.13
A15 2.1GHz = 110.58 (4.8% faster)

I haven't tried 2.2GHz. It has limited use on the stock cooler but under single thread it might be usable, 2 threads would likely need to drop back to 2.1GHz sometimes. Voltages would be right up against the max limit so stability would be questionable on some boards.
Nice to see an increase in performance, if we could safely overclock and run all 4 big cores simultaneously it would be about a 17% bump in performance. Quite a lot really for only 400 mhz more.

I know it's super unlikely to use all 4 overclocked but still nice to think about.

User avatar
mad_ady
Posts: 7558
Joined: Wed Jul 15, 2015 5:00 pm
languages_spoken: english
ODROIDs: XU4, C1+, C2, N1, H2, N2, Go, Go Advance
Location: Bucharest, Romania
Has thanked: 391 times
Been thanked: 296 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by mad_ady » Tue May 09, 2017 1:18 pm

@DarkBahamut: In order to overclock do you need custom kernel patches, or can it be done by changing the dtb only?

DarkBahamut
Posts: 332
Joined: Tue Jan 19, 2016 10:19 am
languages_spoken: english
ODROIDs: XU4, N1
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by DarkBahamut » Tue May 23, 2017 10:51 am

@mad_ady Whoops, missed this one. In the simplest form you just need to edit the cpufreq driver files since the supporting code for most frequencies already exists. Just bump the max support down one level for each cluster to give you 1.5/2.1GHz. Those speeds work very well, but I don't recommend trying higher as you really need to play with the voltages at that point.

A15: https://github.com/Emerald-Phoenix/linu ... 8effb9059f
A7: https://github.com/Emerald-Phoenix/linu ... 8124dbe20a (Set L1 for 1.5GHz)

User avatar
mad_ady
Posts: 7558
Joined: Wed Jul 15, 2015 5:00 pm
languages_spoken: english
ODROIDs: XU4, C1+, C2, N1, H2, N2, Go, Go Advance
Location: Bucharest, Romania
Has thanked: 391 times
Been thanked: 296 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by mad_ady » Tue May 23, 2017 2:32 pm

Thank you, I'll bump up my A7, since in my setup they are doing most of the work.

phaseshifter
Posts: 3520
Joined: Fri May 08, 2015 9:12 am
languages_spoken: english
ODROIDs: U-2,U3+,,XU-3,,XU3-LITE,,XU-4
C1+,,C-2,,,
N-1,,N-2,...other odroid acc`s as well
Has thanked: 41 times
Been thanked: 43 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by phaseshifter » Tue May 23, 2017 3:05 pm

ok i see the git hub pages but dont know the command line files i have to modd could somebody pls put that in laymans term`s ..soz..i lack code exp...phase thanks...

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Sat Jun 03, 2017 11:13 am

(Super) Quiet Please
I have been exploring some ideas to lower the acoustic profile of the XU4 Split Airflow case. The current design allows for the docking of external air sources, thru an enclosed duct that mates with the twin upper and lower case ducts, by a simple adapter.
Split_AirFlow_80mm_Adapter_s.png
Split_AirFlow_80mm_Adapter_s.png (35.57 KiB) Viewed 6727 times
S1340007r.png
S1340007r.png (76.62 KiB) Viewed 6727 times
The Noctua 80mm fans have slightly lower acoustics with more than 6 times the airflow of the current Noctua 40mm fan used in the XU4 Split Airflow Case. The static pressure is better as well. It looks like the case acoustic signature can be significantly reduced while maintaining or possibly increasing thermal performance.

Code: Select all

Model				RPM	dB/A	m3/h	mm H2O	
NF-A4x10 5v     4500	17.9	8.2	 1.78		3 pin
NF-A8 PWM 12v	2200	17.7	55.5	2.37		4 pin PWM
NF-A8 FLX 12v	2000	16.1	50.4	1.96		3 pin
Some quick back of the cranial calculations; assuming the acoustic profiles are similar and scale accordingly, which they probably aren't and don't, the XU4 Split Airflow Case could run the NF-A8 FLX 80mm fan at 16.2% and produce the same airflow of the 40mm fan running at 100% with only 2.6 dB/A noise. Regardless of what the actual numbers are, most likely there will be significant noise reduction and better airflow.

Another question to consider is what method to control and power the fan. The choices are as follows:
  • · Manually Adjustable 3pin or PWM 4pin fan with separate 12v wall adapter.
    · XU4 GPIO PWM driven 3 or 4 pin fan with integrate system power (12v step down to 5v).
    · Use existing XU4 5v fan output to control 5v PWM on 12v fan.
    · 5volt 80mm fan.
They all have pro's and con's depending on the application and deployment. As time allows, I'll be looking at different configurations starting with a simple manually controlled fan run from a 12v wall adapter. It is a simple solution that should work well in certain applications and I happen to have everything already.

I have a NF-A8 PWM 12v fan on order and will look to integrate it as closely as possible. A new XU4 GPIO PWM implementation is the most complex and least desirable in my opinion. The most interesting solution would be to use the existing 5v fan output to control the PWM on the 12v fan. That way no other OS changes would be necessary and a simple wall adapter could be used to power just the fan.

I have seen one 5 volt 80mm fan that claimed to work on a RPI. It had about twice the airflow (16 m3/h) of the current 40mm fan and 22dba of noise. Not the best but it was cheap. I'll share the results of the configurations I evaluate when they are complete. Should be interesting for anyone wanting more cooling and less noise.

I have also been establishing a thermal test baseline of the OEM cooling solutions, active and passive, to use as comparison. This will put into better context the actual performance improvements of the current design and future changes.

Comments and suggestions welcomed.

phaseshifter
Posts: 3520
Joined: Fri May 08, 2015 9:12 am
languages_spoken: english
ODROIDs: U-2,U3+,,XU-3,,XU3-LITE,,XU-4
C1+,,C-2,,,
N-1,,N-2,...other odroid acc`s as well
Has thanked: 41 times
Been thanked: 43 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by phaseshifter » Sat Jun 03, 2017 12:05 pm

interesting concept..will amerdroid take this modd on also the slot where the noctua used to sit can be populated/covered..yes..??

it will allow for a variety of fans to choose from..!!!

funny though it`s starting to look like the Eiffel tower.. :D
Build It And They Will Come...Be Bold And Mighty Forces Will Come To Your Aid..!!!

tmihai20
Posts: 222
Joined: Mon Nov 07, 2016 10:56 pm
languages_spoken: english, french, italian, romanian
ODROIDs: XU4, Go, H2, N2 (4GB)
Location: Romania
Has thanked: 29 times
Been thanked: 8 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by tmihai20 » Thu Jun 08, 2017 10:20 pm

Does anybody have an STL for the case only? I downloaded both scad files, I generated the model, but it has the XU4 inside as well. I also got this warning after generation of the 3D model:

WARNING: Object may not be a valid 2-manifold and may need repair!
Rendering finished.
WARNING: Object may not be a valid 2-manifold and may need repair! See http://en.wikibooks.org/wiki/OpenSCAD_U ... and_Export
Riddle me this, riddle me that
Who is afraid of the big, black bat?
I write (in Romanian mostly) on a blog (see my profile)

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Fri Jun 09, 2017 12:08 am

Hi @tmihai20

A copy of the final stl files is located on the 2nd page of this thread towards the bottom before the copper shim/perch test results. It is labeled XU4_Split_AirFlow_Vertical_Case_v1_Final.zip and contains two stl's, one of the case and another for the accessories.

The warning your getting is normal in model view mode. You are viewing the assembled 3D model and therefore it will not be a valid manifold because it contains multiple individual pieces. To change to print mode review the configuration at the top of the .scad design file, it should look like this.

