XU4 Kernel 4.9 stability test with Coin mining software

Moderators: mdrjr, odroid

XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Sat Jun 03, 2017 11:31 am

As we posted in other thread, we started to test the XU4 Kernel 4.9.27 stability from May 20th (exactly two weeks ago).
20pcs of XU4 boards have been running the Verium coin(VRM) mining software to utilize all 160 CPU cores and 40GB RAM as much as possible.
After two weeks of intensive test, we can proudly say the Kernel 4.9 LTS on XU4 is quite stable.


We had to reboot all of them two times to update the Kernel when 4.9.28 and 4.9.30 were released.
A couple of units had to be manually reset because of 5V DC cable issue.
So we added some more power wires to reinforce the power rails.
Other than that, there was no single bit issue.

I want to disclose how much money we invested first.
20 x XU4 = $1,180
20 x 8G uSD = $160
20 x LAN cable = $10
1 x 5V/80A PSU = $45
1 x 24 Port network Switch = $41
Cables/plugs/PCB spacers are $30
We invested $1,500 approximately.
Ah.. an optional AC power gauge meter was $30.

We measured the power rate and its accumulated power is 65KWh and monthly power consumption could be 130KWh.
So the estimated monthly electricity cost will be around $15 in the US probably. Let me know if I'm wrong.

But we've earned only 100 VRM in two weeks.
If we consider some exchange fees for changing VRM-to-BTC-to-US$, its actual value is $90 at this moment.
So our estimated monthly income will be $90 x 2 - $15 = $165.

Therefore we have to keep running this mysterious Kernel stability test equipment for 9 months to get the ROI (Return on Investment).
I am not sure if it is worth or not.

BTW, we also learned how much XU4 is faster than RPi3 again.
The average hash rate (Hashes per minute) on XU4 is 390H/min while RPi3 is 110H/min only.
Yes, our XU4 computing power is 3.5 times higher than RPi3 as expected.

Anyway, if you are interested in the mining software setup and internal algorithms, refer below links.
Reference Articles :
* CPU Mining is back! A complete how to guide and profit analysis for Verium mining on a farm of single board computers - Part 1: https://goo.gl/XM5ype
* CPU Mining is back! A complete how to guide and profit analysis for Verium mining on a farm of single board computers - Part 2: https://goo.gl/c38MWj
* CPU Mining is back! A complete how to guide and profit analysis for Verium mining on a farm of single board computers - Part 2b: https://goo.gl/4Gi8ax
* Wallet Download: http://www.vericoin.info/
* Miner: https://github.com/effectsToCause/veriumMiner
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby linuxest » Sat Jun 03, 2017 1:41 pm

very fun mining farm for the kernel testing. You have to try 500 or more xu4 boards to check the system scalability. If 500 xu4 boards work stably with such profit rate, you don't need to sell the xu4 boards any more.
I believe it must be more profitable than selling your boards. but don't forget the cryptocurrency or virtual currency market is very risky. it is a sort of gambling not a science engineering.
anyway, this is a really good marketing activity since nobody hates making money. ;)
linuxest
 
Posts: 37
Joined: Tue Nov 08, 2016 6:35 pm
languages_spoken: english
ODROIDs: 1 x HC1 and 2 x C2s with HiFi-Shield 2
And some RPi3 boards

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby memeka » Sat Jun 03, 2017 5:15 pm

@odroid

have you looked for GPU miners that might work on XU4?
User avatar
memeka
 
Posts: 3731
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Sat Jun 03, 2017 5:32 pm

@memeka,
No. We don't expect the Mali GPU mining since the nVidia GPUs must be much more efficient.
Fortunately, Verium is not a GPU/FPGA friendly algorithm at all.

@linuxest,
Don't worry. we will keep selling ODROID boards. We don't like to take any risk. ;)
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby phaseshifter » Sat Jun 03, 2017 7:26 pm

that is an expensive outlay would a personal quote of parts that i require from h-k be at all possible...???? should i mail to odroid@harkenel.com for advice ..
20x xu4 ... 1,180$
20x usd 8gb cards 150$
20 x power cable 25$

1355$ usd to aud =1820 aud
quite a fair deal of outlay..i would like to do it with c-2`s but i am not exp enough to code and set up this project..

this is with an out look of re-purchase every 6 months approx at the same amount of hardware

can it be setup using xu4 minimal image...????
4.9.xx.xx odroids
phaseshifter
 
Posts: 2485
Joined: Fri May 08, 2015 9:12 am
languages_spoken: english
ODROIDs: opp sys.. ubuntu .kernel 4.9.xx.xx.c1+ ..c-2..xu3 lite,xu4...vu7,vu8,c-shell -II..c-shel-II,uart,hi-fi ,,hi-fi2,,show,w-board,6x16GB emmc`s 3.5 inch touch...other odroid acc`s as well

