mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   GPU Computing (https://www.mersenneforum.org/forumdisplay.php?f=92)
-   -   Radeon VII @ newegg for 500 dollars US on 11-27 (https://www.mersenneforum.org/showthread.php?t=24979)

ewmayer 2020-04-30 21:15

Traded PMs with the seller of the ryzen5+mobo combo, he said no onboard gfx, no clue as to why video-out not working with my Radeon VII (and I ended up trying 2 of the 3 R7s I bought to install in this system). Thanks to the nearly-month-long shipping delay with the testbench frame I bought as a chassis for this build we are past the 30-day return window, but seller kindly approved a late return. Bugger, big PITA for us both, it really looks like a sweet little system, was perfect for my needs.

So later today, the hunt for "MoBo/CPU Bundle 2: The Final Insult" begins.

preda 2020-04-30 22:15

[QUOTE=ewmayer;544316]no clue as to why video-out not working with my Radeon VII[/QUOTE]

you did try different PCIe slots on the motherboard?

ewmayer 2020-04-30 22:44

[QUOTE=preda;544323]you did try different PCIe slots on the motherboard?[/QUOTE]

No, because I paid for a mobo with 3 full-length pcie slots, and that means on this one I need all 3 working. Disassembly complete, Return label created, way more than enough time wasted d*cking around with this system, time to move on.

ewmayer 2020-05-01 00:05

How about something along these lines? CPU speed again not not crucial, I can use thrle 2 sticks of ddr4 I bought and the 3 full-length PCI slots again laid out similarly as the first mobo I tried, i.e. 2 slots nearest CPU could each host a Radeon 7 as-is, 3rd slot could use the right-angle adapter I ordered to give its GPU free cooling-air path:

[url]https://www.ebay.com/itm/MSI-Z270-SLI-Plus-Socket-1151-Motherboard-with-a-Celeron-G3900-2-8ghz-CPU/223991583279[/url]

Here is link to the mobo homepage - not clear to me whether that includes onboard video or not - shouldn't matter since I ca use on of the GPUs for gfx, but based on the #fail with my first attempt, it would be nice to be able to do basic setup, OS install without a GPU plugged in:

[url]https://www.msi.com/Motherboard/Z270-SLI-PLUS.html[/url]

paulunderwood 2020-05-01 00:59

[QUOTE=ewmayer;544329]How about something along these lines? CPU speed again not not crucial, I can use thrle 2 sticks of ddr4 I bought and the 3 full-length PCI slots again laid out similarly as the first mobo I tried, i.e. 2 slots nearest CPU could each host a Radeon 7 as-is, 3rd slot could use the right-angle adapter I ordered to give its GPU free cooling-air path:

[url]https://www.ebay.com/itm/MSI-Z270-SLI-Plus-Socket-1151-Motherboard-with-a-Celeron-G3900-2-8ghz-CPU/223991583279[/url]

Here is link to the mobo homepage - not clear to me whether that includes onboard video or not - shouldn't matter since I ca use on of the GPUs for gfx, but based on the #fail with my first attempt, it would be nice to be able to do basic setup, OS install without a GPU plugged in:

[url]https://www.msi.com/Motherboard/Z270-SLI-PLUS.html[/url][/QUOTE]

The board has HDMI. The processor has graphics capability. You should get a signal without the Radeon.

ewmayer 2020-05-12 23:23

Update on the multi-GPU test-frame build is "finally getting somewhere", but that is preceded by a loud "oh dear lord". First to the latter...

So after the rigmarole of repacking mobo/cpu bundle #1 and returning to seller, bundle #2 - the dual-core-Celeron (sse2 only) cheapie, basically a very nice mobo with a minimal CPU - arrived late last week, I unpacked and assembled things and DLed the mobo manual over the weekend, then tried it out. Ruh-roh ... same issue as before - all the diagnostic LEDs indicate all is well, but no video-out signal. This mobo has onboard gfx, had first tried that, then installed one of the still-waiting-for-a-home Radeon VII trio I'd bought for this build. Still nothing. So at that point I start looking for bad video-out stuff of the 'wtf?' variety, and eventually found it in form of the little 6"-long vga-to-hdmi adapter cable I was using. Have used that for last several years to be able to connect to my wifi-less little Odroid C2 for Mlucas ARM builds, but had retired that system (except for future build-only purposes) a few months ago, and just pulled the adapter cable from it for use with the newer mobos which lack vga ports. To see if it might be the adapter, unboxed the Odroid and hooked up as before - no video signal. Tried using the monitor-attached vga cable with the hdmi adapater in my Haswell system (which has a vga port, i.e. normally uses just the vga cable) - no video signal. Somewhere between my final Odroid use and now, the little piece of sh*t just decided to stop functioning.