Code: Select all

// **Case Configuration** //
XU4_WIDTH = 106; // minimum width=83, case side fan min=102
XU4_DEPTH = 63; // minimum depth=59
XU4_THICK = 2; // case thickness
XU4_CS_OFFSET = 4;  // size in mm above min case size, 0 if min size
XU4_COUNTERSINK = 0; // 0=none, 1=countersink case screws

// **Standoff Configuration** //
XU4_BSO_HEIGHT = 8; // bottom standoff height default 8mm
XU4_TSO_HEIGHT = 28; // top standoff height default 28mm
XU4_STANDOFF_HOLESIZE = 1.8; //default 1.8mm

// **Options Configuration** //
XU4_HEATSINK_TYPE = 2; // 0=none, 1=Stock 13.5mm, 2=Gold 36mm 3=northbridge 40mm 4=40mm plate
XU4_HEATSINK_OPEN = 1; // 0=none, 1=normal
XU4_EXT_FAN = 1; // 0=none, 1=case fan and mount; set XU4_WIDTH=102
XU4_DUCT = 1; // 0=none, 1=duct, top and bottom
XU4_DUCT_HEIGHT = 28; // usually the same as XU4_TSO_HEIGHT except for short top standoffs
XU4_UART = 1; // 0=none, 1=UART on ethernet side, 2=UART on USB3 side
XU4_NUT = 1; // M3 nut holder on heatsink slides (accessories only)

// **Display Mode** //
XU4_MODEL = 1; // 1=model mode, 2=print mode, 3=accessories only
The last line, "XU4_MODEL=1" sets the display mode to model mode (assembled). Change it to 2 for print mode and a proper layout for stl generation and printing. The accessories can be viewed and an stl generated by selecting 3. Both of these can be combined on a single platter if needed for printing. The stl's available on the forum have been community designed for a 24mm stand-off by request. You can either generate the design with fully built in standoff's or change the parameter "XU4_TSO_HEIGHT = 28;" to a height that will accommodate the standoff size you'll be using. The picture located up a couple of posts with the 80mm adapter is on a single 20mm standoff which are generally readily available.

As you can see, the case is parametric and can generate other case configurations from the same design file. I don't do any collision checking so make sure to view any changes in model mode for issues. If you change the height yourself know this, the total height is 41mm with the XU4 installed. 2mm top thickness, 2mm bottom thickness, 8mm bottom duct/standoff height, 28mm top duct/standoff height, 1mm PCB=41mm. So to use 20mm top standoff's, the top printed standoff stub needs to be 8mm (8mm+20mm=28mm). so, XU4_TSO_HEIGHT = 8 would be the correct setting for a 20mm standoff.

If you change the configuration and want me to take a quick look to see if I can spot any problems in the stl's before you submit them to your printer, let me know. You can post them here and send me a PM. Likewise, if your not sure and would like me to generate a set of stl's for you based on a different standoff size, I'm willing to do that as well, it only takes me a few minutes. FYI, the 80mm fan adapter is not complete nor a part of the current design file.

tmihai20
Posts: 222
Joined: Mon Nov 07, 2016 10:56 pm
languages_spoken: english, french, italian, romanian
ODROIDs: XU4, Go, H2, N2 (4GB)
Location: Romania
Has thanked: 29 times
Been thanked: 8 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by tmihai20 » Fri Jun 09, 2017 12:16 am

Thank you so much for this. This is the perfect case I had in mind for my XU4 (one that can accommodate a fan). I am using a 120 mm fan right now, I will look for the most silent 40 mm fan I can find.
Riddle me this, riddle me that
Who is afraid of the big, black bat?
I write (in Romanian mostly) on a blog (see my profile)

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Fri Jun 09, 2017 12:21 am

tmihai20 wrote:Thank you so much for this. This is the perfect case I had in mind for my XU4 (one that can accommodate a fan). I am using a 120 mm fan right now, I will look for the most silent 40 mm fan I can find.
That most likely will be the Noctua A4. Let me know if you find something quieter.

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Fri Jun 09, 2017 10:38 am

Here is a STL with the Split Airflow Case and Accessories on a single platter. The case is setup to use a 20mm standoff and 3mm recessed screws. It has both the spring pin and M3-nut slides for the heatsink, the UART strap and 4 feet/nut holders. It is suitable for direct submission to a printer.
XU4_Split_AirFlow_Case_v1_20mm_standoffs.zip
(234.33 KiB) Downloaded 72 times
XU4_Split_AirFlow_Case_v1_20mm_standoffs.jpg
XU4_Split_AirFlow_Case_v1_20mm_standoffs.jpg (72.97 KiB) Viewed 6629 times
**EDIT**
I have created dxf and svg files of the outline of the slides I use to re-orientate the Gold Heatsink for the Split Airflow Case. These slides will rotate and center the thick part of the heatsink directly above the SOC. The outlines can be used to locate the hole and offset and can be easily hand fabricated out of any non-ferrous metal. 14ga copper(1.628mm) fits the heatsink slide opening perfectly and will help conduct heat past the slide openings. LibreCad and Inkscape can be used to view and print.
Gold Heatsink Offset Slides.zip
(2.02 KiB) Downloaded 60 times
Gold Heatsink Offset Slides.png
Gold Heatsink Offset Slides.png (2.54 KiB) Viewed 6598 times

User avatar
mad_ady
Posts: 7558
Joined: Wed Jul 15, 2015 5:00 pm
languages_spoken: english
ODROIDs: XU4, C1+, C2, N1, H2, N2, Go, Go Advance
Location: Bucharest, Romania
Has thanked: 391 times
Been thanked: 296 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by mad_ady » Sun Jun 18, 2017 4:31 am

DarkBahamut wrote:@mad_ady Whoops, missed this one. In the simplest form you just need to edit the cpufreq driver files since the supporting code for most frequencies already exists. Just bump the max support down one level for each cluster to give you 1.5/2.1GHz. Those speeds work very well, but I don't recommend trying higher as you really need to play with the voltages at that point.

A15: https://github.com/Emerald-Phoenix/linu ... 8effb9059f
A7: https://github.com/Emerald-Phoenix/linu ... 8124dbe20a (Set L1 for 1.5GHz)
Sorry for the offtopic, but those patches you mentioned are for kernel 3.10. The file doesn't exist on 4.9. Do you have some patches for 4.9 for overclocking?

User avatar
odroid
Site Admin
Posts: 33649
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English, Korean, Japanese
ODROIDs: ODROID
Has thanked: 500 times
Been thanked: 504 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by odroid » Mon Jun 19, 2017 5:58 pm

@mad_ady,
This patch was tested on XU4 Kernel 4.9. But it might be useless if you don't have a nice cooling system.

Code: Select all

diff --git a/arch/arm/boot/dts/exynos5422-cpus.dtsi b/arch/arm/boot/dts/exynos5422-cpus.dtsi
index 1ac1263..5adb7e0 100644
--- a/arch/arm/boot/dts/exynos5422-cpus.dtsi
+++ b/arch/arm/boot/dts/exynos5422-cpus.dtsi
@@ -78,7 +78,7 @@
                       compatible = "arm,cortex-a15";
                       clocks = <&clock CLK_ARM_CLK>;
                       reg = <0x0>;
-                       clock-frequency = <2000000000>;
+                       clock-frequency = <2100000000>;
                       cci-control-port = <&cci_control1>;
                       operating-points-v2 = <&cluster_a15_opp_table>;
                       cooling-min-level = <0>;
@@ -90,7 +90,7 @@
                       device_type = "cpu";
                       compatible = "arm,cortex-a15";
                       reg = <0x1>;
-                       clock-frequency = <2000000000>;
+                       clock-frequency = <2100000000>;
                       cci-control-port = <&cci_control1>;
                       operating-points-v2 = <&cluster_a15_opp_table>;
                       cooling-min-level = <0>;
@@ -102,7 +102,7 @@
                       device_type = "cpu";
                       compatible = "arm,cortex-a15";
                       reg = <0x2>;
-                       clock-frequency = <2000000000>;
+                       clock-frequency = <2100000000>;
                       cci-control-port = <&cci_control1>;
                       operating-points-v2 = <&cluster_a15_opp_table>;
                       cooling-min-level = <0>;
@@ -114,7 +114,7 @@
                       device_type = "cpu";
                       compatible = "arm,cortex-a15";
                       reg = <0x3>;
-                       clock-frequency = <2000000000>;
+                       clock-frequency = <2100000000>;
                       cci-control-port = <&cci_control1>;
                       operating-points-v2 = <&cluster_a15_opp_table>;
                       cooling-min-level = <0>;
diff --git a/arch/arm/boot/dts/exynos5800.dtsi b/arch/arm/boot/dts/exynos5800.dtsi
index 5db6fa3..693f260 100755
--- a/arch/arm/boot/dts/exynos5800.dtsi
+++ b/arch/arm/boot/dts/exynos5800.dtsi
@@ -24,6 +24,12 @@
};