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby memeka » Sat Jun 03, 2017 9:14 pm

@odroid i saw there's some miners (bitcoin i think?) that use pyopencl, which works with the XU4...
using the GPU as well even if it's not as good as CPU for verium, is still not bad since it can only increase hashes :)
User avatar
memeka
 
Posts: 3731
Joined: Mon May 20, 2013 10:22 am
languages_spoken: english
ODROIDs: XU rev2 + eMMC + UART
U3 + eMMC + IO Shield + UART

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby meveric » Sun Jun 04, 2017 1:20 am

I also think using GPU might be worth a try. the GPU of the XU3/XU4 is very powerful and will all cores is over 100GFlop while this page indicates the GFlops of a Cortex-A15 although kinda the highest in the tests are way below these 100GFlops that the GPU is listed with.
I'd just like to know how the GPU would perform, just to have a comparison not for actual mining ;)
Donate to support my work on the ODROID GameStation Turbo Image for U2/U3 XU3/XU4 X2 X C1 as well as many other releases.
Check out the Games and Emulators section to find some of my work or check the files in my repository to find the software i build for ODROIDs.
If you want to add my repository to your image read my HOWTO integrate my repo into your image.
User avatar
meveric
 
Posts: 7870
Joined: Mon Feb 25, 2013 2:41 pm
languages_spoken: german, english
ODROIDs: ODROID-X2,ODROID-U2,ODROID-U3,ODROID-XU-Lite, ODROID-XU3, ODROID-XU3-Lite, ODROID-C1, ODROID-XU4, ODROID-C2

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby mad_ady » Sun Jun 04, 2017 1:36 am

When I tested WPA2 password cracking last year, using the GPU on the XU4 caused the performance to almost double. But I expect the numbers depend on the workload.
User avatar
mad_ady
 
Posts: 2756
Joined: Wed Jul 15, 2015 5:00 pm
Location: Bucharest, Romania
languages_spoken: english
ODROIDs: XU4, C1+, C2

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Sun Jun 04, 2017 10:42 am

OpenCL and Shader language are very far from our field.

@mad_ady,
Did you use the OpenCL to test the WPA2 password cracking performance?
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Sun Jun 04, 2017 1:05 pm

phaseshifter wrote:that is an expensive outlay would a personal quote of parts that i require from h-k be at all possible...???? should i mail to odroid@harkenel.com for advice ..
20x xu4 ... 1,180$
20x usd 8gb cards 150$
20 x power cable 25$

1355$ usd to aud =1820 aud
quite a fair deal of outlay..i would like to do it with c-2`s but i am not exp enough to code and set up this project..
this is with an out look of re-purchase every 6 months approx at the same amount of hardware
can it be setup using xu4 minimal image...????


Just try to add the items in your cart and check the price on our e- Store. The total price will be very close to what I calculated.
We've run the mining software on our XU4 Ubuntu minimal image. Mate version should be okay too.
We didn't try it on C2 since we are not sure whether it can run on Aarch64 or not. I bet the hash rate on XU4 must be twice faster than C2.

But don't try the mining farm if you can't understand their installation guide. We won't support it.
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby mad_ady » Sun Jun 04, 2017 2:02 pm

@odroid: I was using pyrit which mapped two little cores to the GPU. Details and performance values here: viewtopic.php?t=22488
User avatar
mad_ady
 
Posts: 2756
Joined: Wed Jul 15, 2015 5:00 pm
Location: Bucharest, Romania
languages_spoken: english
ODROIDs: XU4, C1+, C2

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby birty » Sun Jun 04, 2017 11:00 pm

Hi all - I'm the guy that wrote the guides for mining. GPU won't work for this particular alogrithm because of its high memory requirements per thread (128mb). We have had one of the guys that writes mining software for GPUs try to get better results and CPU mining is still best hash per watt and hash per pound / dollar / etc

There are some numbers for C2 and lots of other hardware performance here: https://docs.google.com/spreadsheets/d/ ... jg/pubhtml

Xu4 is still the best hash for power consumption

There is lots of help and info available in our slack channel to get things set up and there are xu4 images with miners ready to go etc.

Slack channel is here https://vericoinandveriuminvite.herokuapp.com/

I am @birty over there but lots of people willing to help!
birty
 
Posts: 16
Joined: Mon May 01, 2017 1:58 pm
languages_spoken: english

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby meveric » Sun Jun 04, 2017 11:27 pm

birty wrote:Hi all - I'm the guy that wrote the guides for mining. GPU won't work for this particular alogrithm because of its high memory requirements per thread (128mb).

That doesn't sound like an issue.128MiB/thread the XU4's GPU has 6 Cores, 6x128MB is 768 MiB RAM and the XU4 has 2GB, so I don't see what would be the issue here. Even if you use twice as many threads as GPU cores that's less RAM than the board has to offer.

