Trouble running 2 RX 570 on Einstein@home

Message boards : Questions and problems : Trouble running 2 RX 570 on Einstein@home
Message board moderation

To post messages, you must log in.

AuthorMessage
Tyco

Send message
Joined: 7 Aug 19
Posts: 2
Germany
Message 92416 - Posted: 7 Aug 2019, 15:20:38 UTC

Dear BOINC community,


Could you please help me with the following issue? Tried to get 2 AMD RX570 GPUs simultaneously calculating for Einstein@home. After restarting the computer several times, and loading web calculation settings it actually worked pretty fine. For a while though.. Latestly the next day it is only calculating with GPU 0, ignoring GPU 1 totally. Right now also reloading the web settings, restarting the computer or resetting the project doesn't help anymore. Interestingly there is no error recorded in the message log..
Would be awesome if anyone got an idea, on what might be wrong. Any help is appreciated, thanks a lot in advance for your efforts!


My Setup is the following:

- BOINC Version 7.14.2(x64)
- Project: Einstein@home only
- CPU: Intel Core i5-4430 @ 3GHz
- GPU: 2xRadeon RX 570 series (Note: Crossfire has been turned off; Both GPUs are visible in Task Manager/Device Manager)
- GPU Drivers: Radeon 19.7.5 (installed via Adrenalin 2019), Note: Windows Driver updates have been disabled
- Windows Version: Windows 10 Education 64-bit

cc_config.xml
The cc_config.xml, which can be found at C:\ProgramData\BOINC has been updated with the line <use_all_gpus>1</use_all_gpus> in the options section and saved using Editor with ANSI encoding.

Excerpt from BOINC Event log:

06.08.2019 17:21:54 | | Starting BOINC client version 7.14.2 for windows_x86_64
06.08.2019 17:21:54 | | log flags: file_xfer, sched_ops, task
06.08.2019 17:21:54 | | Libraries: libcurl/7.47.1 OpenSSL/1.0.2g zlib/1.2.8
06.08.2019 17:21:54 | | Data directory: C:\ProgramData\BOINC
06.08.2019 17:21:54 | | Running under account Corsair
06.08.2019 17:21:56 | | OpenCL: AMD/ATI GPU 0: Radeon RX 570 Series (driver version 2906.10, device version OpenCL 2.0 AMD-APP (2906.10), 4096MB, 4096MB available, 5145 GFLOPS peak)
06.08.2019 17:21:56 | | OpenCL: AMD/ATI GPU 1: Radeon RX 570 Series (driver version 2906.10, device version OpenCL 2.0 AMD-APP (2906.10), 4096MB, 4096MB available, 5145 GFLOPS peak)
06.08.2019 17:21:56 | | Host name: e108corair
06.08.2019 17:21:56 | | Processor: 4 GenuineIntel Intel(R) Core(TM) i5-4430 CPU @ 3.00GHz [Family 6 Model 60 Stepping 3]
06.08.2019 17:21:56 | | 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 fma cx16 sse4_1 sse4_2 movebe popcnt aes f16c rdrandsyscall nx lm avx avx2 vmx tm2 pbe fsgsbase bmi1 smep bmi2
06.08.2019 17:21:56 | | OS: Microsoft Windows 10: Education x64 Edition, (10.00.17134.00)
06.08.2019 17:21:56 | | Memory: 7.94 GB physical, 10.69 GB virtual
06.08.2019 17:21:56 | | Disk: 116.46 GB total, 83.97 GB free
06.08.2019 17:21:56 | | Local time is UTC +2 hours
06.08.2019 17:21:56 | | No WSL found.
06.08.2019 17:21:56 | | Config: use all coprocessors
06.08.2019 17:21:56 | | Last benchmark was 32 days 10:27:07 ago
06.08.2019 17:21:56 | Einstein@Home | URL http://einstein.phys.uwm.edu/; Computer ID 12768267; resource share 100
06.08.2019 17:21:56 | Einstein@Home | General prefs: from Einstein@Home (last modified 05-Jun-2019 11:15:48)
06.08.2019 17:21:56 | Einstein@Home | Host location: none
06.08.2019 17:21:56 | Einstein@Home | General prefs: using your defaults
06.08.2019 17:21:56 | | Reading preferences override file
06.08.2019 17:21:56 | | Preferences:
06.08.2019 17:21:56 | | max memory usage when active: 6501.54 MB
06.08.2019 17:21:56 | | max memory usage when idle: 7314.24 MB
06.08.2019 17:21:56 | | max disk usage: 15.00 GB
06.08.2019 17:21:56 | | (to change preferences, visit a project web site or select Preferences in the Manager)
06.08.2019 17:21:56 | | Setting up project and slot directories
06.08.2019 17:21:56 | | Checking active tasks
06.08.2019 17:21:56 | | Setting up GUI RPC socket
06.08.2019 17:21:56 | | Checking presence of 773 project files
06.08.2019 17:21:56 | | Running CPU benchmarks
06.08.2019 17:21:56 | | Suspending computation - CPU benchmarks in progress
06.08.2019 17:21:56 | Einstein@Home | Sending scheduler request: To report completed tasks.
06.08.2019 17:21:56 | Einstein@Home | Reporting 3 completed tasks
06.08.2019 17:21:56 | Einstein@Home | Requesting new tasks for AMD/ATI GPU
06.08.2019 17:21:59 | Einstein@Home | Scheduler request completed: got 7 new tasks
06.08.2019 17:22:27 | | Benchmark results:
06.08.2019 17:22:27 | | Number of CPUs: 4
06.08.2019 17:22:27 | | 3718 floating point MIPS (Whetstone) per CPU
06.08.2019 17:22:27 | | 12910 integer MIPS (Dhrystone) per CPU
ID: 92416 · Report as offensive
Profile Joseph Stateson
Volunteer tester
Avatar