&cluster_a15_opp_table {
+       opp@2100000000 {
+               opp-hz = /bits/ 64 <2100000000>;
+               opp-microvolt = <1312500>;
+               clock-latency-ns = <140000>;
+       };
+
       opp@2000000000 {
               opp-hz = /bits/ 64 <2000000000>;
               opp-microvolt = <1312500>;

User avatar
mad_ady
Posts: 7558
Joined: Wed Jul 15, 2015 5:00 pm
languages_spoken: english
ODROIDs: XU4, C1+, C2, N1, H2, N2, Go, Go Advance
Location: Bucharest, Romania
Has thanked: 391 times
Been thanked: 296 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by mad_ady » Mon Jun 19, 2017 7:33 pm

Thanks odroid! But you're overclocking the big cores. I only want to add 100MHz to the little cores. I'll see if your patch could work on the little cores as well.
Also, it looks like a dtb change - I should be able to do this without recompiling the kernel, right?

User avatar
odroid
Site Admin
Posts: 33649
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English, Korean, Japanese
ODROIDs: ODROID
Has thanked: 500 times
Been thanked: 504 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by odroid » Mon Jun 19, 2017 9:44 pm

Right. You don't need to change the Kernel source code.
I think there should be another clock table for the LITTLE cores in the dts file.

User avatar
mad_ady
Posts: 7558
Joined: Wed Jul 15, 2015 5:00 pm
languages_spoken: english
ODROIDs: XU4, C1+, C2, N1, H2, N2, Go, Go Advance
Location: Bucharest, Romania
Has thanked: 391 times
Been thanked: 296 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by mad_ady » Tue Jun 20, 2017 8:38 pm

Ok, I started looking into the dtb, but the frequency is listed differently for little cores than for big cores.
For big cores I get:

Code: Select all

# fdtget /media/boot/exynos5422-odroidxu4.dtb /cpus/cpu@0 compatible
arm,cortex-a15

# fdtget /media/boot/exynos5422-odroidxu4.dtb /cpus/cpu@0 clock-frequency
2000000000
While for the little cores the frequency is SrN - which I can't interpret:

Code: Select all

# fdtget /media/boot/exynos5422-odroidxu4.dtb /cpus/cpu@100 compatible
arm,cortex-a7
# fdtget /media/boot/exynos5422-odroidxu4.dtb /cpus/cpu@100 clock-frequency
SrN

phaseshifter
Posts: 3520
Joined: Fri May 08, 2015 9:12 am
languages_spoken: english
ODROIDs: U-2,U3+,,XU-3,,XU3-LITE,,XU-4
C1+,,C-2,,,
N-1,,N-2,...other odroid acc`s as well
Has thanked: 41 times
Been thanked: 43 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by phaseshifter » Wed Jun 28, 2017 2:35 pm

some pics of what i did..be known it is no better than the stock cooler having said this the heat extractor pipes are far too long.
.its cool but not really efficient..next up i will mount a water block cooler to it instead..eventually..
Attachments
Photo0002(3).jpg
Photo0002(3).jpg (169.55 KiB) Viewed 6441 times
Photo0001(5).jpg
Photo0001(5).jpg (155.67 KiB) Viewed 6441 times
Last edited by phaseshifter on Wed Jun 28, 2017 3:04 pm, edited 1 time in total.
Build It And They Will Come...Be Bold And Mighty Forces Will Come To Your Aid..!!!

odroid alc
Posts: 77
Joined: Sun Feb 26, 2017 9:11 am
languages_spoken: english
ODROIDs: xu4
pi ZERO
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by odroid alc » Wed Jun 28, 2017 2:55 pm

I'm thainking about ordering the split flow casse. But the fan I have is the nf-a4x20. Its thiker. Will it fit in the casse and or can if I drimil out some of the material ?

Also did any 1 find out if the casse is beader than mounting the fan on top of the heatsink ?

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Thu Jun 29, 2017 1:37 am

odroid alc wrote:I'm thinking about ordering the split flow case but, the fan I have is the nf-a4x20, Its thicker. Will it fit in the case and or can I drill out some of the material ?

Also, did any 1 find out if the case is better than mounting the fan on top of the heatsink ?
The Noctua A4 20mm will not fit in the case and removing material will not help. They were not available when this project started so were not originally considered. Fans can be mounted on the top to the heatsink. Whether top mounted is better or not depends on the fan and the heatsink combo. The A4 20mm should do better since it has a higher output.

odroid alc
Posts: 77
Joined: Sun Feb 26, 2017 9:11 am
languages_spoken: english
ODROIDs: xu4
pi ZERO
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by odroid alc » Thu Jun 29, 2017 1:17 pm

hominoid wrote:
odroid alc wrote:I'm thinking about ordering the split flow case but, the fan I have is the nf-a4x20, Its thicker. Will it fit in the case and or can I drill out some of the material ?

Also, did any 1 find out if the case is better than mounting the fan on top of the heatsink ?
The Noctua A4 20mm will not fit in the case and removing material will not help. They were not available when this project started so were not originally considered. Fans can be mounted on the top to the heatsink. Whether top mounted is better or not depends on the fan and the heatsink combo. The A4 20mm should do better since it has a higher output.
I have the same heat sink(gold) thainking I culd run the odroid without a fan but just using kodi and delug+PIA+openvpn it was heating to much... I thaink.

I thaink thear might be just bearly space for it.
Image

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Fri Jun 30, 2017 1:43 am

You are using an old picture of the original prototype. If you read the whole post you will find that moving the fan forward was one of the first changes that took place. There is an enclosed twin duct as well that is not shown in that picture. One other big reason to use this case, regardless of the cooling you use, is the case back heatsink hold down system. Read the post or the June issue of Odroid Magazine on using a copper perch/shim under your heatsink for improved thermal performance. All of this and more is indicated in the posts above with more recent photos.
Unfortunately, last I looked Ameridroid is still using the prototype pictures. @Ameridroid, Please update your photos.

odroid alc
Posts: 77
Joined: Sun Feb 26, 2017 9:11 am
languages_spoken: english
ODROIDs: xu4
pi ZERO
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by odroid alc » Fri Jun 30, 2017 2:33 pm

ahh.. ya.. thay shuld updat piks.

I have a bunch of coper shims.. I'll use one. I'm thainking I can cut and past some plastice to space the fan. But is the case werth it. Right now I keep hearing the fan kicke on and off just as much if not more than the stock fan. It is mounted/siting on top of heatsink(gold).

Wear do I find/see the "back heatsink hold down system" ?

Ty for the magazin link. Good info thear. I'l start playing around with diferant heatsink thiknisis and a few other thaings to amke it more responsive on cunducting.

Ate first(a weack ago) I had a psu powering a 120mm 12v fan + the gold heatsink 24/7 runing and it keept it ok to good temp's.

sorry for being eleterit. and ty for your time !