I'm not saying that GPU is better or not, it still may be that the GPU is a lot worse than CPU when it comes for hashing, just the statement "it needs more RAM" sounds a little fishy to me.
Donate to support my work on the ODROID GameStation Turbo Image for U2/U3 XU3/XU4 X2 X C1 as well as many other releases.
Check out the Games and Emulators section to find some of my work or check the files in my repository to find the software i build for ODROIDs.
If you want to add my repository to your image read my HOWTO integrate my repo into your image.
User avatar
meveric
 
Posts: 7870
Joined: Mon Feb 25, 2013 2:41 pm
languages_spoken: german, english
ODROIDs: ODROID-X2,ODROID-U2,ODROID-U3,ODROID-XU-Lite, ODROID-XU3, ODROID-XU3-Lite, ODROID-C1, ODROID-XU4, ODROID-C2

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby birty » Mon Jun 05, 2017 6:17 am

meveric wrote:
birty wrote:Hi all - I'm the guy that wrote the guides for mining. GPU won't work for this particular alogrithm because of its high memory requirements per thread (128mb).

That doesn't sound like an issue.128MiB/thread the XU4's GPU has 6 Cores, 6x128MB is 768 MiB RAM and the XU4 has 2GB, so I don't see what would be the issue here. Even if you use twice as many threads as GPU cores that's less RAM than the board has to offer.

I'm not saying that GPU is better or not, it still may be that the GPU is a lot worse than CPU when it comes for hashing, just the statement "it needs more RAM" sounds a little fishy to me.

If you can make it work then that would be awesome! I didnt think the clockspeed was that high compared to the the A7 or A15 cores though? Same for all GPUs - there are lots of cores, just clocked slower. What i was trying to say was that in general a GPU wont be as good as it doesnt have enough RAM to utilize all its lower clocked cores to take advantage of its massively parallel nature; the XU4 may have enough memory for all its GPU cores (dont know how much system memory can be shared with the GPU - im new to XU4 apart from mining) but those cores will be clocked slower than the make ones?

But like i said - happy to be proved wrong and it would be awesome to get even more performance out of the XU4's!!
birty
 
Posts: 16
Joined: Mon May 01, 2017 1:58 pm
languages_spoken: english

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby meveric » Mon Jun 05, 2017 7:44 am

birty wrote:I didnt think the clockspeed was that high compared to the the A7 or A15 cores though?

One has nothing to do with the other. If that would be the case why would we need Graphics cards at all? Graphics cards are "clocked" a lot slower than CPUs The first 3DFx Voodoo graphics card was running at 50MHz that was extremely slow compared to the CPUs that were available at that time. If "clock speed" would be the only thing that matters, no one would have used graphic cards in the first place, since they are very slow compared to CPUs.

birty wrote:Same for all GPUs - there are lots of cores, just clocked slower. What i was trying to say was that in general a GPU wont be as good as it doesnt have enough RAM to utilize all its lower clocked cores to take advantage of its massively parallel nature;

First of all you can't compare x86 and arm architecture that are WORLDS that lay in between.
NVIDIA GPUs have THOUSANDS of shader cores similar things goes for AMDs ATi GPUs. They are completely different to the Mali GPU and ARM architecture.

birty wrote:the XU4 may have enough memory for all its GPU cores (dont know how much system memory can be shared with the GPU - im new to XU4 apart from mining) but those cores will be clocked slower than the make ones?

Clockspeed alone does say NOTHING about the capabilities of a CPU. There are so many differences in the ARM world, that clock speed alone means NOTHING!
The ODROID C1 has a default clock of 1536 MHz by your "just clock speed matters" comparison that would mean, they should be faster than the A7 cores of the XU4 and they also should be as fast as a C2 which also runs on 1536 MHz.
Still the C2 is probably twice as fast as the C1, in some cases even 10times as fast.
sysbench on the C1 running @1728MHz (overclocked):
Code: Select all
sysbench --num-threads=1 --test=cpu run
sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 10000


Test execution summary:
    total time:                          192.7101s
    total number of events:              10000
    total time taken by event execution: 192.6983
    per-request statistics:
         min:                                 19.16ms
         avg:                                 19.27ms
         max:                                 34.43ms
         approx.  95 percentile:              19.55ms

Threads fairness:
    events (avg/stddev):           10000.0000/0.00
    execution time (avg/stddev):   192.6983/0.00


sysbench on the C2 running @1752MHz (overclocked):
Code: Select all
sysbench --num-threads=1 --test=cpu run
sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 1

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 10000


Test execution summary:
    total time:                          8.4746s
    total number of events:              10000
    total time taken by event execution: 8.4694
    per-request statistics:
         min:                                  0.85ms
         avg:                                  0.85ms
         max:                                  2.65ms
         approx.  95 percentile:               0.85ms

Threads fairness:
    events (avg/stddev):           10000.0000/0.00
    execution time (avg/stddev):   8.4694/0.00


8.5 seconds vs 192.5 seconds that is more than 22 times faster although the clock speed is nearly identical!
But one is a Cortex-A5 processor running on armhf the other is a Cortex-A53 running on arm64. Once again specialization is everything.

