Generic Multiple GPU discusssion

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4,982
Credit: 18,807,765,968
RAC: 7,848,393

Just in case you didn't know,

Just in case you didn't know, Bridgman is an AMD Linux developer. He should know of what he speaks.

 

petri33
petri33
Joined: 4 Mar 20
Posts: 124
Credit: 4,096,165,819
RAC: 6,162,951

Hi   TOM M,   To your

Hi   TOM M,

 

To your 2) ? "version of BOINC Manager that would allow me to change the number of application tasks depending on each GPU or class of GPUs."

 

There is a way to run different GPUs with the settings they each like.

Set up a plethora (multiple) of boinc clients on the same machine!

 

Each of them will have a ....xml file where you can tell 'use/do not use' gpu 1, 2, n, m, and the other boinc instances will obey their own set of rules set by their individual ...xml files.

I have not done that myself, but I have read it is possible.

If I'm correct it is possible to run one GPU with one boinc - should you have N boinc instances runnig (each in its own directory/folder).

 

Those who know better can correct/clarify me. (Please!!)

--

petri33

 

 

 

 

 

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4,982
Credit: 18,807,765,968
RAC: 7,848,393

Yes, that is the correct

Yes, that is the correct method to run different gpu parameters/projects with concurrent independent BOINC instances.

https://www.overclock.net/threads/guide-setting-up-multiple-boinc-instances.1628924/

 

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3,985
Credit: 47,425,992,642
RAC: 61,078,469

petri33 wrote: To your 2) ?

petri33 wrote:

To your 2) ? "version of BOINC Manager that would allow me to change the number of application tasks depending on each GPU or class of GPUs."

if you have something like a mixed nvidia/AMD system, you can setup different parameters this way in the app_config since the application and plan class will be different for nvidia vs AMD/"ati" using the <app_version> flag in app_config. see the specifics here: https://boinc.berkeley.edu/wiki/Client_configuration

 

otherwise you have to setup multiple BOINC instances as mentioned above, but this comes with the caveat that it splits up your stats into "logical" hosts.

 

it would be nice if BOINC had some tighter project level configuration on a per device level (not just by plan class or type). For example one of my systems has an RTX 3070 with a GTX 1660Super. I get a little better production with the RTX3070 running 2x, but the GTX1660S occasionally will freeze up when running 2x, and doesnt see much benefit in doing so. So I want to run the RTX3070 at 2x but the 1660S at 1x on the same GR app without splitting the host into two instances, but at the moment BOINC just can't be configured this way since they both have the same plan class and the same app. the only solution for my situation is to give up some production and run 1x to avoid the occasional freezing on my 1660S

 

_________________________________________________________________________

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6,483
Credit: 9,608,994,374
RAC: 5,396,301

Tom M wrote:Keith,I have now

Tom M wrote:
Keith,I have now skimmed the thread.  Once I have re-created the inability to run an Rx 570/580 card and an Rx 5700 card under Ubuntu 2.0 I will see if I can correctly explain what I did and how it is not working.

It appears my Boinc Manager (I run the A-I-O under Linux) may not be working.

Here is the results of: lspci | grep VGA

tom@gpu-server:~$ lspci | grep VGA
00:02.0 VGA compatible controller: Intel Corporation UHD Graphics 630 (Desktop 9 Series) (rev 02)
01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Ellesmere [Radeon RX 470/480/570/570X/580/580X/590] (rev ef)
02:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Ellesmere [Radeon RX 470/480/570/570X/580/580X/590] (rev ef)
0f:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Navi 10 [Radeon RX 5600 OEM/5600 XT / 5700/5700 XT] (rev c4)
tom@gpu-server:~$


And here is the first 30 odd lines of my Boinc Mgr log.