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Fri Jun 30, 2017 10:59 pm

odroid alc wrote: Wear do I find/see the "back heatsink hold down system" ?
The second page of posts, at the beginning, has a long write up and pictures on the "case back hold down system".
Changing your governor setting and/or lowering the clock frequency to 1.8ghz or 1.9ghz will significantly help with temps with minimal performance reduction. This SOC uses a lot of energy and produces a lot of heat for the last 100-200mhz. There are plenty of posts on these subjects if you search the forum.

tmihai20
Posts: 222
Joined: Mon Nov 07, 2016 10:56 pm
languages_spoken: english, french, italian, romanian
ODROIDs: XU4, Go, H2, N2 (4GB)
Location: Romania
Has thanked: 29 times
Been thanked: 8 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by tmihai20 » Wed Sep 13, 2017 4:09 am

Can you please share with us the stl for the model that allows docking of external air sources? It is in this post viewtopic.php?f=97&t=26373&start=100#p192202. I already have a 80 mm fan that I am using now and I want it to be as silent as possible.
Riddle me this, riddle me that
Who is afraid of the big, black bat?
I write (in Romanian mostly) on a blog (see my profile)

klunkerbus
Posts: 9
Joined: Fri Oct 06, 2017 9:42 am
languages_spoken: english
ODROIDs: XU4
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by klunkerbus » Wed Oct 18, 2017 5:49 am

New to the Odroid community here. I wanted to thank hominoid for the openSCAD script. In searching for a SBC with USB 3.0 that I can use with OpenMediaVault, the XU4 won out in part because of the script being there. I've ended up working on a pretty hefty overhaul of the scripts for both the split airflow case and the XU4 model that provides more flexibility for the user through parameter changes. Until I get the updates fleshed out some more, here are a few comments and observations -

A few in the thread have asked about the load capacity of the fan driver on the XU4. I haven't seen those questions answered. Page 4 of the XU4 schematic shows the fan is controlled by a PBRN123YT transistor. The amount of copper around the transistor will be the limiting factor in how much power the transistor can dissipate, but the PBRN123YT is specified for a 600mA collector current. It should be able to handle a typical 40mm 5V fan. A 10mm thick Noctua requires 50mA, and the 20mm thick Noctua requires 100mA.

I first read about flatness concerns with the gdstime gold heatsink in the XU4 heatsink without Cooler (or low noise) thread at viewtopic.php?p=142926#p142926. Sure enough, checking mine with a simple metal ruler the bottom of the heatsink is pretty convex. I can see hominoid's "rocking test" of the heatsink on the SOC being helpful with that. I can also see how raising the heatsink up on a copper perch would allow lapping of the heatsink bottom to a flatter surface, since that would remove the isolation properties of the anodized coating.

klunkerbus
Posts: 9
Joined: Fri Oct 06, 2017 9:42 am
languages_spoken: english
ODROIDs: XU4
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by klunkerbus » Sun Nov 19, 2017 4:15 pm

I've completed futzing around with tweaking the openSCAD design. Rather than hijack hominoid's thread, I've posted information on it in the Configurable openSCAD Open-Frame XU4 Case thread at viewtopic.php?f=97&t=29066

Captain_Data
Posts: 5
Joined: Tue Nov 14, 2017 3:09 am
languages_spoken: english
ODROIDs: 3 x XU4 + CloudShell2 + Hominoid Case
Has thanked: 0
Been thanked: 0
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by Captain_Data » Mon Mar 12, 2018 9:08 pm

What is supposed to be the length of the black spacers that support the top cover?

When I assemble the case, the top cover does not align with the fan. The top cover is too low.

My spacers are 24mm. Is that correct?

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Thu Mar 15, 2018 8:30 am

Captain_Data wrote:What is supposed to be the length of the black spacers that support the top cover?

When I assemble the case, the top cover does not align with the fan. The top cover is too low.

My spacers are 24mm. Is that correct?
Hi @Captain_Data,
The case Ameridroid is selling is using 24mm spacers from earlier in the development cycle. I don't believe they ever updated what they were doing. I had updated the design early on because 20mm standoffs were easier to come by. That is what the stl's from the above June 24th post are using; the last one I posted. Each stub on the top and bottom are 8mm tall. More recently @kluckerbus has made a bunch of mods in their own fork of the project which I'm not familiar with at all. Where or how did you acquire your case? Can you post a picture or tell me the length of the stubs on the top and bottom?

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Fri Mar 16, 2018 5:10 am

"Hi hominoid,

I didn't want to clutter the forum post you made regarding your case design, so I'm sending this PM instead.

I am planning to buy the Odroid XU4, print the case you provided, and use the same heat sink and fan you used. I would love to use shims like you mentioned, however, I can't seem to find any shims that are as thick as yours. I seem to only find 1.2mm or 2.0mm in a few instances. Do you suspect 1mm shims would work just as well as the thickness you tried? I considered stacking a few shims to gain thickness, but that just creates more points where I would need to add thermal compound and heat transfer would probably diminish noticeably.

Also, did you lap the heat sink? I have read that the heat sink isn't very flat and I'm concerned about poor cooling performance if it is not flat.

Thanks!"


I received this PM from a user regarding the Split Airflow Case. I don't mind receiving PM's from forum users but they were asking good questions so I wanted to reply on the forum so others could benefit from the questions they asked.

The shims/perch I used were handmade from a 1/8" stock C110 copper bar I purchased on ebay. You would have to buy maybe more then you might need and have the tools (hacksaw, file, emery cloth) to cut and finish the shim. If you buy a bar you then have extra material to make a larger spreader shim for better heat transfer to your heat sink. That will help a lot with cooling. If that is too much, then buy the thickest ones you can get. 1mm shims aren't going to perform well (see P1 test results of 3x20ga CU) but you can always try it, do so some testing and see were your at; 2mm or greater would be much better. You can always upgrade the shim in the future.

I wouldn't be too worried about stacking two of them together solely because of diminished thermal transfer. It's not the ideal situation but the loss is negligible in the big picture, IMO. I had used a perch and a heat spreader stacked. Also, how often are you going to need to disassemble them? I find once I'm setup, they are there for a long time. If you need to disassemble they will have a tendency to stay together anyway.

Regardless if you stack two together, what's more important IMO is using a good thermal compound (2.5W/mK > preferably 5W/mK>) and bond them first with a good squeeze. For the testing I was using a paste rated at 5.3W/mK(Antec Forumla 6). I've also have used Antec Formula 7 rated at 8.3W/mK, there are others that will work equally as well, do some research. Be wary of anything that isn't clearly labeled or uses a different metric.

The heatsink is not the flattest, I didn't lap mine. Here's some rhetorical questions to consider before making your choice. Do you have a trued lapping stone? What effect is the anodization going to have on it? Are you willing to possibly sacrifice it(they can be expensive)? Are you going to lap the SOC? The SOC is a small percentage of the total area of the heat sink so that works in your favor in that the change in flatness won't be as great as it is over the entirety of the heat sink. If you consider the cost, gain and possible electromagnetic problems with removing the anodization, is it worth it? It was not for me. I decided to use the extra time and resources to make a good shim, spreader and buy a better fan. Your choice might be different.

Lastly, I have found there is a great group of professional engineers, developer and hobbyist with varying skills and experience on this forum that are always open, willing and gracious in sharing what they know. Don't be afraid to use them; it's the whole point of community collaboration. If your on topic for the thread, there is no such thing as clutter. And, even the professionals learn something from us hobbyist from time to time. ;)

hominoid
Posts: 356
Joined: Tue Feb 28, 2017 3:55 am
languages_spoken: english
ODROIDs: C2, XU4, MC1, N1, N2
Location: Lake Superior Basin, USA
Has thanked: 19 times
Been thanked: 47 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by hominoid » Mon Aug 27, 2018 3:40 pm

XU4 Split Air Flow 80mm Fan Adapter
I have been working on getting the previously purposed XU4 Split Air Flow Case 80mm fan adapter design finished and a test system constructed and running. After addressing several design issues, 3D printer upgrades/problems and the sourcing all of the parts needed the initial design and construction is complete.
XU4-SAF80.jpg
XU4-SAF80.jpg (46.99 KiB) Viewed 4106 times
XU4 Split Air Flow 80mm fan adapter with Noctua 80mm PWM Fan
Model: NF-A8 PWM 12v
RPM: 2200
dB/A: 17.7
m3/h: 55.5
mm H2O: 2.37