The XU4's A7 cores are also a lot faster as the C1 although it's "clocked lower". Clock speed alone says NOTHING about the capabilities. It also matter how many instructions can be performed PER CLOCK.
And if one core does 1 instruction per clock and another does 2.5 per clock it doesn't matter if the MHz is "lower" in the end it will still be faster!

GPUs are specialized processors that are made to perform specific tasks very fast.
CPUs HAVE to be universal, cause they are suppose to run a lot of different tasks run programs of all kinds and do a lot of different operations, they are NOT as specialized as GPUs are. Therefore they have to be a lot faster to get the same kind of performance.
Just as an example, there are super computers that consists of thousands of graphics cards, only cause they are a HELL LOT FASTER for some specific cases than regular CPUs are. For example to crack passwords and encryptions in general it's better to use GPUs rather than CPUs.

You see similar things between the ARM and the x86 worlds. ARM processors SUCK at calculating double values but they are very good at calculating floats. "specialization" is the magic word.

The link that I showed earlier also shows this nicely:
Add (fp32) -> Cortex-A15 @1.55GHz = 1.69 GFlops, so since our ODROIDs run at 2GHz it's probably save to say we can get around 2GFlops on 32bit floating point addition operations.
Now 2 GFlops is not very much, that's rather low. It gets better if there's some specific hardware components used, like NEON optimization and such, IF your program actually supports this.
Still the values are rather slow. At highest probably 15 GFlops per core, so a total of 60 GFlops in best case. For that specific operation!
The Mali T628MP6 is tested to have over 100 GFlops -> FOR A SPECIAL CASE it was made for (that is nearly twice as fast).

The question here is if the GPUs can do the hashing better than the CPU can, or is the specialization of the GPUs slower than the CPUs.

birty wrote:But like i said - happy to be proved wrong and it would be awesome to get even more performance out of the XU4's!!

And that's exactly what I'm pointing at.. It should be tried, to see what the differences are. The GPU can totally suck on hashing, or it can outperform the CPU hashing 1:10, we don't know if no one tries.
Donate to support my work on the ODROID GameStation Turbo Image for U2/U3 XU3/XU4 X2 X C1 as well as many other releases.
Check out the Games and Emulators section to find some of my work or check the files in my repository to find the software i build for ODROIDs.
If you want to add my repository to your image read my HOWTO integrate my repo into your image.
User avatar
meveric
 
Posts: 7870
Joined: Mon Feb 25, 2013 2:41 pm
languages_spoken: german, english
ODROIDs: ODROID-X2,ODROID-U2,ODROID-U3,ODROID-XU-Lite, ODROID-XU3, ODROID-XU3-Lite, ODROID-C1, ODROID-XU4, ODROID-C2

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby phaseshifter » Mon Jun 05, 2017 11:07 am

ok it`s like i try out a fair deal of ideas put forward on this forum..but my lack of knowledge in coding holds me back...

that said if i have the precise code that is needed i can input it and try it..

but without the code well,,half the time i just run into knots..

that said i was able to get an xu4 minning..but it throws errors and i dont know how to fix it..
also i tried an c-2 and a c-1 both failed..the make file code...
..but the xu4 was the only one that i actually got it sort of working..not to a complete stage..i used the notes from the posts above..
4.9.xx.xx odroids
phaseshifter
 
Posts: 2485
Joined: Fri May 08, 2015 9:12 am
languages_spoken: english
ODROIDs: opp sys.. ubuntu .kernel 4.9.xx.xx.c1+ ..c-2..xu3 lite,xu4...vu7,vu8,c-shell -II..c-shel-II,uart,hi-fi ,,hi-fi2,,show,w-board,6x16GB emmc`s 3.5 inch touch...other odroid acc`s as well

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby phaseshifter » Mon Jun 05, 2017 11:18 am

yes instuctions per cycle ..risc-3 v`s risc-4 etc i see where mavereic is coming from..work done is always a factor and if one chip does 3 times the work per cycle..then speed does not count rather...
4.9.xx.xx odroids
phaseshifter
 
Posts: 2485
Joined: Fri May 08, 2015 9:12 am
languages_spoken: english
ODROIDs: opp sys.. ubuntu .kernel 4.9.xx.xx.c1+ ..c-2..xu3 lite,xu4...vu7,vu8,c-shell -II..c-shel-II,uart,hi-fi ,,hi-fi2,,show,w-board,6x16GB emmc`s 3.5 inch touch...other odroid acc`s as well

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby birty » Mon Jun 05, 2017 1:46 pm

phaseshifter wrote:that said i was able to get an xu4 minning..but it throws errors and i dont know how to fix it..
also i tried an c-2 and a c-1 both failed..the make file code...