This is intensely irritating, to have such a basic, unpowered component simply go flaky from one day to the next, with no obvious signs of wear or damage, resulting in a multi-week delay and many hours of wasted time and frustration. Good grief.

Anyhow, moving forward, no use crying over spilled milk, water under the bridge, etc - gonna contact the seller of mobo #1, apologize and explain the cable issue, and offer to refund at least his return-shipping expense. Mobo #2 cost a bit less than #1, but further annoying in that the CPU is vastly inferior (#1 was 8-core Ryzen5 with avx2 support, #2 is 2-core Celeron with sse2, no avx-anything). But mybe better to not be tempted to use the CPU for GIMPS work, because in either case the watts-per-FLOP pales versus the expected R7 throughput.

So trying to install gpuOwl, hit a snag:
[code]git clone https://github.com/preda/gpuowl && cd gpuowl && make
Cloning into 'gpuowl'...
remote: Enumerating objects: 107, done.
remote: Counting objects: 100% (107/107), done.
remote: Compressing objects: 100% (88/88), done.
remote: Total 5126 (delta 58), reused 41 (delta 19), pack-reused 5019
Receiving objects: 100% (5126/5126), 12.59 MiB | 7.08 MiB/s, done.
Resolving deltas: 100% (3684/3684), done.
./tools/expand.py < gpuowl.cl > gpuowl-expanded.cl
cat head.txt gpuowl-expanded.cl tail.txt > gpuowl-wrap.cpp
echo \"`git describe --long --dirty --always`\" > version.new
diff -q -N version.new version.inc >/dev/null || mv version.new version.inc
echo Version: `cat version.inc`
Version: "v6.11-278-ga39cc1a"
g++ -MT Pm1Plan.o -MMD -MP -MF .d/Pm1Plan.Td -Wall -O2 -std=c++17 -c -o Pm1Plan.o Pm1Plan.cpp
g++ -MT GmpUtil.o -MMD -MP -MF .d/GmpUtil.Td -Wall -O2 -std=c++17 -c -o GmpUtil.o GmpUtil.cpp
In file included from GmpUtil.cpp:3:
GmpUtil.h:6:10: fatal error: gmpxx.h: No such file or directory
6 | #include <gmpxx.h>
| ^~~~~~~~~
compilation terminated.
make: *** [Makefile:30: GmpUtil.o] Error 1[/code]
I had previously tried to install GMP like so:
[code]sudo apt install libgmp
Reading package lists... Done
Building dependency tree
Reading state information... Done
E: Unable to locate package libgmp[/code]
Then did package-search on the Ubntu site, closest I found was libgmp10:
[code]Reading package lists... Done
Building dependency tree
Reading state information... Done
libgmp10 is already the newest version (2:6.1.2+dfsg-4).
0 upgraded, 0 newly installed, 0 to remove and 264 not upgraded.[/code]
Any clues as to the GMP-incompatibility issue welcome.

Prime95 2020-05-13 00:16

sudo apt install libgmp-dev

ewmayer 2020-05-13 03:19

[QUOTE=Prime95;545200]sudo apt install libgmp-dev[/QUOTE]

Thanks - build successful, but attempt to run gives shared-lib error:
[i]
../gpuowl: error while loading shared libraries: libOpenCL.so.1: cannot open shared object file: No such file or directory
[/i]
No rush to run stuff tonight, since already clear that while used-R7 #1 is very clean, its fan noise is much louder than the new R7 I bought for my Haswell back in January. So especially with 3 cards intended, gonna need to find/create a soundproof location to put the thing.

paulunderwood 2020-05-13 03:22

@ernst, I am glad your found the source of the hardware problem albeit frustrating, especially when you could have double checked the vga-hdmi gadget before you bought the Celeron system. We have all learnt from this at your expense. If you do go for the AMD system, you might find the Celeron system useful for testing purposes at a future date.