Design files
This 80mm adapter should work with all released versions of the XU4 Split Airflow 40mm case including the ones from Ameridroid.
XU4_Split_AirFlow_Adapter_v.1.scad
XU4_Split_AirFlow_Adapter_v.1.stl
Odroid XU4 Split AirFlow Adapter_v.1.zip
(167.66 KiB) Downloaded 34 times
I have incorporated the following changes to the original XU4 Split Airflow 40mm case implementation in the thermal pipeline area.
The shape and mass of the copper spreader has been changed (20mm x 27mm at 23g) to allow for increased cooling area of the SOC and the underside of the heatsink. The perch has been permanently bonded (hard soldered or silver soldered)to the copper spreader. The 12pin connector has been removed to allow significantly more airflow around both sides of the SOC and thermal pipeline.
Thermal_pipeline.jpg
Thermal_pipeline.jpg (67.39 KiB) Viewed 4106 times
3D Print
As you can see in the photo above, this was a difficult print for me. I wanted to make this design so that it could be printed without support. I have accomplished that but I still get some filament pull out inside at the top. This is a printing problem and not a design issue. One of the reasons is the cuts that are necessary to accommodate the fasteners in a manner that allows them to be perpendicular to the horizontal plane of the design. Otherwise, they insert at angles that do not accommodate tightening and a good fit. The following design image illustrates this point.
XU4-SAFA-Bottom Angle-clip-s.jpg
XU4-SAFA-Bottom Angle-clip-s.jpg (60.29 KiB) Viewed 4106 times
Your printing an overhang at the top of the adapter so make sure your slicing software is finely tuned and your printer is slowed down for this one. No speed printing this model. One other note of interest is that I used the feet of the standard XU4 Split Airflow case as evident in the lead photo. They are too short for the airflow of most 80mm fans I tried. They do not allow enough airflow to feed the fan which manifests itself in a lower CFM, velocity and added back pressure in the air channel. I simply raised the case on some small blocks for the initial test. The bottom of the fan needs to have about double the clearance in order to allow for adequate air intake for the highest volume fans. This will vary depending on the over capability of the fan used. The fans I tried that had less velocity and lower CFM did better at the designed height then ones with more airflow. I'll go back and change the feet height at a later date, for now I wanted to get on with testing and see what this baby could do!

Fan Testing
First off I have to say that I had a full compliment of testing that I was going to do before releasing anything. It included comprehensive testing of the design across multiple fans and controls. As well as full comparison testing of the original XU4 Split Airflow Case with it's 40mmx10mm 5v fan, HK's OEM active and passive cooled XU4's and active cooled MC1 cluster. I also have been working on interfacing a true PWM driven fan off the SOC's PWM signal. Even though HK's OEM heatsink fan is part of a PWM system, it uses a linear voltage controlled fan and relies on the kernel driver for proper startup and operation. This is fine but introduces some potential problem areas because the kernel needs to be adjusted to accommodate different fans, more kernel code needs to be used to implement the design and difference within a given fan model can become problematic when at the extremes of the duty cycle. I believe this is one of the reasons there have been lots of tweaks to new kernels to address the chirping, low duty cycle noise and failed starts.

The PWM specification requires all 4-pin PWM fans to include the necessary circuitry to address all of the these problems. Per the PWM specification, fan vendors must clearly divulge and guarantee the operational duty cycle of their fans, provide the correct initial startup speed and provide one of several modes of operation if no PWM signal is provided. They are also required to provide continues power (usually 12v) regardless of the PWM signal. And, there are many more fan choices for end users to choose from that can be implement without worry or kernel change. This is why I always bring it up during discussions about future product wish lists. IMO it provides a lot of value for very little capital, PCB real estate and all or most of the hardware is already there if the SBC has 12 volts available. This becomes more important I believe as the use cases for SBC's continue to expand and the processing demands and capabilities across both the CPU and GPU cores increase. If your interested, I had previously posted the PWM Fan Specification 1.3 in this thread.

Preliminary Test
I used the same testing strategy as I did for the initial XU4 Split Air Flow 40mm case so comparisons could be made and improvements judged.

sysbench CPU prime for load and heat generation

Code: Select all

sysbench cpu --cpu-max-prime=300000 --threads=8 run
watchtemp script for temperature and frequency data collection

Code: Select all

watchtemp | tee noctua80mm-prime300k-test1.txt
GNUPlot scripts for graph generation and analysis
noctua80mm-prime300k-2ghz-test1.gpl

Code: Select all

#!/usr/bin/gnuplot --persist
#   
#
# Decription: GNUPLOT script file that plots temperature and 
#             frequency for all cores from a watchtemp_all.sh data file
#             formated as "count,freq1,freq2,freq3,freq4,core1,core2,core3,core4,gpu"
#
# Version 5.x
# Use: gnuplot multicore_temp.gpl outputchart.png file1
#
# Version 4.x
# Use: gnuplot multicore_temp.gpl
#
# Author: @hominoid www.forum.odroid.com
# Date: 5/3/2017
# Released to the public domain with no warranty of any kind
#

# for version 4.x only, uncomment and enter file names instead of on the commandline, see use above
#ARG1 = "noctua80mm-prime300k-2ghz-test1.png"
#ARG2 = "test3.txt"

set term pngcairo size 1024,768 enhanced font 'Verdana,10'
set output ARG1
set datafile separator ","

# title and line style definitions

data_title1 = "Core 4 Temp"
set style line 1 lc rgb "blue" lw 1

data_title2 = "Core 5 Temp"
set style line 2 lc rgb "forest-green" lw 1

data_title3 = "Core 6 Temp"
set style line 3 lc rgb "cyan" lw 1

data_title4 = "Core 7 Temp"
set style line 4 lc rgb "dark-violet" lw 1

data_title5 = "GPU Temp"
set style line 5 lc rgb "red" lw 1

# border and grid line styles
set style line 9 lc rgb "black" lt 1 lw 1
set style line 10 lc rgb "black" lt 0 lw 1
 
# graph style
set border 3 front ls 9
set tics nomirror
set grid back ls 10
set grid ytics ls 10

# legend top right corner
set key noauto font 'Verdana,13'

set multiplot layout 2,1 title "CPU/GPU Core Temp vs. Freq for Sysbench Prime 300K" font 'Verdana,16'

# temperature plot
set size ratio 0 1,.8
set origin 0,0.2
set lmargin 11
# y axis
set ylabel 'Temperature (c)' font 'Verdana,12'
set yrange [30:100]
set ytics 30,2 border nomirror out
set mytics
# x axis
set xlabel 'Time (seconds)'
set xtics 0,30 border nomirror out
set mxtics

plot ARG2 using 1:6 with lines ls 1 axes x1y1 title data_title1, ARG2 using 1:7 with lines ls 2 axes x1y1 title data_title2, ARG2 using 1:8 with lines ls 3 axes x1y1 title data_title3, ARG2 using 1:9 with lines ls 4 axes x1y1 title data_title4, ARG2 using 1:10 with lines ls 5 axes x1y1 title data_title5

# frequency plot
set size ratio 0 1,.2
set origin 0,0
set lmargin 11
# y axis
set ylabel 'Frequency (ghz)' font 'Verdana,12'
set yrange [0:2]
set ytics 0,.5 border nomirror out
set format y '%.1f'
# x axis
set noxlabel

plot ARG2 using 1:($2/1000.0) with lines ls 2 axes x1y2 notitle, ARG2 using 1:3 with lines ls 3 axes \
					x1y2, ARG2 using 1:4 with lines ls 4 axes x1y2, ARG2 using 1:5 with lines ls 5 axes x1y2

unset multiplot
The original XU4 Split Airflow 40mm Case tests were run at 2Ghz or lower on kernel 3.10.104 at a slower DRAM speed(866Mhz I believe) and the current 80mm adapter tests were run at 2Ghz on 933Mhz DRAM on kernel 4.14.66-147

The initial test results seemed to be way off and I believe there is a problem. This is why I decided to post now and have not completed the remainder of the testing I had planned. I worked on this all weekend and can not figure out what's wrong. The results are not believable and there is at least one anomaly that I have no explanation for and need the knowledge and wisdom of the HK staff and Odroid community. Some results and comparisons to the original XU4 Split Airflow 40mm Case testing to illustrate my point:

I was using sysbench prime 300k for the original testing so I started with that test for the 80mm fan adapter. Below are the summary results

Code: Select all

hominoid@c0n1:~$ uname -a
Linux c0n1 4.14.66-147 #1 SMP PREEMPT Wed Aug 22 13:24:49 -03 2018 armv7l armv7l armv7l GNU/Linux
hominoid@c0n1:~$ sysbench cpu --cpu-max-prime=300000 --threads=8 run
sysbench 1.0.11 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:
Number of threads: 8
Initializing random number generator from current time


Prime numbers limit: 300000

Initializing worker threads...