Someone said they fixed the C2 one by editing sysinfos.c to change the #ifndef __arm__ to #ifdef __arm__ don't have a C2 myself so can't test. Head over to slack if you need more help - bit easier live
birty
 
Posts: 16
Joined: Mon May 01, 2017 1:58 pm
languages_spoken: english

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby birty » Mon Jun 05, 2017 1:51 pm

@meveric sorry if I offended, wasn't my intention. Was showing my ignorance - kinda thought you = lots of cores clocked slower working in parallel. Will go read up on Mali, as I said I'm new to arm really - just started using them for this application and was reporting what I and others have said. Didn't mean to give the impression I thought you were wrong
birty
 
Posts: 16
Joined: Mon May 01, 2017 1:58 pm
languages_spoken: english

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby phaseshifter » Mon Jun 05, 2017 3:02 pm

@ birty the c2 threw atomic pointers needed issue i will try with your code...which is your forum so i can join pls
4.9.xx.xx odroids
phaseshifter
 
Posts: 2485
Joined: Fri May 08, 2015 9:12 am
languages_spoken: english
ODROIDs: opp sys.. ubuntu .kernel 4.9.xx.xx.c1+ ..c-2..xu3 lite,xu4...vu7,vu8,c-shell -II..c-shel-II,uart,hi-fi ,,hi-fi2,,show,w-board,6x16GB emmc`s 3.5 inch touch...other odroid acc`s as well

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Mon Jun 05, 2017 4:49 pm

@phaseshifter
Please open a new thread in C2 Project sub-form.
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby meveric » Mon Jun 05, 2017 5:15 pm

birty wrote:@meveric sorry if I offended, wasn't my intention. Was showing my ignorance - kinda thought you = lots of cores clocked slower working in parallel. Will go read up on Mali, as I said I'm new to arm really - just started using them for this application and was reporting what I and others have said. Didn't mean to give the impression I thought you were wrong

I'm not offended. I'm just curious of the results. The GPU was a BEAST in it's time. Back then there weren't many OpenGL ES 3.x devices out there. The SoC was Samsungs flagship and designed to exceed in any possible way.
I'm just curious how good it might actual be.
There are a couple possible scenarios to look out for like:
* The GPU sucks for hasing and doesn't help at all
* The GPU is extremly powerful in hashing and outshines the CPUs power
* The GPU is moderate in hashing and might assist the CPU
* The GPU is moderate in hashing but uses way less energy than the CPU and therefore can be used as a replacement to save energy costs.
* The GPU can't be used at all for hashing, since no code exist to use the GPU

As I said, it's just an interesting topic.
Imagine the GPU just being moderately good, and does 200-250 hashes rather than over 300 as the CPU, but the GPU uses less energy than the CPU, so the value/time ratio might just better due to lower energy cost or something.
Or the GPU can assist the CPU to generate even more hashes (maybe the XU4 goes over 400 or 500 hashes if the GPU assists the CPU).
Donate to support my work on the ODROID GameStation Turbo Image for U2/U3 XU3/XU4 X2 X C1 as well as many other releases.
Check out the Games and Emulators section to find some of my work or check the files in my repository to find the software i build for ODROIDs.
If you want to add my repository to your image read my HOWTO integrate my repo into your image.
User avatar
meveric
 
Posts: 7870
Joined: Mon Feb 25, 2013 2:41 pm
languages_spoken: german, english
ODROIDs: ODROID-X2,ODROID-U2,ODROID-U3,ODROID-XU-Lite, ODROID-XU3, ODROID-XU3-Lite, ODROID-C1, ODROID-XU4, ODROID-C2

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby hominoid » Mon Jun 05, 2017 11:36 pm

meveric wrote:...Or the GPU can assist the CPU to generate even more hashes (maybe the XU4 goes over 400 or 500 hashes if the GPU assists the CPU).

Already solidly over 400h/m on released kernel 3.10.104+ running in the mid to high 70's c :D working on more...
XU4 Split Airflow Mining.png
XU4 Split Airflow Mining.png (64.3 KiB) Viewed 61682 times
hominoid
 
Posts: 72
Joined: Tue Feb 28, 2017 3:55 am
Location: Lake Superior Basin, USA
languages_spoken: english
ODROIDs: XU4

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby jconsole » Tue Jun 06, 2017 1:35 am

Can you state how much power one unit consumed in an hour. Or how much current was the unit drawing.

i am using an Rpi2 and Rpi3 in a land based drone. attention to power consumption in a mobile environment is a priority. thanks
jconsole
 
Posts: 3
Joined: Tue Jun 06, 2017 1:30 am
languages_spoken: english

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby gigoplast » Tue Jun 06, 2017 8:19 am

Who switch better and more efficient to 40 xu4 an Netgear ProSafe 48 Port 10/100/1000 Smart Switch GS748T V4 or Netgear 48 Port 10/100 Switch ?

