Dual GPU
log in

Advanced search

Message boards : AMD : Dual GPU

Author Message
Profile Michael H.W. Weber
Send message
Joined: 18 Nov 16
Posts: 11
Credit: 60,819,518
RAC: 0
Message 142 - Posted: 19 Jun 2017, 14:04:43 UTC

Would it be possible to make the log file for each of the tasks a bit more informative regarding what type of GPU was used? I have a dual GPU system (1x AMD Tahiti [280X] & 1x AMD Hawaii [290X]) and if one card gives errors it becomes a bit uncomfortable finding out which is the faulty one.

Michael.
____________
President of Rechenkraft.net

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 322
Credit: 103,382
RAC: 0
Message 146 - Posted: 19 Jun 2017, 18:46:09 UTC - in response to Message 142.

Would it be possible to make the log file for each of the tasks a bit more informative regarding what type of GPU was used? I have a dual GPU system (1x AMD Tahiti [280X] & 1x AMD Hawaii [290X]) and if one card gives errors it becomes a bit uncomfortable finding out which is the faulty one.

Michael.

Actually, the app writes the info about the GPU it was run on to one of the two resulting files (namely the one ending with '0'). Unfortunately, for now I don't know whether it is possible or not to include this information into the result's form.
I think, this info can be written to stderr.txt file created by client. However, it may be a bad idea to modify the file used by another process.

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 322
Credit: 103,382
RAC: 0
Message 149 - Posted: 19 Jun 2017, 19:39:18 UTC - in response to Message 146.

I think, this info can be written to stderr.txt file created by client.

I think, it is possible to include GPU info into stderr.txt with boinc_finish_message(int status, const char* msg, bool is_notice). I'll try this out.

Profile Michael H.W. Weber
Send message
Joined: 18 Nov 16
Posts: 11
Credit: 60,819,518
RAC: 0
Message 160 - Posted: 20 Jun 2017, 5:52:23 UTC

This is all the information, I as a contributer can extract from my project web account for a given task:

Stderr Ausgabe

<core_client_version>7.6.33</core_client_version>
<![CDATA[
<message>
exceeded elapsed time limit 749.98 (432000.00G/600.47G)
</message>
<stderr_txt>

</stderr_txt>
]]>

Somewhere here I need to see the hardware information for my graphics boards as it is included in e.g. Primegrid:

<core_client_version>7.6.33</core_client_version>
<![CDATA[
<stderr_txt>
geneferocl 3.3.2-6 (Windows/OpenCL/32-bit)

Copyright 2001-2017, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2017, Iain Bethune
Genefer is free source code, under the MIT license.

Running on platform 'AMD Accelerated Parallel Processing', device 'Hawaii', vendor 'Advanced Micro Devices, Inc.', version 'OpenCL 1.2 AMD-APP (2348.3)' and driver '2348.3'.
44 computeUnits @ 1150MHz, memSize=3072MB, cacheSize=16kB, cacheLineSize=64B, localMemSize=32kB, maxWorkGroupSize=256.
Supported transform implementations: ocl ocl2 ocl3 ocl4 ocl5

Command line: projects/www.primegrid.com/primegrid_genefer_3_3_2_3.17_windows_intelx86__openclatiGFN17MEGA.exe -boinc -q 46553866^131072+1

Low priority change succeeded.
Checking available transform implementations...
OCL transform is past its b limit.
OCL3 transform is past its b limit.
OCL4 transform is past its b limit.
OCL5 transform is past its b limit.
Using OCL2 transform
Starting initialization...
Initialization complete (0.645 seconds).
Testing 46553866^131072+1...
Estimated time for 46553866^131072+1 is 0:13:00
46553866^131072+1 is complete. (1005055 digits) (err = 0.0000) (time = 0:13:21) 11:52:06
11:52:06 (7380): called boinc_finish(0)

</stderr_txt>
]]>

You may just ask the guys at Primegrid how to do that.

It is important, however, to really extract the correct model. The 280X and 290X are both AMD R9 cards. The information in BAM! or BOINC therefore is not sufficient where for my system it would just indicate the presence of two AMD R9 OpenCL devicec.

Michael.
____________
President of Rechenkraft.net

Vlad
Project administrator
Project developer
Project tester
Project scientist
Help desk expert
Send message
Joined: 26 Oct 16
Posts: 322
Credit: 103,382
RAC: 0
Message 176 - Posted: 20 Jun 2017, 23:28:41 UTC - in response to Message 160.

It is important, however, to really extract the correct model. The 280X and 290X are both AMD R9 cards. The information in BAM! or BOINC therefore is not sufficient where for my system it would just indicate the presence of two AMD R9 OpenCL devicec.

Done. Here are two examples for Nvidia and AMD.

Profile Michael H.W. Weber
Send message
Joined: 18 Nov 16
Posts: 11
Credit: 60,819,518
RAC: 0
Message 185 - Posted: 21 Jun 2017, 12:03:22 UTC
Last modified: 21 Jun 2017, 12:09:35 UTC

Very good!

I checked it for my system and indeed it does distinguish properly between the two R9 cards.
Thank you very much.
I believe it might help a lot in future error diagnostics on systems with more than one GPU.

Michael.

P.S.: It seems your initial tweaks caused a couple of tasks to produce errors? I am not 100% sure but I think the app version # has increased by two from yesterday to today (errors occurred this night)? Howver, the final one now seems safe. ;-)
____________
President of Rechenkraft.net

Message boards : AMD : Dual GPU


Main page · Your account · Message boards


© 2021 Vladislav Neverov (NRC 'Kurchatov institute'), Nikolay Khrapov (Institute for Information Transmission Problems of RAS)