| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: GPU Support Forum
Thread: Graphics Card Performance |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 254
|
|
| Author |
|
|
[VENETO] boboviz
Senior Cruncher Joined: Aug 17, 2008 Post Count: 184 Status: Offline Project Badges: 
|
It seems that most distributing projects use Nvidia cards, probably for performance reasons related to CUDA. Not really. Cuda first version was released in 2007, OpenCL in 2009..... 2 years is a LOT of time, in gpgpu field |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
CUDA is a lock-in, Stream is a lock in, OpenCL is not... it widens choice and allows to look at performance and price as what the owner specifies for her/himself. Anyone who ran out to shop to grab a card and had not been listening to "It will be on OpenCL" messaged for over a year made a choice for which s/he is responsible. No one else.
|
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: 
|
Both Nvidia and AMD are capable of OpenCL. The only project that I know of that has both a CUDA and OpenCL version is Folding@home, and it is at least twice as fast on Nvidia cards as on the comparable AMD cards. If you start development on OpenCL, then AMD cards are usually more efficient. That is probably due to fact that the AMD cards run OpenCL more efficiently, and more importantly don't have a CUDA version to compare with. (There is some debate on the POEM forum about which card does better. POEM runs on OpenCL, but it depends on how much CPU power you are willing to devote to it, and some say that Nvidia does better.)
|
||
|
|
Promilus
Cruncher Joined: Apr 11, 2008 Post Count: 9 Status: Offline Project Badges: 
|
The only project that I know of that has both a CUDA and OpenCL version is Folding@home, and it is at least twice as fast on Nvidia cards as on the comparable AMD cards Because AMD cards run OpenCL and NV cards run CUDA - there's no comparison of OpenCL@NV vs CUDA@NV and OpenCL@NV vs OpenCL@AMD. At least it wasn't some year ago... |
||
|
|
deltavee
Ace Cruncher Texas Hill Country Joined: Nov 17, 2004 Post Count: 4894 Status: Offline Project Badges: 
|
"It will be on OpenCL" messaged for over a year made a choice for which s/he is responsible. No one else. I bought my 6870 when knreed said back in March that they had been using that for testing as he posted here. It seemed that AMD would be the safe way to go. |
||
|
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges:
|
The only project that I know of that has both a CUDA and OpenCL version is Folding@home, and it is at least twice as fast on Nvidia cards as on the comparable AMD cards Because AMD cards run OpenCL and NV cards run CUDA - there's no comparison of OpenCL@NV vs CUDA@NV and OpenCL@NV vs OpenCL@AMD. At least it wasn't some year ago... Nvidia runs both CUDA and OpenCL. AMD runs only OpenCL. The comparison on Folding@home shows that Nvidia cards running CUDA are faster than comparable (in power, price and games ability) AMD cards running OpenCL. On the other hand, POEM runs on OpenCL, and it is a tossup as to whether the AMD or Nvidia cards do best. But I see that SETI, which started on CUDA, now has an OpenCL version, though I don't see any comparisons yet. As for GPUGrid, they run only CUDA. They have tried to develop an OpenCL version, but have not gotten it to work. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
just installed my new xfx amd 7880. Works great. I will continue to use my gtx560M on my laptop also; but only when away and at night..
|
||
|
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges: 
|
just installed my new xfx amd 7880. Works great. I will continue to use my gtx560M on my laptop also; but only when away and at night.. Didn't you mean 7870? I've updated the list at the top of this thread. Glad to see my card added to the list!  10/16/2012 3:29:40 PM | | ATI GPU 0: Pitcairn (CAL version 1.4.1703, 2048MB, 2009MB available, 6464 GFLOPS peak) 10/16/2012 3:29:40 PM | | OpenCL: ATI GPU 0: Pitcairn (driver version CAL 1.4.1703 (VM), device version OpenCL 1.1 AMD-APP (898.1), 2048MB, 2009MB available) ---------------------------------------- [Edit 2 times, last edit by BladeD at Oct 18, 2012 3:12:45 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
just installed my new xfx amd 7880. Works great. I will continue to use my gtx560M on my laptop also; but only when away and at night.. Didn't you mean 7870? Here's a spanish post suggesting 7880 exist yo tambien! y ahi va a salir la Ati HD 7880 XT GT3 V8 con carburador holley 40/40 y esta va a ser malisima... esta todo arreglado para cagarnos!! constantemente con el dedito..pero bue..399 en usa? bue aca va a estar como 1000 dolares... odio a todos Mucho deniro and "odio" would be the Italian word for hate... them all. |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
Both Nvidia and AMD are capable of OpenCL. The only project that I know of that has both a CUDA and OpenCL version is Folding@home, and it is at least twice as fast on Nvidia cards as on the comparable AMD cards. Folding@home is a bad example, since Nvidia & Amd-cards has never ran the same FAH-wu's. Also, FAH has a severely screwed-up points-system, so any measures like "Nvidia gives 2x more points/day than Amd" is also not really a good indication of the actual performance. Even a GPU can easily do 10x more FLOPS/day than a 16-way CPU-system, at Folding@home the 16-way CPU-system will easily get more points/day than the GPU... Still, one thing is interesting with FAH, going by the status-page the GROGPU2 max handles 1254 atoms while OPENMMGPU max handles 2262 atoms. If not mistaken the GPUGPU2 is CUDA while OPENMMGPU is Amd... Oh, as for projects with both CUDA & OpenCL, SETI@home have for years had this, but granted the Amd-versions have until resently only been available as "anonymous platform". Interestingly, while NVIDIA craps-out on some of the SETI-wu's, Amd-cards has no problems crunching these wu's. Other projects with CUDA & OpenCL includes Einstein@home, Collatz conjecture and DistrRTgen. Was sure Milkyway also had CUDA-version, but strangely enough they currently only lists OpenCL-versions for both Nvidia & Amd. Poem@home also have only OpenCL-versions.  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
|