Sat 13 Mar 2021 04:48:54 PM CST |  | Starting BOINC client version 7.16.5 for x86_64-pc-linux-gnu
Sat 13 Mar 2021 04:48:54 PM CST |  | log flags: file_xfer, sched_ops, task, sched_op_debug
Sat 13 Mar 2021 04:48:54 PM CST |  | Libraries: libcurl/7.68.0 GnuTLS/3.6.13 zlib/1.2.11 brotli/1.0.7 libidn2/2.2.0 libpsl/0.21.0 (+libidn2/2.2.0) libssh/0.9.3/openssl/zlib nghttp2/1.40.0 librtmp/2.3
Sat 13 Mar 2021 04:48:54 PM CST |  | Data directory: /home/tom/Desktop/BOINC
Sat 13 Mar 2021 04:48:59 PM CST |  | OpenCL: AMD/ATI GPU 0: Radeon RX 570 Series (driver version 3188.4, device version OpenCL 1.2 AMD-APP (3188.4), 8183MB, 8183MB available, 4833 GFLOPS peak)
Sat 13 Mar 2021 04:48:59 PM CST |  | OpenCL: AMD/ATI GPU 1: Radeon RX 570 Series (driver version 3188.4, device version OpenCL 1.2 AMD-APP (3188.4), 8183MB, 8183MB available, 4506 GFLOPS peak)
Sat 13 Mar 2021 04:48:59 PM CST |  | All projects have zero resource share; setting to 100
Sat 13 Mar 2021 04:48:59 PM CST |  | libc: Ubuntu GLIBC 2.31-0ubuntu9 version 2.31
Sat 13 Mar 2021 04:48:59 PM CST |  | Host name: gpu-server
Sat 13 Mar 2021 04:48:59 PM CST |  | Processor: 16 GenuineIntel Intel(R) Core(TM) i9-9900 CPU @ 3.10GHz [Family 6 Model 158 Stepping 13]
Sat 13 Mar 2021 04:48:59 PM CST |  | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb invpcid_single ssbd ibrs ibpb stibp ibrs_enhanced tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 avx2 smep bmi2 erms invpcid mpx rdseed adx smap clflushopt intel_pt xsaveopt xsavec xgetbv1 xsaves dtherm ida arat pln pts hwp hwp_notify hwp_act_window hwp_epp md_clear flush_l1d arch_capabilities
Sat 13 Mar 2021 04:48:59 PM CST |  | OS: Linux Ubuntu: Ubuntu 20.04.1 LTS [5.4.0-42-generic|libc 2.31 (Ubuntu GLIBC 2.31-0ubuntu9)]
Sat 13 Mar 2021 04:48:59 PM CST |  | Memory: 15.54 GB physical, 2.00 GB virtual
Sat 13 Mar 2021 04:48:59 PM CST |  | Disk: 227.74 GB total, 205.82 GB free
Sat 13 Mar 2021 04:48:59 PM CST |  | Local time is UTC -6 hours
Sat 13 Mar 2021 04:48:59 PM CST |  | Config: use all coprocessors
Sat 13 Mar 2021 04:48:59 PM CST |  | General prefs: from http://www.worldcommunitygrid.org/ (last modified 08-Mar-2021 16:19:49)
Sat 13 Mar 2021 04:48:59 PM CST |  | Host location: none
Sat 13 Mar 2021 04:48:59 PM CST |  | General prefs: using your defaults
Sat 13 Mar 2021 04:48:59 PM CST |  | Reading preferences override file
Sat 13 Mar 2021 04:48:59 PM CST |  | Preferences:
Sat 13 Mar 2021 04:48:59 PM CST |  | max memory usage when active: 7954.07 MB
Sat 13 Mar 2021 04:48:59 PM CST |  | max memory usage when idle: 14317.33 MB
Sat 13 Mar 2021 04:48:59 PM CST |  | max disk usage: 204.96 GB
Sat 13 Mar 2021 04:48:59 PM CST |  | max CPUs used: 14
Sat 13 Mar 2021 04:48:59 PM CST |  | don't use GPU while active
Sat 13 Mar 2021 04:48:59 PM CST |  | suspend work if non-BOINC CPU load exceeds 25%
Sat 13 Mar 2021 04:48:59 PM CST |  | (to change preferences, visit a project web site or select Preferences in the Manager)
Sat 13 Mar 2021 04:48:59 PM CST |  | Setting up project and slot directories
Sat 13 Mar 2021 04:48:59 PM CST |  | Checking active tasks
Sat 13 Mar 2021 04:48:59 PM CST | Einstein@Home | URL http://einstein.phys.uwm.edu/; Computer ID 12876469; resource share 100
Sat 13 Mar 2021 04:48:59 PM CST | Einstein@Home | Your settings do not allow fetching tasks for CPU.  To fix this, you can change Project Preferences on the project's web site.
Sat 13 Mar 2021 04:48:59 PM CST |  | Setting up GUI RPC socket
Sat 13 Mar 2021 04:48:59 PM CST |  | Checking presence of 56 project files
Sat 13 Mar 2021 04:54:31 PM CST | Einstein@Home | Computation for task LATeah3003L02_948.0_0_0.0_28457226_1 finished
Sat 13 Mar 2021 04:54:31 PM CST | Einstein@Home | Starting task LATeah3003L02_948.0_0_0.0_28709631_1


So apparently it is not a driver issue.  Apparently it is a Boinc Mgr.