You can now chase me up the leader board :wink:

paulunderwood 2020-05-13 03:34

[QUOTE=ewmayer;545208]Thanks - build successful, but attempt to run gives shared-lib error:
[i]
../gpuowl: error while loading shared libraries: libOpenCL.so.1: cannot open shared object file: No such file or directory
[/i]
No rush to run stuff tonight, since already clear that while used-R7 #1 is very clean, its fan noise is much louder than the new R7 I bought for my Haswell back in January. So especially with 3 cards intended, gonna need to find/create a soundproof location to put the thing.[/QUOTE]

On my Debian system which is similar to Ubuntu:

[CODE]apt-cache search libOpenCL
ocl-icd-libopencl1 - Generic OpenCL ICD Loader
nvidia-libopencl1 - NVIDIA OpenCL ICD Loader library
libopencl1-amdgpu-pro - AMD OpenCL ICD Loader library
[/CODE]

[CODE]apt list --installed | grep libopencl1

WARNING: apt does not have a stable CLI interface. Use with caution in scripts.

ocl-icd-libopencl1/stable,now 2.2.12-2 amd64 [installed,automatic]
[/CODE]

Note: don't install libopencl1-amdgpu-pro if you are using rocm.

ewmayer 2020-05-13 19:21

[QUOTE=paulunderwood;545209]@ernst, I am glad your found the source of the hardware problem albeit frustrating, especially when you could have double checked the vga-hdmi gadget before you bought the Celeron system. We have all learnt from this at your expense. If you do go for the AMD system, you might find the Celeron system useful for testing purposes at a future date.

You can now chase me up the leader board :wink:[/QUOTE]

Thanks, Paul - to clarify, the cheapie-Celeron-CPU (but nice mobo) rig #2 is what I have now - in theory I could've discovered the vga/hdmi-adapter borkage before the return-#1-and-buy-#2 rigmarole, but the possibility of such a basic component which never caused any problems before being the culprit only occurred to me once I saw the same no-video-out issue with rig #2. Basically we had to let 1 patient die in order to properly diagnose the problem in time to save patient #2.

Here the OpenCL-install info from the system:
[code]apt-cache search libOpenCL
ocl-icd-libopencl1 - Generic OpenCL ICD Loader
libopencl-clang-dev - thin wrapper for clang -- development files
libopencl-clang9 - thin wrapper for clang
nvidia-libopencl1-331 - Transitional package for nvidia-libopencl1-340
nvidia-libopencl1-331-updates - Transitional package for nvidia-libopencl1-340
nvidia-libopencl1-340 - NVIDIA OpenCL Driver and ICD Loader library
nvidia-libopencl1-340-updates - Transitional package for nvidia-libopencl1-340
nvidia-libopencl1-384 - Transitional package for nvidia-headless-390[/code]
Does that tell you anything about what might be needed or need-to-be-removed?

EDIT: Whoops, forgot the 2nd thing you said to check:
[code]apt list --installed | grep libopencl1

WARNING: apt does not have a stable CLI interface. Use with caution in scripts.[/code]
Nothing further - so did 'sudo apt install ocl-icd-libopencl1', that produces this entry in the above listing:
[code]ocl-icd-libopencl1/eoan,now 2.2.11-1ubuntu1 amd64 [installed][/code]
and solves the missing-shared-lib problem, now gpuowl starts but immediately coredumps:
[code]2020-05-13 13:31:31 gpuowl v6.11-278-ga39cc1a
2020-05-13 13:31:31 Note: not found 'config.txt'
2020-05-13 13:31:31 device 0, unique id 'df7080c172fd5d6e'
2020-05-13 13:31:31 df7080c172fd5d6e 104954387 FFT: 5.50M 1K:11:256 (18.20 bpw)
2020-05-13 13:31:31 df7080c172fd5d6e Expected maximum carry32: 50D10000
Segmentation fault (core dumped)[/code]


All times are UTC. The time now is 12:59.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.