Which CPU task(s) generate the most RAC?

JHMarshall
JHMarshall
Joined: 24 Jul 12
Posts: 17
Credit: 1018018169
RAC: 0

It's normal operation for

It's normal operation for BOINC to only use the better performing GPU if a system has two different GPUs types.

Have you tried using a cc_config file with use_all_gpus  set to 1?

 

Joe

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1364
Credit: 3562358667
RAC: 1580

It's been a few years since I

It's been a few years since I did a comparison, but the GW CPU tasks offered significantly higher credit than the Fermi ones on my intel/win10 boxes.

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4704
Credit: 17546245278
RAC: 6404912

Yes, 1000 versus 680 credits.

Yes, 1000 versus 680 credits.

 

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109398736666
RAC: 35694941

As I did mention, it's credit

As I did mention, it's credit per unit of time that would be the important metric for someone trying to squeeze every last ounce of RAC from the CPU cores.

I don't run CPU tasks but the last time I noticed (admittedly some time ago), it was possible to get nearly 2 GRP tasks done in the time to complete a single GW task.  It probably differs with different architectures so Tom would need to do a comparison on his host to be sure.  In the end, it probably wont make a great deal of difference.

Cheers,
Gary.

Tom M
Tom M
Joined: 2 Feb 06
Posts: 5589
Credit: 7674622754
RAC: 1809600

JHMarshall wrote: It's

JHMarshall wrote:

It's normal operation for BOINC to only use the better performing GPU if a system has two different GPUs types.

Have you tried using a cc_config file with use_all_gpus  set to 1?

 

Joe

Joe,

If you check the first 30 odd lines of your Boinc log after you first start up Boinc it lists the available gpu even if you are not using them using the "use_all_gpus" parameter.

12/2/2020 4:33:17 AM |  | Starting BOINC client version 7.16.11 for windows_x86_64
12/2/2020 4:33:17 AM |  | log flags: file_xfer, sched_ops, task, sched_op_debug
12/2/2020 4:33:17 AM |  | Libraries: libcurl/7.47.1 OpenSSL/1.0.2s zlib/1.2.8
12/2/2020 4:33:17 AM |  | Data directory: C:\ProgramData\BOINC
12/2/2020 4:33:17 AM |  | Running under account tlgal
12/2/2020 4:33:17 AM |  | OpenCL: AMD/ATI GPU 0: AMD Radeon RX 5700 (driver version 3110.7 (PAL,LC), device version OpenCL 2.0 AMD-APP (3110.7), 8176MB, 8176MB available, 7488 GFLOPS peak)
12/2/2020 4:33:17 AM |  | OpenCL: AMD/ATI GPU 1: AMD Radeon RX 5700 (driver version 3110.7 (PAL,LC), device version OpenCL 2.0 AMD-APP (3110.7), 8176MB, 8176MB available, 7488 GFLOPS peak)
12/2/2020 4:33:18 AM |  | All projects have zero resource share; setting to 100
12/2/2020 4:33:18 AM |  | Windows processor group 0: 8 processors
12/2/2020 4:33:18 AM |  | Host name: Moonshot-W
12/2/2020 4:33:18 AM |  | Processor: 32 AuthenticAMD AMD Ryzen 3 3100 4-Core Processor [Family 23 Model 113 Stepping 0]
12/2/2020 4:33:18 AM |  | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 htt pni ssse3 fma cx16 sse4_1 sse4_2 movebe popcnt aes f16c syscall nx lm avx avx2 svm sse4a osvw ibs skinit wdt tce topx page1gb rdtscp fsgsbase bmi1 smep bmi2
12/2/2020 4:33:18 AM |  | OS: Microsoft Windows 10: Professional x64 Edition, (10.00.19041.00)
12/2/2020 4:33:18 AM |  | Memory: 15.95 GB physical, 18.33 GB virtual
12/2/2020 4:33:18 AM |  | Disk: 465.05 GB total, 421.86 GB free
12/2/2020 4:33:18 AM |  | Local time is UTC -6 hours
12/2/2020 4:33:18 AM |  | No WSL found.
12/2/2020 4:33:18 AM |  | VirtualBox version: 6.1.12
12/2/2020 4:33:18 AM | Einstein@Home | Found app_config.xml
12/2/2020 4:33:18 AM | PrimeGrid | Found app_config.xml
12/2/2020 4:33:18 AM | PrimeGrid | Max 3 concurrent jobs
12/2/2020 4:33:18 AM | Einstein@Home | einstein_O2MD1: Max 4 concurrent jobs
12/2/2020 4:33:18 AM |  | Config: simulate 32 CPUs
12/2/2020 4:33:18 AM |  | General prefs: from http://www.worldcommunitygrid.org/ (last modified 27-Nov-2020 12:20:44)
12/2/2020 4:33:18 AM |  | Host location: none
12/2/2020 4:33:18 AM |  | General prefs: using your defaults
12/2/2020 4:33:18 AM |  | Reading preferences override file
12/2/2020 4:33:18 AM |  | Preferences:
12/2/2020 4:33:18 AM |  | max memory usage when active: 12251.67 MB
12/2/2020 4:33:18 AM |  | max memory usage when idle: 14702.00 MB
12/2/2020 4:33:18 AM |  | max disk usage: 418.54 GB
12/2/2020 4:33:18 AM |  | max CPUs used: 24
12/2/2020 4:33:18 AM |  | (to change preferences, visit a project web site or select Preferences in the Manager)
12/2/2020 4:33:18 AM |  | Setting up project and slot directories
12/2/2020 4:33:18 AM |  | Checking active tasks
12/2/2020 4:33:18 AM | Einstein@Home | URL http://einstein.phys.uwm.edu/; Computer ID 12858369; resource share 150
12/2/2020 4:33:18 AM | Einstein@Home | Not using CPU: project preferences
12/2/2020 4:33:18 AM | PrimeGrid | URL http://www.primegrid.com/; Computer ID 1040050; resource share 100
12/2/2020 4:33:18 AM | PrimeGrid | Not using AMD/ATI GPU: project preferences
12/2/2020 4:33:18 AM |  | Setting up GUI RPC socket
12/2/2020 4:33:18 AM |  | Checking presence of 6204 project files
 

I am reporting that when I install an Rx 5700 GPU and an Rx 580 on the same system I get one GPU or the other GPU but I can't get both to be listed at the same time.  Don't get me wrong, the non-listed gpu is still functioning as a monitor driver.  But not as a cruncher.

I think I have replicated this across both Linux and Windows.  So something about the two versions of OpenCL don't play well together.

It would be really nice if we could run heterogeneous Rx video card setups.  It would make it cheaper to set up 5-7 GPU rigs with very high production levels (aka RAC).

It would also make the top 50 producers list even more production heavy.

Tom M

 

 

 

 

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)

Comment viewing options

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