I could "revert" to the generic Ubuntu install for the Boinc Mgr system.

I could try out "Pandora".  I could download yet another copy of Tbar's All-In-One.

Any other ideas?  I will start with another download of the AIO archive.

====edit===

The link on the GPUUG page to the Wiki via S@H seems to have stopped working.  So I can't download another copy.

===edit====

Apparently the Rx 5700 is not being recognized at all.  Will try adding the OCL driver manually like we used to have to do with the Nvidia cards.

Nope.

===edit===

Installed the generic Ubuntu version of Boinc client/manager.

And figured out where the cc_config.xml file lives. 

Now have a mixed rx 570 (2) and rx 5700 (1) running on system.

Now to find where the app_config.xml files live.

===edit---

Tom M

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6,483
Credit: 9,608,994,374
RAC: 5,396,301

Tom M wrote: Installed the

Tom M wrote:

Installed the generic Ubuntu version of Boinc client/manager.

And figured out where the cc_config.xml file lives. 

Now have a mixed rx 570 (2) and rx 5700 (1) running on system.

Now to find where the app_config.xml files live.

The app_config.xml file lives in: tom@gpu-server:/var/lib/boinc/projects/einstein.phys.uwm.edu$

And I now have 3 gpu Gamma Ray tasks per GPU running on the gpu-server (Ubuntu).

I am much more used to the AIO from a compaq and non-security conscious form.

Tom M

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4,982
Credit: 18,807,765,968
RAC: 7,848,393

If you feel brave, maybe you

If you feel brave, maybe you can try out the new BoincTasks-Js application from eFMer.

He announced it awhile ago today on the BOINC website Questions and Answers forum.

I tried the Snap install and it didn't work because of missing dependencies.

But the normal Debian install via a regular .deb file worked fine.

But I can't get it to connect to any of my hosts.  Seems it is looking for a regular stock Linux BOINC installation in /var/lib/boinc/ which I don't use since I run the AIO version.

But since that is how you installed your latest version, maybe you can give the new Javascript version of BoincTasks a tryout and see if it works for you.

Just another datapoint for Fred to debug.

 

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6,483
Credit: 9,608,994,374
RAC: 5,396,301

Keith Myers wrote: If you

Keith Myers wrote:

If you feel brave, maybe you can try out the new BoincTasks-Js application from eFMer.

While I am not feeling brave I am feeling motivated. Will investigate.

I can now see a clear path to a 9 GPU system without buying another larger mining rack.  5 Rx 5700's, 2 Rx 580's, and 2 Rx 570's. Just have to get those pesky riser cards combo's all working (ordered 4 more this morning).

This does mean the box that is currently running the 5 Rx 5700's is going to have to start subsisting on two Rx 570's. But it also means I might make it to a much higher RAC (plus two 1600 watt PSU's?).

Tom M

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6,483
Credit: 9,608,994,374
RAC: 5,396,301

After running into assorted

After running into assorted gpu hardware glitches that required "cable" troubleshooting I ended up with 9 Amd video cards on one box.

And then the dreaded "Logon Loop" struck.  There are still to many ways Linux/Ubuntu can end up with those symptoms.  So I swapped HD's and now have the rig running under a "test" version of Windows 10.

And was able to issue an (automated) Amd driver bug report shortly thereafter.

I now have the joy, once again, of having so many gpu tasks active I can't get them on one screen (dinky/old/modest sized) of my Monitor  :)

Tom M

 

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4,982
Credit: 18,807,765,968
RAC: 7,848,393

You should never get a "logon

You should never get a "logon loop" problem if you do one simple thing to preempt it.

I have told you multiple times but you seem to forget every install you do.

 

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.