| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: GPU Support Forum
Thread: GPU Work Units - Post Your Tech Support Questions Here |
| No member browsing this thread |
|
Thread Status: Active Thread Type: Sticky Thread Total posts in this thread: 290
|
|
| Author |
|
|
mbotmiller
Cruncher Joined: Dec 1, 2005 Post Count: 47 Status: Offline Project Badges: 
|
And just to add to my previous post, I'm now also seeing similar behavior on a system running Windows as well of an OPNG WU failing almost immediately:
Host information: 5/12/2021 4:28:47 PM | | Starting BOINC client version 7.16.11 for windows_x86_64 5/12/2021 4:28:47 PM | | log flags: file_xfer, sched_ops, task 5/12/2021 4:28:47 PM | | Libraries: libcurl/7.47.1 OpenSSL/1.0.2s zlib/1.2.8 5/12/2021 4:28:47 PM | | Data directory: C:\ProgramData\BOINC 5/12/2021 4:28:49 PM | | CUDA: NVIDIA GPU 0: GeForce GT 750M (driver version 376.54, CUDA version 8.0, compute capability 3.0, 2048MB, 1703MB available, 711 GFLOPS peak) 5/12/2021 4:28:49 PM | | OpenCL: NVIDIA GPU 0: GeForce GT 750M (driver version 376.54, device version OpenCL 1.2 CUDA, 2048MB, 1703MB available, 711 GFLOPS peak) 5/12/2021 4:28:49 PM | | Windows processor group 0: 8 processors 5/12/2021 4:28:49 PM | | Processor: 8 GenuineIntel Intel(R) Core(TM) i7-4850HQ CPU @ 2.30GHz [Family 6 Model 70 Stepping 1] BOINC log: 5/12/2021 5:29:07 PM | World Community Grid | Starting task OPNG_0008773_00088_0 5/12/2021 5:30:21 PM | World Community Grid | Computation for task OPNG_0008773_00088_0 finished 5/12/2021 5:30:21 PM | World Community Grid | Output file OPNG_0008773_00088_0_r366864428_0 for task OPNG_0008773_00088_0 absent 5/12/2021 5:30:21 PM | World Community Grid | Output file OPNG_0008773_00088_0_r366864428_1 for task OPNG_0008773_00088_0 absent Result log from failed WU: Result Name: OPNG_ 0008773_ 00088_ 0-- <core_client_version>7.16.11</core_client_version> <![CDATA[ <message> (unknown error) - exit code 3221225477 (0xc0000005)</message> <stderr_txt> projects/www.worldcommunitygrid.org/wcgrid_opng_autodockgpu_7.28_windows_x86_64__opencl_nvidia_102 -jobs OPNG_0008773_00088.job -input OPNG_0008773_00088.zip -seed 1472766101 -wcgruns 900 -wcgdpf 18 INFO: Using gpu device from app init data 0 INFO:[17:29:07] Start AutoGrid... autogrid4: Successful Completion. INFO:[17:29:24] End AutoGrid... INFO:[17:29:25] Start AutoDock for ZINC000164511848-ACR1.15_RX1--fr2325benz_006--CYS142.dpf(Job #0)... OpenCL device: GeForce GT 750M Unhandled Exception Detected... - Unhandled Exception Record - Reason: Access Violation (0xc0000005) at address 0x00007FF78F95DEE0 read attempt to address 0x00000000 |
||
|
|
nyanthiss
Cruncher Joined: Nov 23, 2012 Post Count: 15 Status: Offline Project Badges: 
|
Hi Jay,
----------------------------------------There are three possible driver options for AMD cards in Linux. Option one, Mesa - the default that comes with every Linux. This will give you OpenCL 1.1 and that hasn't changed for years, so it's the same in all Ubuntu versions. Option two, ROCm. This is the newest driver, it has some extra perks, but it only supports a very limited set of cards. Your card - AMD 7750 - is definitely not supported, so no point in trying. Option three, AMDGPU-PRO. This is a proprietary driver that gets you an OpenCL 2.0 or 2.1 on your system. It is possible to install this driver in a way that installs minimal amount of packages (so the possibility of it breaking stuff is limited). Unfortunately they have been phasing out support for older GPUs from the PRO driver, so you need to find a version that supports your GPU and the OS. Quick googling says AMDGPU-PRO version 18.40 should work for you. But you'll have to install it on a OS supported by the driver. So the steps would be:
... this should install the minimum amount of packages to get you a working OpenCL 2.something.