The Netgear 10/100 is more shiper in price and the 10/100/1000 is more expensive, but i don't know if there are no problem at the current level with the ethernet in xu4 board? Gigabit switch can handle at better performence the cluster than a megabit switch ?
gigoplast
 
Posts: 3
Joined: Fri Jun 02, 2017 1:52 pm
languages_spoken: english
ODROIDs: xu4

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Tue Jun 06, 2017 10:15 am

@hominoid
Yes, we could see over 410H/m when the cooling efficiency is higher.
We might need to consider a liquid cooling system something like this to keep that hash rate. :o


@jconsole
When we run the Verium mining software, XU4 consumes around 2.5Amp.
There is no USB device attached.

@gigoplast
100Mbps switch is more than enough for the Verium mining.
But the Gbit switch must be better for the cluster application because XU4 Gbit throughput is near 900Mbps.
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby linuxest » Tue Jun 06, 2017 10:55 am

Do you have any plan to make another XU4 variant of [stripped-down] version something like XU0?
No HDMI, No eMMC, No USB 3.0 port, No GPIO ports. it will reduce the board size and material cost significantly.
It will be absolutely suitable for the clustering and mining application.
I saw your latest mainline u-boot can support the Ethernet port and you can use the TFTP & PXE booting for much easier maintenance with cheap 1~2GB micro-SD cards.
linuxest
 
Posts: 37
Joined: Tue Nov 08, 2016 6:35 pm
languages_spoken: english
ODROIDs: 1 x HC1 and 2 x C2s with HiFi-Shield 2
And some RPi3 boards

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby gigoplast » Tue Jun 06, 2017 11:09 am

PSU FOR CLUSTER ??
Can i use ATX desktop power supply to run 10 or * xu4 without damage the board ?
what's better reference desktop atx *watt ?
I prefer desktop power supply because is shipper than lrs-200-5 form me , atx i can shipped locally :D , The lrs-200-5 cost me Triple price on shipping from Internet , I'm from north Africa and delivery is expensive to me . :oops:
exp 400 watt atx =
https://sgcdn.startech.com/005329/media ... 0PRO.C.jpg
finally how make the connection between them with better solution .
Thank's for every clarification
gigoplast
 
Posts: 3
Joined: Fri Jun 02, 2017 1:52 pm
languages_spoken: english
ODROIDs: xu4

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Tue Jun 06, 2017 11:26 am

@gigoplast
ATX power supply has 5V/3.3V/12V/-12V output rails. So it is not easy to confirm if the 5Volt rail output current can be 40Amp or not.
You have to check the 5V rail specification of the ATX power.

@linuxest
That's a good idea. I think the XU0 can be $50 if we can remove all of those features.
But I am not sure whether we have enough market volume for the ARM cluster computing application or not.

Anyway, please stop writing off topic posts.
Open a new thread in the XU4 sub-forum if you need further discussion.
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby mad_ady » Thu Jun 08, 2017 5:01 pm

@odroid: I have an idea to reduce your mining investment cost. You probably have a case of broken odroid xu3/4s from all sorts of RMAs. If you use the ones with the broken USB bus (and broken network, sdcard), but with working emmc and serial, you can use them for hashing by converting the onboard serial adapter to a network interface with ppp.
Details about ppp networking are here:

viewtopic.php?t=25527
http://www.instructables.com/id/Connect ... sing-UART/

You would need 1 serial/usb adapter and 1 emmc (since those with fried USB bus usually can't boot from sd anymore) and you would need a working odroid (either a xu4 miner, or a cheaper c1) to aggregate 3/4 broken miners via USB and act as a router with NAT. This way, the boards can use networking and report their results and still earn you money. And you could sell the functional XU4s once replaced, thus reducing your investment.

By the way, what do you guys do with broken hardware that is returned? I hope you don't just throw it away (I can't throw away anything even if it's broken - seems I'm a hoarder).
User avatar
mad_ady
 
Posts: 2756
Joined: Wed Jul 15, 2015 5:00 pm
Location: Bucharest, Romania
languages_spoken: english
ODROIDs: XU4, C1+, C2

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Thu Jun 08, 2017 7:22 pm

Normally we are throwing away the returned boards if we can't repair it within 10~20 minutes.
Failure analysis and BGA rework process takes too long time and they are very expensive (more expensive than board cost).

Anyway, it is a good idea to recycle the returned boards if any damaged block can be overcome with software solution like the PPP network.
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby mad_ady » Thu Jun 08, 2017 8:39 pm

odroid wrote:Normally we are throwing away the returned boards if we can't repair it within 10~20 minutes.


