A few remarks about the MC1 solo

Post Reply
o_lampe
Posts: 174
Joined: Wed Jan 01, 2020 12:35 am
languages_spoken: english german
ODROIDs: MC1 solo
Has thanked: 11 times
Been thanked: 16 times
Contact:

A few remarks about the MC1 solo

Unread post by o_lampe » Mon Jan 27, 2020 2:00 am

Hi folks,
I finally got my 8 MC1 solos( late, but it was my fault )
I've installed the Docker swarm image based on 16.04, but I'm not happy with it. But that's not the subject, I want to discuss.

Three of the nodes are fed by an ATX powersupply and they are stresstested by BOINC workloads. The first thing I noticed, was the "pumping" noise of the PSU.
The workload was on 7 cores at 60% on all nodes and it seems, the cores all switch off and on at the same time.
All task-schedulers seem to be synced with the RTC? If so, there must be a better solution. ( imagine a 100 node cluster on a single PSU : :o )
The pumping sound was gone when I changed the load on each node to 67%,71%, 73% . These are all primenumbers, so they will never sync again.

Then I wrote a little script to 'watch' all four thermal_zones on each node. It seems, that zone 2 + 3 on all nodes is 8-10°C hotter, than zone 0 + 1 ?!
Did anyone else notice that?
I wouldn't be surprised if 2+3 are placed where the SoC is closest to the PCB-edge ( less cooling from the GND-heatspreader layer ).
Did I mention I'm a hardware guy? ;)
Or is there another explanation? Maybe uneven pressure between the chip-dye and the heatsink? These three springloaded PCB-holders aren't placed well, force-vector wise...

The third observation is the heartbeat LED: It's very well hidden, I can only see it from a very narrow angle....

Then there is a very odd task balance on one of the nodes. On the picture it looks like 'all' cores are working on one thread?! Weird...maybe a BOINC issue
.
async_boinc.jpg
async_boinc.jpg (78.35 KiB) Viewed 530 times
.
.
Forgive me, if I sound like a Troll, I just wanted to share my 1st impression. Maybe it helps on the next PCB-project?
Olaf
These users thanked the author o_lampe for the post:
odroid (Tue Jan 28, 2020 9:55 am)

o_lampe
Posts: 174
Joined: Wed Jan 01, 2020 12:35 am
languages_spoken: english german
ODROIDs: MC1 solo
Has thanked: 11 times
Been thanked: 16 times
Contact:

Re: A few remarks about the MC1 solo

Unread post by o_lampe » Mon Jan 27, 2020 4:02 pm

Here I add some pics of my current cooling rig for testing. The stacked MC1 cluster might have problems with temps of the lowest node, I don't know?
I decided to use 120x20mm fans, because they were laying around. I arranged the solos in a quare pattern ( they lock in nicely! ) and made a fan duct from cardboard to guide the air along the heatsinks. With rubberband and toothpicks the whole thing's kept together, I felt like MacGyver ;)
2020-01-24 19.13.43.jpg
2020-01-24 19.13.43.jpg (98.86 KiB) Viewed 492 times
2020-01-24 15.10.34.jpg
2020-01-24 15.10.34.jpg (107.59 KiB) Viewed 492 times
2020-01-26 20.09.56.jpg
2020-01-26 20.09.56.jpg (110.14 KiB) Viewed 492 times

User avatar
mad_ady
Posts: 7572
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: 396 times
Been thanked: 300 times
Contact:

Re: A few remarks about the MC1 solo

Unread post by mad_ady » Mon Jan 27, 2020 4:18 pm

Good job
Image
But won't the rubber bands age quickly because of the heat and snap?
I'm sure a 21st century MacGuyver would quickly 3d print a gadget to get him out of that mess.

o_lampe
Posts: 174
Joined: Wed Jan 01, 2020 12:35 am
languages_spoken: english german
ODROIDs: MC1 solo
Has thanked: 11 times
Been thanked: 16 times
Contact:

Re: A few remarks about the MC1 solo

Unread post by o_lampe » Mon Jan 27, 2020 4:29 pm

Thanks for the compliments!
The heatsinks never get hot, sadly...good for the rubberband, but bad for the longevity of the SoC. I can see the temps jump up to 90°C but the heatsink stays 'warm'. Hopefully watercooling with a chunk of copper will help eventually.


BTW: I have three 3D-printers at hand, all build by myself. But the test rig is a temporary solution, no need to waste filament ;)

o_lampe
Posts: 174
Joined: Wed Jan 01, 2020 12:35 am
languages_spoken: english german
ODROIDs: MC1 solo
Has thanked: 11 times
Been thanked: 16 times
Contact:

Re: A few remarks about the MC1 solo

Unread post by o_lampe » Tue Jan 28, 2020 3:39 pm

I rearranged the heatsinks. Now they build a smaller quare. Will see today, if that makes any changes regarding temps...
.
2020-01-27 18.23.15.jpg
2020-01-27 18.23.15.jpg (101.14 KiB) Viewed 430 times
2020-01-27 18.26.57.jpg
2020-01-27 18.26.57.jpg (106.29 KiB) Viewed 430 times
These users thanked the author o_lampe for the post (total 2):
odroid (Tue Jan 28, 2020 4:07 pm) • joy (Thu Feb 13, 2020 10:40 am)

Post Reply

Return to “Issues”

Who is online

Users browsing this forum: No registered users and 6 guests