Threads started!

CPU speed:
    events per second:     3.98

General statistics:
    total time:                          11.2872s
    total number of events:              45

Latency (ms):
         min:                               1281.84
         avg:                               1888.62
         max:                               3094.84
         95th percentile:                   2932.60
         sum:                              84987.73

Threads fairness:
    events (avg/stddev):           5.6250/1.87
    execution time (avg/stddev):   10.6235/0.39

hominoid@c0n1:~$
Here is a link to the original Split Airflow Case 40mm fan results for the sysbench 300k test. Examining the sysbench 300k test summary for each configuration it took 1891.66 seconds to complete originally compared to the 80mm fan adapter results which took 11.66 seconds.

The original 40mm fan sysbench 1M test took 10783.89 seconds and only 18.66 seconds for the 80mm fan adapter testing.

Code: Select all

root@odroid-xu4:/home/odroid# sysbench --test=cpu --cpu-max-prime=1000000 --num-threads=8 run
sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 8

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 1000000


Test execution summary:
    total time:                          10783.8903s
    total number of events:              10000
    total time taken by event execution: 86246.5639
    per-request statistics:
         min:                               6020.55ms
         avg:                               8624.66ms
         max:                              16558.43ms
         approx.  95 percentile:           15014.19ms

Threads fairness:
    events (avg/stddev):           1250.0000/185.15
    execution time (avg/stddev):   10780.8205/1.74
80mm Adapter Results

Code: Select all

hominoid@c0n1:~$ sysbench cpu --cpu-max-prime=1000000 --threads=8 run
sysbench 1.0.11 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:
Number of threads: 8
Initializing random number generator from current time


Prime numbers limit: 1000000

Initializing worker threads...

Threads started!

CPU speed:
    events per second:     0.55

General statistics:
    total time:                          18.0668s
    total number of events:              10

Latency (ms):
         min:                               7248.39
         avg:                              13185.29
         max:                              18064.62
         95th percentile:                  18075.36
         sum:                             131852.90

Threads fairness:
    events (avg/stddev):           1.2500/0.43
    execution time (avg/stddev):   16.4816/1.20

hominoid@c0n1:~$
I didn't run a sysbench 10M in the original testing but I did for the 80mm fan adapter because of the low times I was receiving.
sysbench Summary

Code: Select all

hominoid@c0n1:~$ sysbench cpu --cpu-max-prime=10000000 --threads=8 run
sysbench 1.0.11 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:
Number of threads: 8
Initializing random number generator from current time


Prime numbers limit: 10000000

Initializing worker threads...

Threads started!

CPU speed:
    events per second:     0.03

General statistics:
    total time:                          313.8163s
    total number of events:              8

Latency (ms):
         min:                             201861.57
         avg:                             257793.02
         max:                             313815.99
         95th percentile:                 100000.00
         sum:                            2062344.14

Threads fairness:
    events (avg/stddev):           1.0000/0.00
    execution time (avg/stddev):   257.7930/55.62