Intel Xeon E3-1231 v3
AMD A10 7800 AMD Ryzen 5 3500U AMD Ryzen 1700X AMD Ryzen 5900X 2x RaspberryPi, 1x Odroid |
||
|
|
Ulrich Metzner
Cruncher Joined: Apr 8, 2020 Post Count: 7 Status: Offline Project Badges: 
|
Hi there,
----------------------------------------will the issue with the GT 640 ever be cared of? At the moment i'm running Einstein-GPU-tasks just fine on this card, because milkyway is out of business recently. Since the Einstein GPU-tasks require OpenCL 1.2 i'm pretty sure, my card meets the requirements just fine. Only WCG-GPU-tasks refuse to run on this card. I think, it's the same problem for the GT 730, which is mostly a rebranded GT 640.
Aloha, Uli
|
||
|
|
jay_Orlando
Senior Cruncher USA Joined: Jan 4, 2006 Post Count: 181 Status: Offline Project Badges: 
|
@Nyanthiss,
----------------------------------------Thank you for your response!! I am running Ubuntu 21.04 on my machine with the GPU and would rather not downgrade to Ubuntu 18. any idea when the OpenCL Mesa drivers will upgrade to 2.1.? I have been poking around in 1) https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/docs/features.txt where it looks like one piec is complete and others, not started. 2) https://wiki.archlinux.org/title/GPGPU Trying to learn more... Any suggestion!! T H A N K Y O U !! Jay  |
||
|
|
HatmanStack
Cruncher Joined: Aug 26, 2017 Post Count: 1 Status: Offline Project Badges: 
|
Hey Guys,
I'm having a problem getting any GPU WU and updating my boinc desktop manager to use the web settings. Here's the event log: Host Info: 5/12/2021 9:37:35 AM | | Starting BOINC client version 7.14.3 for windows_intelx86 5/12/2021 9:37:35 AM | | log flags: file_xfer, sched_ops, task 5/12/2021 9:37:35 AM | | Libraries: libcurl/7.47.1 OpenSSL/1.0.2o zlib/1.2.8 5/12/2021 9:37:35 AM | | Data directory: C:\ProgramData\BOINC 5/12/2021 9:37:35 AM | | Running under account COUNT 5/12/2021 9:37:37 AM | | CUDA: NVIDIA GPU 0: GeForce GTX 1650 (driver version 457.51, CUDA version 11.1, compute capability 7.5, 4096MB, 3325MB available, 3037 GFLOPS peak) 5/12/2021 9:37:37 AM | | OpenCL: NVIDIA GPU 0: GeForce GTX 1650 (driver version 457.51, device version OpenCL 1.2 CUDA, 4096MB, 3325MB available, 3037 GFLOPS peak) 5/12/2021 9:37:38 AM | | Host name: DESKTOP-U2VQC4P 5/12/2021 9:37:38 AM | | Processor: 4 GenuineIntel Intel(R) Core(TM) i5-3470 CPU @ 3.20GHz [Family 6 Model 58 Stepping 9] 5/12/2021 9:37:38 AM | | 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 htt tm pni ssse3 cx16 sse4_1 sse4_2 popcnt aes f16c rdrandnx lm avx vmx smx tm2 pbe fsgsbase smep 5/12/2021 9:37:38 AM | | OS: Microsoft Windows 10: Professional x64 Edition, (10.00.19041.00) 5/12/2021 9:37:38 AM | | Memory: 11.95 GB physical, 21.45 GB virtual 5/12/2021 9:37:38 AM | | Disk: 465.15 GB total, 347.60 GB free 5/12/2021 9:37:38 AM | | Local time is UTC -5 hours When trying to switch to web preferences: 5/14/2021 10:24:11 PM | World Community Grid | Host location: none 5/14/2021 10:24:11 PM | World Community Grid | General prefs: using your defaults 5/14/2021 10:24:11 PM | | Reading preferences override file 5/14/2021 10:24:11 PM | | Preferences: 5/14/2021 10:24:11 PM | | max memory usage when active: 9174.47 MB 5/14/2021 10:24:11 PM | | max memory usage when idle: 11009.36 MB 5/14/2021 10:24:11 PM | | max disk usage: 100.00 GB 5/14/2021 10:24:11 PM | | (to change preferences, visit a project web site or select Preferences in the Manager) 5/14/2021 10:24:30 PM | World Community Grid | General prefs: from World Community Grid (last modified 04-May-2021 12:32:09) 5/14/2021 10:24:30 PM | World Community Grid | Host location: none 5/14/2021 10:24:30 PM | World Community Grid | General prefs: using your defaults 5/14/2021 10:24:30 PM | | Preferences: 5/14/2021 10:24:30 PM | | max memory usage when active: 9174.47 MB 5/14/2021 10:24:30 PM | | max memory usage when idle: 11009.36 MB 5/14/2021 10:24:30 PM | | max disk usage: 10.00 GB 5/14/2021 10:24:30 PM | | (to change preferences, visit a project web site or select Preferences in the Manager) My manager keeps grabbing my default online preferences which I've updated and also created a new profile and tried to switch to it. Any ideas? |
||
|
|
nyanthiss
Cruncher Joined: Nov 23, 2012 Post Count: 15 Status: Offline Project Badges:
|
@Jay
----------------------------------------any idea when the OpenCL Mesa drivers will upgrade to 2.1.? Considering they have been sitting at 1.1 for about 8 years, i suspect the answer is "never". If you don't want to downgrade your Ubuntu, you could (in theory) find some kernel version that is supported by AMDGPU-PRO and also works with the version of your Ubuntu; but very likely the whole setup would be fragile.