You monsters! I need to arrange a dumpster diving session next time I'm in South Korea :lol:
User avatar
mad_ady
 
Posts: 2756
Joined: Wed Jul 15, 2015 5:00 pm
Location: Bucharest, Romania
languages_spoken: english
ODROIDs: XU4, C1+, C2

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby hominoid » Sat Jun 10, 2017 12:15 pm

odroid wrote:@hominoid
Yes, we could see over 410H/m when the cooling efficiency is higher.
We might need to consider a liquid cooling system something like this to keep that hash rate. :o

XU4 with stock cooler running at 1.9ghz...No mineral oil needed :ugeek:
Screen-1.png
Screen-1.png (62.35 KiB) Viewed 60465 times
hominoid
 
Posts: 72
Joined: Tue Feb 28, 2017 3:55 am
Location: Lake Superior Basin, USA
languages_spoken: english
ODROIDs: XU4

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby Anoy74 » Sat Jun 10, 2017 8:40 pm

how??? :o
Anoy74
 
Posts: 1
Joined: Mon Jan 23, 2017 6:38 am
languages_spoken: english

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby mad_ady » Sat Jun 10, 2017 9:34 pm

Check out hominid's article in this month odroid magazine issue :)
User avatar
mad_ady
 
Posts: 2756
Joined: Wed Jul 15, 2015 5:00 pm
Location: Bucharest, Romania
languages_spoken: english
ODROIDs: XU4, C1+, C2

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby birty » Sun Jun 11, 2017 3:37 pm

But that last post says with stock cooler. Looks like hominid had done some great work!

Having read the article as well - I'm off to make some copper shims. Great how much knowledge is shared in this community!
birty
 
Posts: 16
Joined: Mon May 01, 2017 1:58 pm
languages_spoken: english

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby phaseshifter » Mon Jun 12, 2017 9:18 am

getting an average of 395-405 hash 12 threads with copper shim sink compound and an extra small 5 volt fan

my c-2 does 198-200 constant 12 threads clocked at 1752

code for neon for C-2 is
Code: Select all
ARMv8=neon
Last edited by phaseshifter on Mon Jun 19, 2017 5:44 am, edited 3 times in total.
4.9.xx.xx odroids
phaseshifter
 
Posts: 2485
Joined: Fri May 08, 2015 9:12 am
languages_spoken: english
ODROIDs: opp sys.. ubuntu .kernel 4.9.xx.xx.c1+ ..c-2..xu3 lite,xu4...vu7,vu8,c-shell -II..c-shel-II,uart,hi-fi ,,hi-fi2,,show,w-board,6x16GB emmc`s 3.5 inch touch...other odroid acc`s as well

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Mon Jun 12, 2017 11:04 am

hominoid wrote:
odroid wrote:@hominoid
Yes, we could see over 410H/m when the cooling efficiency is higher.
We might need to consider a liquid cooling system something like this to keep that hash rate. :o

XU4 with stock cooler running at 1.9ghz...No mineral oil needed :ugeek:
Screen-1.png


I could see only 410~420H/m Max with the stock cooler.
How did you get over 500H/m? :o
Are you living in Alaska?
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby hominoid » Tue Jun 13, 2017 4:35 am

-.png
-.png (263.48 KiB) Viewed 57814 times
hominoid
 
Posts: 72
Joined: Tue Feb 28, 2017 3:55 am
Location: Lake Superior Basin, USA
languages_spoken: english
ODROIDs: XU4

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Tue Jun 13, 2017 7:43 am

The most amazing picture in this year! :o
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby obroken » Tue Jun 13, 2017 8:48 pm

Do you guys have a smart person I can chain in my basement with an xbox and and unlimited supply of skittles? Native born American brains are just too lazy to figure all of this stuff out. I really want to build one of these coin mining networks, probably for etherium, but omg I haven't been in the Linux stuff for 20 years and it's really not all coming back to me now. Plus everything is different. I wrote my first website in perl 4, and even published a book with wiley on website automation, but you guys are lightyears ahead.
obroken
 
Posts: 24
Joined: Fri Jun 09, 2017 10:01 am
languages_spoken: english
ODROIDs: Xu4

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby phaseshifter » Tue Jun 13, 2017 10:52 pm

@obroken here is a great place to start...

https://vericoinandveriuminvite.herokuapp.com/
4.9.xx.xx odroids
phaseshifter
 
Posts: 2485
Joined: Fri May 08, 2015 9:12 am
languages_spoken: english
ODROIDs: opp sys.. ubuntu .kernel 4.9.xx.xx.c1+ ..c-2..xu3 lite,xu4...vu7,vu8,c-shell -II..c-shel-II,uart,hi-fi ,,hi-fi2,,show,w-board,6x16GB emmc`s 3.5 inch touch...other odroid acc`s as well

