Plz Update "Top GPU Models" page


log in

Advanced search

Message boards : Number crunching : Plz Update "Top GPU Models" page

Author Message
Profile BMaytum
Volunteer tester
Avatar
Send message
Joined: 3 Apr 99
Posts: 101
Credit: 3,974,295
RAC: 3,100
United States
Message 1348734 - Posted: 20 Mar 2013, 15:30:23 UTC

I don't know who maintains the "Top GPU Models" webpage:
http://setiathome.berkeley.edu/gpu_list.php
but it is rather out of date in that Nvidia GeForce GTX 680 and GTX 690 models are not currently included (even though the contemporary GTX670 is listed). Likewise the Quadro series GPUs are absent (and the new GeForce Titan <- I'm assuming Titan GPU crunches S@H).

I request the maintainer to update the subject page in their spare time.

Thanks!
____________
Sabertooth Z77, i7-3770K@4.2GHz, GTX680, W8.1Pro x64
P5N32-E SLI, C2D E8400@3Ghz, GTX580, Win7SP1Pro x64 & PCLinuxOS2014 x64

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8760
Credit: 52,711,005
RAC: 23,962
United Kingdom
Message 1348735 - Posted: 20 Mar 2013, 15:45:01 UTC - in response to Message 1348734.

It will be an automatic script, rather than a person. Maybe it got broken in one or more of the server re-assignments. Let them recover from this outage, then drop Matt a note.

Josef W. SegurProject donor
Volunteer developer
Volunteer tester
Send message
Joined: 30 Oct 99
Posts: 4332
Credit: 1,113,678
RAC: 906
United States
Message 1348787 - Posted: 20 Mar 2013, 18:18:46 UTC - in response to Message 1348735.

It will be an automatic script, rather than a person. Maybe it got broken in one or more of the server re-assignments. Let them recover from this outage, then drop Matt a note.

That boinc/html/user/gpu_list.php code is probably working as well as it ever has. It considers only results from hosts running stock, which gives an unrealistic picture for this project where most of the top hosts are running anonymous platform.

Another limitation is it doesn't include results with run time less than 100 seconds, which may mean the proportion of shorties will shift the rankings a lot.

Finally, since BOINC considers each GPU on dual-GPU cards separately, the x9x cards have always been lower in the list than seems right to users.

The script is part of BOINC, and Dr. Anderson is the only author.
Joe

ClaggyProject donor
Volunteer tester
Send message
Joined: 5 Jul 99
Posts: 4212
Credit: 34,470,642
RAC: 17,957
United Kingdom
Message 1348803 - Posted: 20 Mar 2013, 18:57:34 UTC - in response to Message 1348787.
Last modified: 20 Mar 2013, 19:01:46 UTC

It will be an automatic script, rather than a person. Maybe it got broken in one or more of the server re-assignments. Let them recover from this outage, then drop Matt a note.

That boinc/html/user/gpu_list.php code is probably working as well as it ever has. It considers only results from hosts running stock, which gives an unrealistic picture for this project where most of the top hosts are running anonymous platform.

Another thing is the Stock 6.10 app doean't work on Keplers unless the CUDA_GRID_SIZE_COMPAT environment variable has been applied, so out of the Box there no Keplers would have been in the list.

If BMaytum wants this fixing, he'll have to join the Beta testing of Setiathome v7, once the v7 apps and server setup are complete, and deployed here, then Keplers will be shown.

Claggy

Horacio
Send message
Joined: 14 Jan 00
Posts: 536
Credit: 75,836,754
RAC: 21,728
Argentina
Message 1348821 - Posted: 20 Mar 2013, 19:18:35 UTC

Does the "most productive GPUs" mean that the number of GPUs beeing used by the users affect the ranking (I meant, a million GT9500 will be more productive than just 1 GTX TITAN) or is just some kind of "average" performance for that kind of GPU?
____________

Profile Wiggo
Avatar
Send message
Joined: 24 Jan 00
Posts: 7854
Credit: 98,319,655
RAC: 33,136
Australia
Message 1348827 - Posted: 20 Mar 2013, 19:25:53 UTC - in response to Message 1348821.

Does the "most productive GPUs" mean that the number of GPUs beeing used by the users affect the ranking (I meant, a million GT9500 will be more productive than just 1 GTX TITAN) or is just some kind of "average" performance for that kind of GPU?

The "most productive GPUs" is based on the amount of work returned by a certain model card so its really based on numbers and popularity of that card.

It does not reflect the actual performance of a card.

Cheers.

Message boards : Number crunching : Plz Update "Top GPU Models" page

Copyright © 2014 University of California