Send message
Joined: 27 Jun 08
Posts: 641
United States
Message 92419 - Posted: 7 Aug 2019, 16:20:23 UTC - in response to Message 92416.  

i recall that AMD can enable crossfire mode by default unlike nvidia. if enabled try disabling it
ID: 92419 · Report as offensive
Tyco

Send message
Joined: 7 Aug 19
Posts: 2
Germany
Message 92421 - Posted: 8 Aug 2019, 7:56:15 UTC

Dear JStateson

thank you for your fast reply! Just double-checked it, Crossfire is still disabled.

What's interesting though, yesterday it only calculated with one GPU, today, even i changed absolulty nothing, it is calculating with both GPUs once again.. Let's see how long it is working this time. Will keep you posted!
ID: 92421 · Report as offensive
Gary Roberts

Send message
Joined: 7 Sep 05
Posts: 130
Australia
Message 92427 - Posted: 9 Aug 2019, 0:26:28 UTC - in response to Message 92416.  

... Would be awesome if anyone got an idea, on what might be wrong.
You really should be asking this on the Einstein project forum since there are many regulars with the necessary experience there to tell you what the problem is.

Simply put, your computer is trying to do all the searches and you have lots of CPU tasks close to or exceeding the deadline. You must have too large a work cache size for either the hours your computer runs or for the project mix you are trying to support. It looks very likely that BOINC is going into high priority mode in order to get more CPU tasks completed before the deadline. The effect of that is to prevent GPU tasks from running in order to release an extra CPU core to crunch the 'at risk' CPU tasks.

As an example of why I think this, take the GW O2 All-sky CPU search. You have a total of 232 tasks of which 147 have already exceeded the deadline. Some of these are quite recent failures and BOINC would have stopped crunching on one of your GPUs to allow an extra CPU task to start. It's an impossible situation for BOINC to cope with. If you leave things as they are, very shortly it will be a further case of 'rinse and repeat'.

The very first thing you need to do is lower your work cache settings to something very low in order to stop further work requests. You may also need to assess tasks already on board to see if there are any that can't possibly be completed by the deadline and abort them. I suspect you will find some but you need to work that out for yourself. If you want both your GPUs to crunch without interruption, you need to get rid of excess CPU tasks.

If you need help with sorting this out, post details about which projects you wish to support, what resource shares you wish them to have, how many hours per day that your computer crunches, etc., on the Einstein 'Problems' forum. Also post what the setting for your work cache size has been up till now.
Cheers,
Gary.
ID: 92427 · Report as offensive
Sandman192

Send message
Joined: 28 Aug 19
Posts: 49
United States
Message 93436 - Posted: 31 Oct 2019, 3:55:14 UTC

On the Einstein home page goto Account then Preferences. Change the one in RED to .5 for 2 work units to run on 1 GPU. Since you have 2 GPUs you'll be running 4 GPU work units.

GPU utilization factor of BRP apps: .5
WARNING! Only change this preference if you are absolutely sure of what you are doing! Wrong settings may even damage your computer! Use solely at your own risk!
See our FAQ page for more information about this GPU utilization setting. Also, ask questions in our help forum to learn more.

Default value: 1.0

GPU utilization factor of FGRP apps: .5
WARNING! Only change this preference if you are absolutely sure of what you are doing! Wrong settings may even damage your computer! Use solely at your own risk!
See our FAQ page for more information about this GPU utilization setting. Also, ask questions in our help forum to learn more.

Default value: 1.0

GPU utilization factor of GW apps: .5
WARNING! Only change this preference if you are absolutely sure of what you are doing! Wrong settings may even damage your computer! Use solely at your own risk!
See our FAQ page for more information about this GPU utilization setting. Also, ask questions in our help forum to learn more.

Default value: 1.0

.5 will run two tasks on one GPU but keep an eye on the GPU TEMPS. With stock Overclocking a GTX 1080 Ti with 3 fans has not overheated for ME yet.
ID: 93436 · Report as offensive
Profile Keith Myers
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 17 Nov 16
Posts: 863
United States
Message 93438 - Posted: 31 Oct 2019, 6:10:33 UTC

You still need to reduce your cache settings. I would try 0.1 days of cache and 0.1 days of additional cache to start. As long as you have a regular internet connection, you will always get work from the project and will be able to finish all before deadline.
ID: 93438 · Report as offensive

Message boards : Questions and problems : Trouble running 2 RX 570 on Einstein@home

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.