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby obroken » Tue Jun 13, 2017 11:28 pm

Yea I joined the slack board but I don't know when I'll have time. Plus Vericoin is not one I've heard about. I think the ship has sailed on mining BTC, but Etherium is pretty extensive in it's adoption of late. The founder met with Putin the other day as a potential currency in place of gold once the dollar goes away.
obroken
 
Posts: 24
Joined: Fri Jun 09, 2017 10:01 am
languages_spoken: english
ODROIDs: Xu4

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby pnosker » Sat Jun 17, 2017 4:36 am

Verium/VeriCoin co-creator and developer here, thanks to Hardkernel for the XU4! I bought 20 of them and have been mining for a bit. Great ROI so far.
pnosker
 
Posts: 1
Joined: Sat Jun 17, 2017 4:35 am
languages_spoken: english
ODROIDs: 20 XU4

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Mon Jun 19, 2017 5:55 pm

Nice to meet the Verium co-creator in this forum. :D
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Wed Aug 02, 2017 4:55 pm

We made this monstrous cluster a month ago. It has 1600 CPU cores and 400GB RAM. :o
Image
200pcs of XU4 are running together and its uptime is around 30 days.
All of them are working well so far and we can say again the Kernel 4.9 stability is quite acceptable.
It is a very useful test machine even we invested $15,000 approx.
We also ported the mainline u-boot to implement the PXE/TFTP remote booting for easier maintenance.
We will release the mainline u-boot for XU4 users within a few weeks.


Anyway, is it really worth to build and run the Verium miner with ODROID?
Probably NOT!
It has consumed ~1200KWh over the last 30 days and it found ~250 Veriums only even we made it 10 times bigger.
The probability of finding a Verium is much much lower than in June.
The price of Verium is around $1.5 today. A couple of weeks ago it jumped to $4.5. A few days ago it was around $1.1 only.
It has been really unstable and not predictable at all like a gambling.

Anyway, let's estimate the ROI again.
VRM250 * $1.5 = $375
Electricity cost = $150
We made only $225.

Right! We need to run it over 5 years to reach the ROI point. We are completely ruined. :D
:evil: DON'T TRY TO BUILD A MINER MACHINE WITH ODROID :evil:
If my calculation is wrong, please let me know.

We may need to find other application like an amazing dist-cc machine to have a fun with "gcc -j1600" or a Docker-Swarm platform.
Any idea will be appreciated.
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby mad_ady » Thu Aug 03, 2017 1:18 am

You could use it for boinc and advance science and technology, but you'd have a harder time recuperating your money.
You could crack passwords with it :). Might be a faster way to make a buck. Though a gpu is worth about 150 XU4s in processing power.
User avatar
mad_ady
 
Posts: 2756
Joined: Wed Jul 15, 2015 5:00 pm
Location: Bucharest, Romania
languages_spoken: english
ODROIDs: XU4, C1+, C2

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby wtarreau » Thu Aug 10, 2017 11:05 pm

odroid wrote:We may need to find other application like an amazing dist-cc machine to have a fun with "gcc -j1600" or a Docker-Swarm platform.
Any idea will be appreciated.

Yep, like this : http://wiki.ant-computing.com/Choosing_a_processor_for_a_build_farm :-)
I suspect the XU4 can be on par with the MiQi and its RK3399 for build performance. I'd probably disable the A7 cores to avoid slowing down on large files though.
wtarreau
 
Posts: 5
Joined: Thu Jan 21, 2016 1:22 am
languages_spoken: english, french

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby elatllat » Fri Aug 11, 2017 8:02 pm

odroid wrote:...the Kernel 4.9 stability is quite acceptable...

That is a very broad statement for a very narrow test.
Anyone run https://github.com/autotest/autotest yet?
I'm happy with 4.9 but last I checked it's quite crashable if you saturate USB or go looking for issues with a broad test, some are easily fixed by just merging upstream.
elatllat
 
Posts: 618
Joined: Tue Sep 01, 2015 8:54 am
languages_spoken: english
ODROIDs: XU4

Re: XU4 Kernel 4.9 stability test with Coin mining software

Unread postby odroid » Sat Aug 12, 2017 10:54 am

@elatllat,

Can you tell me more about the USB issue?
We've already tested it with very heavy USB traffics on Ethernet and Storage simultaneously by 30TB transferring for 4 days.
viewtopic.php?f=146&t=26016&start=250#p198451

BTW, we have a plan to release a new u-boot with Kernel 4.9.4x in a week.
http://www.hardkernel.com/?b188
User avatar
odroid
Site Admin
 
Posts: 25018
Joined: Fri Feb 22, 2013 11:14 pm
languages_spoken: English
ODROIDs: ODROID

Next

Return to General Chat

Who is online

Users browsing this forum: mrxiong and 4 guests