exclude_gpu device_num with 2 AMD GPUs: download and abort loop

Message boards : GPUs : exclude_gpu device_num with 2 AMD GPUs: download and abort loop
Message board moderation

To post messages, you must log in.

AuthorMessage
sam6861

Send message
Joined: 1 May 20
Posts: 1
Message 98255 - Posted: 1 May 2020, 18:49:16 UTC

I have 2 AMD GPUs. If I exclude ATI/AMD with device 0 from a project, then all AMD GPU from a project goes into download and abort loop with "missing coprocessor". This will constantly download and abort tasks until I stop it. I wonder why BOINC fails to start AMD GPU on device 1, as in, the other AMD GPU device that isn't device 0?

Both device 0 and 1 will work if I do not exclude. I would like an option to easily pause one AMD GPU but keep the other AMD GPU going. The only good way to try to do so is edit config file and BOINC options, read config files.

I used use_all_gpus as well. If without use_all_gpus then BOINC client only use device 0, slow Vega 11, not the other device.

cc_config and some logs:
<cc_config>
<options>
<allow_remote_gui_rpc>1</allow_remote_gui_rpc>
<use_all_gpus>1</use_all_gpus>
<exclude_gpu>
 <url>einstein.phys.uwm.edu</url>
 <type>ATI</type>
 <device_num>0</device_num>
</exclude_gpu>
</options>
</cc_config>

27-Apr-2020 08:59:10 [---] Starting BOINC client version 7.16.5 for windows_x86_64
27-Apr-2020 08:59:10 [---] Libraries: libcurl/7.47.1 OpenSSL/1.0.2s zlib/1.2.8
27-Apr-2020 08:59:10 [---] Data directory: C:\ProgramData\BOINC
27-Apr-2020 08:59:11 [---] CUDA: NVIDIA GPU 0: GeForce GT 1030 (driver version 442.74, CUDA version 10.2, compute capability 6.1, 2048MB, 1661MB available, 1127 GFLOPS peak)
27-Apr-2020 08:59:11 [---] OpenCL: NVIDIA GPU 0: GeForce GT 1030 (driver version 442.74, device version OpenCL 1.2 CUDA, 2048MB, 1661MB available, 1127 GFLOPS peak)
27-Apr-2020 08:59:11 [---] OpenCL: AMD/ATI GPU 0: AMD Radeon(TM) RX Vega 11 Graphics (driver version 3004.8 (PAL,HSAIL), device version OpenCL 2.0 AMD-APP (3004.8), 12148MB, 12148MB available, 100 GFLOPS peak)
27-Apr-2020 08:59:11 [---] OpenCL: AMD/ATI GPU 1: Radeon RX 5500 XT (driver version 3004.8 (PAL,LC), device version OpenCL 2.0 AMD-APP (3004.8), 8176MB, 8176MB available, 2418 GFLOPS peak)

01-May-2020 12:18:45 [Einstein@Home] Config: excluded GPU.  Type: ATI.  App: all.  Device: 0
01-May-2020 12:18:45 [---] Config: use all coprocessors
01-May-2020 12:18:45 [---] log flags: file_xfer, sched_ops, task
01-May-2020 12:18:45 [Einstein@Home] Found app_config.xml
01-May-2020 12:18:45 [Milkyway@Home] Found app_config.xml
01-May-2020 12:18:45 [Rosetta@home] Found app_config.xml
01-May-2020 12:18:49 [Einstein@Home] Sending scheduler request: To fetch work.
01-May-2020 12:18:49 [Einstein@Home] Requesting new tasks for AMD/ATI GPU
01-May-2020 12:18:51 [Einstein@Home] Scheduler request completed: got 3 new tasks
01-May-2020 12:18:51 [Einstein@Home] Project requested delay of 60 seconds
01-May-2020 12:18:51 [Einstein@Home] Missing coprocessor for task LATeah1063L32_436.0_0_0.0_17934476_1; aborting
01-May-2020 12:18:51 [Einstein@Home] Missing coprocessor for task LATeah1063L32_436.0_0_0.0_16701440_1; aborting
01-May-2020 12:18:51 [Einstein@Home] Missing coprocessor for task LATeah1063L32_436.0_0_0.0_16999913_0; aborting
ID: 98255 · Report as offensive
Ian&Steve C.

Send message
Joined: 24 Dec 19
Posts: 228
United States
Message 98261 - Posted: 2 May 2020, 0:53:40 UTC - in response to Message 98255.  

are you trying to use the Vega 11 GPU for anything else? if you don't want BOINC to use it for any projects. you could try using the ignore flag instead

<ignore_ati_dev>N</ignore_ati_dev>
Ignore (don't use) a specific ATI GPU. You can ignore more than one. Requires a client restart.

ID: 98261 · Report as offensive

Message boards : GPUs : exclude_gpu device_num with 2 AMD GPUs: download and abort loop

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.