Intel Xeon E3-1231 v3
AMD A10 7800 AMD Ryzen 5 3500U AMD Ryzen 1700X AMD Ryzen 5900X 2x RaspberryPi, 1x Odroid |
||
|
|
merboy
Cruncher CA Joined: Nov 17, 2004 Post Count: 8 Status: Offline Project Badges: 
|
Maybe someone can shed some light on the behavior of the OpenPandemics Covid19 gpu workunits. I'm running BOINC on Win7, nvidia 1070 and an i5 3570K. I'm monitoring hardware activity and status with hwinfo64 and the msi afterburner. The gpu is mostly just idle on those WUs and it looks like there is nearly no possible benefit of running the task on the gpu if only a few seconds gpu time come on top of about 10 minutes cpu time. While one could argue that those typical spikes in gpu load is easily misleading in pure visual representation the temperature is a pretty good indirect indicator for power usage which is guaranteed to come with computation. No power usage, no computation. While the typical GPU temperature under load is at about 67-70° C (with games or primegrid WUs) with OpenPandemics Covid19 it was at 53 and since there was quite low activity i would suspect that most of the power consumption came from the gpu just being in high performance state. So is this an error? Usual behavior of those WUs? Maybe my setup? Or is the GPU part in those just some spice on top of a cpu workload? It certainly gives me the feeling that this gpu time is wasted. In addition to that those WUs produce distinctly lag on the windows desktop. Which makes the observed idling gpu even more vexing. I am finding the exact same situation on my system. Dual video cards, very little activity, same temp monitoring as you also and same with the very laggy desktop when trying to monitor the situation. |
|||
|
||||
|
amygdala
Cruncher Joined: Jan 18, 2017 Post Count: 2 Status: Offline Project Badges:
|
Hi All!
I decided to throw some OPN tasks at my RX 560 gpu. Not the best around the block, but should still give decent computing power. Or so I thought. I am running one task at a time on the gpu, yet it still takes 30 minutes on average for a single work unit. Is that normal? (Hint: that card is spec-wise similar to Nvidia's 1050 ti) |
||
|
|
Falconet
Master Cruncher Portugal Joined: Mar 9, 2009 Post Count: 3295 Status: Offline Project Badges:
|
Hi All! I decided to throw some OPN tasks at my RX 560 gpu. Not the best around the block, but should still give decent computing power. Or so I thought. I am running one task at a time on the gpu, yet it still takes 30 minutes on average for a single work unit. Is that normal? (Hint: that card is spec-wise similar to Nvidia's 1050 ti) Looks normal to me. My RX 550 took around 45 minutes to complete 1 task in the BETA and early days of the GPU release. Remember that each GPU task has dozens or even more jobs inside it while a CPU task only has 1 or 2 jobs in it. AMD Ryzen 5 1600AF 6C/12T 3.2 GHz - 85W AMD Ryzen 5 2500U 4C/8T 2.0 GHz - 28W AMD Ryzen 7 7730U 8C/16T 3.0 GHz |
||
|
|
amygdala
Cruncher Joined: Jan 18, 2017 Post Count: 2 Status: Offline Project Badges:
|
Hi All! I decided to throw some OPN tasks at my RX 560 gpu. Not the best around the block, but should still give decent computing power. Or so I thought. I am running one task at a time on the gpu, yet it still takes 30 minutes on average for a single work unit. Is that normal? (Hint: that card is spec-wise similar to Nvidia's 1050 ti) Looks normal to me. My RX 550 took around 45 minutes to complete 1 task in the BETA and early days of the GPU release. Remember that each GPU task has dozens or even more jobs inside it while a CPU task only has 1 or 2 jobs in it. Thanks man! I appreciate your quick response, and your focus on power consumption (as in your signature). |
||
|
|
|