hominoid@c0n1:~$ 
sysbench 1M Data

Code: Select all

T, Freq4,   Freq5,   Freq6,   Freq7,   T4, T5, T6, T7, TGPU
0, 2000000, 2000000, 2000000, 2000000, 33, 34, 36, 35, 30
2, 2000000, 2000000, 2000000, 2000000, 51, 48, 55, 54, 36
4, 2000000, 2000000, 2000000, 2000000, 61, 57, 65, 65, 42
6, 2000000, 2000000, 2000000, 2000000, 64, 59, 69, 68, 45
8, 2000000, 2000000, 2000000, 2000000, 65, 61, 70, 69, 46
10, 2000000, 2000000, 2000000, 2000000, 66, 62, 71, 70, 47
12, 2000000, 2000000, 2000000, 2000000, 67, 63, 72, 71, 47
14, 2000000, 2000000, 2000000, 2000000, 67, 63, 73, 72, 48
16, 2000000, 2000000, 2000000, 2000000, 68, 64, 73, 72, 48
18, 2000000, 2000000, 2000000, 2000000, 68, 64, 73, 72, 48
20, 2000000, 2000000, 2000000, 2000000, 68, 64, 74, 73, 49
22, 2000000, 2000000, 2000000, 2000000, 68, 65, 74, 73, 49
24, 2000000, 2000000, 2000000, 2000000, 69, 65, 75, 74, 49
26, 2000000, 2000000, 2000000, 2000000, 69, 65, 75, 74, 49
28, 2000000, 2000000, 2000000, 2000000, 69, 65, 75, 74, 50
30, 2000000, 2000000, 2000000, 2000000, 70, 66, 76, 75, 50
32, 2000000, 2000000, 2000000, 2000000, 70, 66, 76, 75, 50
34, 2000000, 2000000, 2000000, 2000000, 70, 66, 76, 75, 50
36, 2000000, 2000000, 2000000, 2000000, 70, 66, 76, 75, 51
38, 2000000, 2000000, 2000000, 2000000, 70, 67, 77, 75, 51
40, 2000000, 2000000, 2000000, 2000000, 71, 67, 77, 75, 51
42, 2000000, 2000000, 2000000, 2000000, 71, 67, 77, 76, 51
44, 2000000, 2000000, 2000000, 2000000, 71, 67, 77, 76, 51
46, 2000000, 2000000, 2000000, 2000000, 71, 67, 77, 76, 51
48, 2000000, 2000000, 2000000, 2000000, 71, 67, 77, 76, 51
50, 2000000, 2000000, 2000000, 2000000, 71, 67, 77, 76, 51
52, 2000000, 2000000, 2000000, 2000000, 71, 67, 77, 76, 52
54, 2000000, 2000000, 2000000, 2000000, 72, 68, 78, 77, 52
56, 2000000, 2000000, 2000000, 2000000, 72, 68, 78, 77, 52
58, 2000000, 2000000, 2000000, 2000000, 72, 68, 78, 77, 52
60, 2000000, 2000000, 2000000, 2000000, 72, 68, 78, 77, 52
62, 2000000, 2000000, 2000000, 2000000, 73, 68, 79, 77, 52
64, 2000000, 2000000, 2000000, 2000000, 72, 69, 79, 78, 52
66, 2000000, 2000000, 2000000, 2000000, 73, 69, 79, 78, 53
68, 2000000, 2000000, 2000000, 2000000, 73, 69, 79, 78, 53
70, 2000000, 2000000, 2000000, 2000000, 73, 69, 79, 78, 53
72, 2000000, 2000000, 2000000, 2000000, 73, 69, 79, 78, 53
74, 2000000, 2000000, 2000000, 2000000, 73, 69, 79, 78, 53
76, 2000000, 2000000, 2000000, 2000000, 73, 70, 80, 79, 53
78, 2000000, 2000000, 2000000, 2000000, 74, 70, 80, 79, 53
80, 2000000, 2000000, 2000000, 2000000, 74, 70, 80, 79, 54
82, 2000000, 2000000, 2000000, 2000000, 74, 70, 80, 79, 54
84, 2000000, 2000000, 2000000, 2000000, 74, 70, 80, 79, 54
86, 2000000, 2000000, 2000000, 2000000, 74, 70, 80, 79, 54
88, 2000000, 2000000, 2000000, 2000000, 74, 70, 80, 79, 54
90, 2000000, 2000000, 2000000, 2000000, 75, 70, 80, 79, 54
92, 2000000, 2000000, 2000000, 2000000, 75, 71, 81, 80, 54
94, 2000000, 2000000, 2000000, 2000000, 75, 71, 81, 80, 54
96, 2000000, 2000000, 2000000, 2000000, 75, 71, 81, 80, 55
98, 2000000, 2000000, 2000000, 2000000, 75, 71, 81, 80, 55
100, 2000000, 2000000, 2000000, 2000000, 75, 71, 81, 80, 55
102, 2000000, 2000000, 2000000, 2000000, 75, 71, 81, 80, 55
104, 2000000, 2000000, 2000000, 2000000, 75, 71, 81, 80, 55
106, 2000000, 2000000, 2000000, 2000000, 75, 71, 81, 80, 55
108, 2000000, 2000000, 2000000, 2000000, 75, 71, 81, 80, 55
110, 2000000, 2000000, 2000000, 2000000, 75, 71, 82, 80, 55
112, 2000000, 2000000, 2000000, 2000000, 75, 71, 82, 80, 55
114, 2000000, 2000000, 2000000, 2000000, 76, 71, 82, 81, 55
116, 2000000, 2000000, 2000000, 2000000, 76, 72, 82, 80, 55
118, 2000000, 2000000, 2000000, 2000000, 76, 72, 82, 81, 55
120, 2000000, 2000000, 2000000, 2000000, 76, 72, 82, 81, 55
122, 2000000, 2000000, 2000000, 2000000, 76, 72, 82, 81, 56
124, 2000000, 2000000, 2000000, 2000000, 76, 72, 82, 81, 56
126, 2000000, 2000000, 2000000, 2000000, 76, 72, 82, 81, 56
128, 2000000, 2000000, 2000000, 2000000, 76, 72, 82, 81, 56
130, 2000000, 2000000, 2000000, 2000000, 76, 72, 82, 81, 56
132, 2000000, 2000000, 2000000, 2000000, 76, 72, 83, 81, 56
134, 2000000, 2000000, 2000000, 2000000, 77, 72, 83, 81, 56
136, 2000000, 2000000, 2000000, 2000000, 77, 72, 83, 82, 56
138, 2000000, 2000000, 2000000, 2000000, 77, 72, 83, 82, 56
140, 2000000, 2000000, 2000000, 2000000, 77, 73, 83, 82, 56
142, 2000000, 2000000, 2000000, 2000000, 77, 73, 83, 82, 56
144, 2000000, 2000000, 2000000, 2000000, 77, 73, 83, 82, 57
146, 2000000, 2000000, 2000000, 2000000, 77, 73, 84, 82, 57
148, 2000000, 2000000, 2000000, 2000000, 77, 73, 84, 82, 57
150, 2000000, 2000000, 2000000, 2000000, 77, 73, 84, 83, 57
152, 2000000, 2000000, 2000000, 2000000, 78, 73, 84, 83, 57
154, 2000000, 2000000, 2000000, 2000000, 78, 74, 84, 83, 57
156, 2000000, 2000000, 2000000, 2000000, 78, 74, 84, 83, 57
158, 2000000, 2000000, 2000000, 2000000, 78, 74, 84, 83, 57
160, 2000000, 2000000, 2000000, 2000000, 78, 74, 84, 83, 57
162, 2000000, 2000000, 2000000, 2000000, 78, 74, 84, 83, 57
164, 2000000, 2000000, 2000000, 2000000, 78, 74, 84, 83, 57
166, 2000000, 2000000, 2000000, 2000000, 78, 74, 84, 83, 57
168, 2000000, 2000000, 2000000, 2000000, 78, 74, 84, 83, 57
170, 1900000, 1900000, 1900000, 1900000, 76, 72, 82, 80, 57
172, 2000000, 2000000, 2000000, 2000000, 78, 74, 84, 83, 57
174, 2000000, 2000000, 2000000, 2000000, 78, 74, 84, 83, 57
176, 2000000, 2000000, 2000000, 2000000, 78, 74, 84, 83, 58
178, 2000000, 2000000, 2000000, 2000000, 77, 73, 83, 82, 57
180, 2000000, 2000000, 2000000, 2000000, 55, 53, 56, 56, 50
182, 2000000, 2000000, 2000000, 2000000, 52, 50, 52, 52, 47
184, 2000000, 2000000, 2000000, 2000000, 50, 48, 51, 50, 46
186, 2000000, 2000000, 2000000, 2000000, 50, 47, 50, 49, 45
188, 2000000, 2000000, 2000000, 2000000, 49, 46, 49, 49, 45
190, 2000000, 2000000, 2000000, 2000000, 49, 46, 48, 48, 44
192, 2000000, 2000000, 2000000, 2000000, 48, 46, 48, 48, 44
194, 2000000, 2000000, 2000000, 2000000, 48, 46, 48, 47, 43
196, 2000000, 2000000, 2000000, 2000000, 48, 45, 47, 47, 43
198, 2000000, 2000000, 2000000, 2000000, 47, 45, 47, 47, 43
200, 2000000, 2000000, 2000000, 2000000, 47, 45, 47, 46, 43
202, 2000000, 2000000, 2000000, 2000000, 47, 45, 47, 46, 43
204, 2000000, 2000000, 2000000, 2000000, 47, 45, 47, 46, 43
206, 2000000, 2000000, 2000000, 2000000, 47, 44, 46, 46, 42
208, 2000000, 2000000, 2000000, 2000000, 47, 44, 46, 46, 42
210, 2000000, 2000000, 2000000, 2000000, 46, 44, 46, 46, 42
212, 2000000, 2000000, 2000000, 2000000, 46, 44, 46, 46, 42
214, 2000000, 2000000, 2000000, 2000000, 46, 44, 46, 46, 42
216, 2000000, 2000000, 2000000, 2000000, 46, 44, 46, 46, 42
218, 2000000, 2000000, 2000000, 2000000, 46, 44, 46, 45, 42
220, 2000000, 2000000, 2000000, 2000000, 46, 44, 46, 45, 42
222, 2000000, 2000000, 2000000, 2000000, 46, 43, 46, 45, 42
224, 2000000, 2000000, 2000000, 2000000, 46, 43, 45, 45, 42
226, 2000000, 2000000, 2000000, 2000000, 46, 43, 45, 45, 41
228, 2000000, 2000000, 2000000, 2000000, 45, 43, 45, 45, 41
230, 2000000, 2000000, 2000000, 2000000, 46, 43, 45, 45, 41
232, 2000000, 2000000, 2000000, 2000000, 46, 43, 45, 45, 41
234, 2000000, 2000000, 2000000, 2000000, 46, 43, 45, 45, 41
236, 2000000, 2000000, 2000000, 2000000, 45, 43, 45, 45, 41
238, 2000000, 2000000, 2000000, 2000000, 45, 43, 45, 45, 41
240, 2000000, 2000000, 2000000, 2000000, 45, 43, 45, 45, 41
242, 2000000, 2000000, 2000000, 2000000, 45, 43, 45, 45, 41
244, 2000000, 2000000, 2000000, 2000000, 45, 43, 45, 44, 41
246, 2000000, 2000000, 2000000, 2000000, 45, 43, 45, 44, 41
248, 2000000, 2000000, 2000000, 2000000, 45, 42, 45, 44, 41
250, 2000000, 2000000, 2000000, 2000000, 45, 43, 44, 44, 41
252, 2000000, 2000000, 2000000, 2000000, 45, 42, 44, 44, 41
254, 2000000, 2000000, 2000000, 2000000, 45, 42, 44, 44, 40
256, 2000000, 2000000, 2000000, 2000000, 45, 42, 44, 44, 40
258, 2000000, 2000000, 2000000, 2000000, 45, 42, 44, 44, 40
260, 2000000, 2000000, 2000000, 2000000, 45, 42, 44, 44, 40
262, 2000000, 2000000, 2000000, 2000000, 45, 42, 44, 44, 40
264, 2000000, 2000000, 2000000, 2000000, 44, 42, 44, 44, 40
266, 2000000, 2000000, 2000000, 2000000, 45, 42, 44, 44, 40
268, 2000000, 2000000, 2000000, 2000000, 44, 42, 44, 44, 40
270, 2000000, 2000000, 2000000, 2000000, 44, 42, 44, 44, 40
272, 2000000, 2000000, 2000000, 2000000, 44, 42, 44, 43, 40
274, 2000000, 2000000, 2000000, 2000000, 44, 42, 44, 44, 40
276, 2000000, 2000000, 2000000, 2000000, 45, 42, 44, 44, 40
278, 2000000, 2000000, 2000000, 2000000, 44, 42, 44, 44, 40
280, 2000000, 2000000, 2000000, 2000000, 44, 42, 44, 44, 40
282, 2000000, 2000000, 2000000, 2000000, 44, 42, 44, 43, 40
284, 2000000, 2000000, 2000000, 2000000, 44, 42, 44, 43, 40
286, 2000000, 2000000, 2000000, 2000000, 44, 42, 44, 43, 40
288, 2000000, 2000000, 2000000, 2000000, 44, 42, 44, 44, 40
290, 2000000, 2000000, 2000000, 2000000, 40, 41, 43, 42, 37
292, 2000000, 2000000, 2000000, 2000000, 40, 40, 42, 41, 37
294, 2000000, 2000000, 2000000, 2000000, 40, 40, 42, 41, 36
296, 2000000, 2000000, 2000000, 2000000, 40, 40, 42, 41, 36
