980s performing poorly

Message boards : Number crunching : 980s performing poorly
Message board moderation

To post messages, you must log in.

AuthorMessage
elec999 Project Donor

Send message
Joined: 24 Nov 02
Posts: 375
Credit: 416,969,548
RAC: 141
Canada
Message 1896115 - Posted: 19 Oct 2017, 12:52:24 UTC

My 670 are beating my 980s. Not sure what to try?
ID: 1896115 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1896156 - Posted: 19 Oct 2017, 16:27:20 UTC - in response to Message 1896115.  
Last modified: 19 Oct 2017, 16:28:12 UTC

You are running a CUDA 3.2 app which is incompatible with Maxwell cards like your 980. You would need to run at least the CUDA 50 app or better still the SoG app. I would run the Lunatics installer again and this time not let it select defaults. You need to make the choice yourself for the SoG app in the menu.
From one of your GPU tasks stderr.txt outputs
setiathome_CUDA: device 1703268 is Maxwell or newer generation, 5.x compute compatibility or higher. Cuda 3.2 or lower doesn't support these.
setiathome_CUDA: device 1703268 is Maxwell or newer generation, 5.x compute compatibility or higher. Cuda 3.2 or lower doesn't support these.

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1896156 · Report as offensive
elec999 Project Donor

Send message
Joined: 24 Nov 02
Posts: 375
Credit: 416,969,548
RAC: 141
Canada
Message 1896189 - Posted: 19 Oct 2017, 19:47:02 UTC - in response to Message 1896156.  

You are running a CUDA 3.2 app which is incompatible with Maxwell cards like your 980. You would need to run at least the CUDA 50 app or better still the SoG app. I would run the Lunatics installer again and this time not let it select defaults. You need to make the choice yourself for the SoG app in the menu.
From one of your GPU tasks stderr.txt outputs
setiathome_CUDA: device 1703268 is Maxwell or newer generation, 5.x compute compatibility or higher. Cuda 3.2 or lower doesn't support these.
setiathome_CUDA: device 1703268 is Maxwell or newer generation, 5.x compute compatibility or higher. Cuda 3.2 or lower doesn't support these.

I reinstalled. How can I check again?
ID: 1896189 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1896198 - Posted: 19 Oct 2017, 20:24:30 UTC - in response to Message 1896189.  

Look at the stderr output of one of your latest reported tasks after you made the change. It doesn't matter if it is validated yet. You should see in the output what application is being used. It should look something like this:
CPU features: FPU TSC PAE CMPXCHG8B APIC SYSENTER MTRR CMOV/CCMP MMX FXSAVE/FXRSTOR SSE SSE2 HT SSE3 SSSE3 FMA3 SSE4.1 SSE4.2 AVX SSE4A XOP FMA4 
OpenCL-kernels filename : MultiBeam_Kernels_r3584.cl 
ar=0.396164  NumCfft=203835  NumGauss=1192503516  NumPulse=226461361118  NumTriplet=452888103600
Currently allocated 2149 MB for GPU buffers
In v_BaseLineSmooth: NumDataPoints=1048576, BoxCarLength=8192, NumPointsInChunk=32768

Windows optimized setiathome_v8 application
Based on Intel, Core 2-optimized v8-nographics V5.13 by Alex Kan
SSE3xj Win32 Build 3584 , Ported by : Raistmer, JDWhale

SETI8 update by Raistmer

OpenCL version by Raistmer, r3584


What you should look for is the application called out as the OpenCL version by Raistmer. Version should be r3557 as installed by the Lunatics installer when choosing the SoG app in the installer menu. I just looked at your latest tasks and I still see the CUDA32 app, so you didn't make a choice again and let the installer install the defaults. You have to select the CUDA50 or the SoG button on the Nvidia MB app select page. The installer defaults to CUDA32 if you make no other choice.

I assume you are referring to this machine hostid=7368589 as it is the only one I see with a GTX 980.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1896198 · Report as offensive
mmonnin
Volunteer tester

Send message
Joined: 8 Jun 17
Posts: 58
Credit: 10,176,849
RAC: 0
United States
Message 1896405 - Posted: 20 Oct 2017, 19:36:08 UTC - in response to Message 1896115.  
Last modified: 20 Oct 2017, 19:36:24 UTC

My 670 are beating my 980s. Not sure what to try?


Lunatics installer:
http://mikesworld.eu/download.html
ID: 1896405 · Report as offensive
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 34253
Credit: 79,922,639
RAC: 80
Germany
Message 1896418 - Posted: 20 Oct 2017, 20:27:16 UTC
Last modified: 20 Oct 2017, 20:28:12 UTC

You should add -sbs 1024 -period_iterations_num 10 to your command line text file to speed up a little more.
Make sure its in the same line.
Like this.

-sbs 1024 -period_iterations_num 10 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64 (*requires testing)


With each crime and every kindness we birth our future.
ID: 1896418 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1897573 - Posted: 26 Oct 2017, 16:04:51 UTC - in response to Message 1897570.  

As I understand it, no. The r3584 was introduced for the lesser cards improvement. On the faster cards r3557 is supposed to be faster than r3584. I wouldn't expect the r3584 app ever to show up in the Lunatics installer if and when Richard ever decides to write a new installer (slim chance) or some other developer takes on the Lunatics installer project.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1897573 · Report as offensive

Message boards : Number crunching : 980s performing poorly


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.