298, 2000000, 2000000, 2000000, 2000000, 40, 40, 42, 41, 36
300, 2000000, 2000000, 2000000, 2000000, 40, 40, 42, 41, 36
302, 2000000, 2000000, 2000000, 2000000, 39, 40, 41, 40, 36
304, 2000000, 2000000, 2000000, 2000000, 39, 40, 42, 40, 36
306, 2000000, 2000000, 2000000, 2000000, 39, 40, 41, 41, 36
308, 2000000, 2000000, 2000000, 2000000, 39, 40, 41, 41, 36
310, 2000000, 2000000, 2000000, 2000000, 39, 39, 41, 40, 36
312, 2000000, 2000000, 2000000, 2000000, 39, 39, 41, 40, 36
314, 2000000, 2000000, 2000000, 2000000, 39, 39, 41, 40, 35
316, 2000000, 2000000, 2000000, 2000000, 39, 39, 41, 40, 35
318, 2000000, 2000000, 2000000, 2000000, 39, 39, 41, 40, 35
320, 2000000, 2000000, 2000000, 2000000, 39, 39, 41, 40, 35
322, 2000000, 2000000, 2000000, 2000000, 39, 39, 41, 40, 35
324, 2000000, 2000000, 2000000, 2000000, 39, 39, 41, 40, 35
326, 2000000, 2000000, 2000000, 2000000, 38, 39, 41, 40, 35
328, 2000000, 2000000, 2000000, 2000000, 38, 39, 41, 40, 35
330, 2000000, 2000000, 2000000, 2000000, 38, 39, 41, 40, 35
332, 2000000, 2000000, 2000000, 2000000, 38, 39, 41, 40, 35
334, 2000000, 2000000, 2000000, 2000000, 38, 39, 41, 39, 35
336, 2000000, 2000000, 2000000, 2000000, 38, 39, 41, 39, 35
338, 2000000, 2000000, 2000000, 2000000, 38, 39, 41, 40, 35
340, 2000000, 2000000, 2000000, 2000000, 38, 39, 41, 39, 35
342, 2000000, 2000000, 2000000, 2000000, 38, 39, 40, 39, 35
344, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 35
346, 2000000, 2000000, 2000000, 2000000, 38, 39, 40, 39, 35
348, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 35
350, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 35
352, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 35
354, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
356, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
358, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
360, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
362, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
364, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
366, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
368, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
370, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
372, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
374, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
376, 2000000, 2000000, 2000000, 2000000, 37, 38, 40, 39, 34
378, 2000000, 2000000, 2000000, 2000000, 38, 38, 40, 39, 34
380, 2000000, 2000000, 2000000, 2000000, 37, 38, 40, 39, 34
382, 2000000, 2000000, 2000000, 2000000, 37, 38, 39, 39, 34
384, 2000000, 2000000, 2000000, 2000000, 37, 38, 39, 38, 34
386, 2000000, 2000000, 2000000, 2000000, 37, 38, 40, 39, 34
388, 2000000, 2000000, 2000000, 2000000, 37, 38, 39, 39, 34
390, 2000000, 2000000, 2000000, 2000000, 37, 38, 39, 39, 34
392, 2000000, 2000000, 2000000, 2000000, 37, 38, 39, 38, 34
394, 2000000, 2000000, 2000000, 2000000, 37, 38, 39, 38, 34
396, 2000000, 2000000, 2000000, 2000000, 37, 38, 39, 38, 34
398, 2000000, 2000000, 2000000, 2000000, 37, 38, 39, 38, 34
400, 2000000, 2000000, 2000000, 2000000, 37, 38, 39, 38, 34
402, 2000000, 2000000, 2000000, 2000000, 37, 38, 39, 38, 34
404, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 34
406, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 34
408, 2000000, 2000000, 2000000, 2000000, 37, 38, 39, 38, 34
410, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 34
412, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 34
414, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
416, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 34
418, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
420, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
422, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
424, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
426, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
428, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
430, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
432, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
434, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
436, 2000000, 2000000, 2000000, 2000000, 36, 37, 39, 38, 33
438, 2000000, 2000000, 2000000, 2000000, 36, 37, 39, 38, 33
440, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
442, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
444, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
446, 2000000, 2000000, 2000000, 2000000, 36, 37, 39, 38, 33
448, 2000000, 2000000, 2000000, 2000000, 37, 37, 39, 38, 33
450, 2000000, 2000000, 2000000, 2000000, 36, 37, 39, 37, 33
452, 2000000, 2000000, 2000000, 2000000, 36, 37, 39, 38, 33
454, 2000000, 2000000, 2000000, 2000000, 36, 37, 39, 38, 33
456, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 38, 33
458, 2000000, 2000000, 2000000, 2000000, 36, 37, 39, 37, 33
460, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 37, 33
462, 2000000, 2000000, 2000000, 2000000, 36, 37, 39, 37, 33
464, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 37, 33
466, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 38, 33
468, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 37, 33
470, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 37, 33
472, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 37, 33
474, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 37, 33
476, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 37, 33
478, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 37, 33
480, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 37, 33
482, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 33
484, 2000000, 2000000, 2000000, 2000000, 36, 37, 38, 37, 33
486, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 33
488, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 33
490, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 33
492, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 33
494, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 33
496, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 33
498, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
500, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 33
502, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
504, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
506, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
508, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
510, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
512, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
514, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
516, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
518, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
520, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
522, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
524, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
526, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
528, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
530, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
532, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
534, 2000000, 2000000, 2000000, 2000000, 35, 36, 38, 37, 32
536, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
538, 2000000, 2000000, 2000000, 2000000, 36, 36, 38, 37, 32
It only took 313.81 seconds to complete but there was also an anomaly in the middle of the test right after a brief throttling to 1.9Ghz at 170 seconds; all the core temperatures dropped from approximately 80c to the high 40's around 180 seconds until test completion at 313.81 seconds.
noctua80mm-prime10M-2ghz-test1.png
noctua80mm-prime10M-2ghz-test1.png (59.21 KiB) Viewed 4106 times
I was watching the test run and I originally thought the system finished but saw it had not completed. I then thought it hung and I tried to open another shell and examine what was going on but the test finished "normally" before I could type a command. In examining the system afterwards everything seemed to be operating fine. I spent the rest of the weekend checking my work and trying to figure out what's going. With no further ideas other than sysbench is/was broke, I decided to stop further testing and seek the wisdom and knowledge of the community. All thoughts, ideas and speculation appreciated!

User avatar
rooted
Posts: 7336
Joined: Fri Dec 19, 2014 9:12 am
languages_spoken: english
Location: Gulf of Mexico, US
Has thanked: 495 times
Been thanked: 137 times
Contact:

Re: Split Air Flow Vertical or Horizontal 40mm Case

Unread post by rooted » Tue Aug 28, 2018 5:27 am

I believe sysbench is semi broken on ARM, I searched for a discussion I had with someone about vastly different sysbench results on latest sysbench vs older revisions.

My suggestion is use an older version of sysbench.

As a second benchmark or alternative, compiling the linux kernel is great.

Between compilations just make sure to:

Code: Select all

make mrproper
make odroidxu4_defconfig

Post Reply

Return to “Hardware and peripherals”

Who is online

Users browsing this forum: No registered users and 6 guests