| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: GPU Support Forum
Thread: Which GPU is best "bang for buck" for crunching? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 37
|
|
| Author |
|
|
Jack007
Master Cruncher CANADA Joined: Feb 25, 2005 Post Count: 1604 Status: Offline Project Badges: 
|
What do you guys think about the R290 for around 400 bucks?
---------------------------------------- |
||
|
|
Mamajuanauk
Master Cruncher United Kingdom Joined: Dec 15, 2012 Post Count: 1900 Status: Offline Project Badges: 
|
What do you guys think about the R290 for around 400 bucks? Hi Jack, don't know much, but a quick 'google' reveals some heat issues and a smear campain by NVidia!Personally I would stand back and watch what happens with the heat issues, ignoring the NVidia smear and make up my own mind from the latest info and reviews available on the various review sites & forums...
Mamajuanauk is the Name! Crunching is the Game!
  |
||
|
|
jay_Orlando
Senior Cruncher USA Joined: Jan 4, 2006 Post Count: 189 Status: Offline Project Badges:
|
Hi there,
----------------------------------------I have searched through benchmark comparison of GPU. But, these seem to be based on video (game) performance - not based on "typical" number crunching used in BOINC projects. I have a Radeon (ATI) 7750 and have been trying to compare my results with my "wing-Man" when the project prove a comparison of all the clients assigned to the same WU. Then I noticed that the ratio between mine and another (like a 7790) can vary. That hints hat the code running can make comparisons difficult. OK, the point.... It seems to be that the web administrators could do a database search on GPU tasks - where the tasks are the same and get a average processing time per group of GPU manufacturer and model. How about it? Would that cause any legal problems? Any suggestions? I realize that the WCG forum may not be best for tha request, since they don't have any GPU tasks at this time, but IMHO, their forums have better discussions. Thanks, Jay  |
||
|
|
deltavee
Ace Cruncher Texas Hill Country Joined: Nov 17, 2004 Post Count: 4894 Status: Offline Project Badges: 
|
I realize that the WCG forum may not be best for tha request, since they don't have any GPU tasks at this time, but IMHO, their forums have better discussions. Thanks, Jay The best place to ask this question would be at the forum for the particular Boinc project you are crunching. That being said, the charts in this link directcompute are comparable for OpenCL computing. [Edit 1 times, last edit by deltavee at Apr 22, 2014 7:39:06 PM] |
||
|
|
Coleslaw
Veteran Cruncher USA Joined: Mar 29, 2007 Post Count: 1343 Status: Offline Project Badges:
|
Even then, your numbers will differ from manufacturer, how many work units were ran at a time on the same card, whether a full core/thread was reserved for the GPU, heat, over clocking, etc...
----------------------------------------You can of course get a rough estimate though. And as deltavee, it is usually best to go to the forums of the project in question since WCG has no GPU sub projects right now.     [Edit 1 times, last edit by Coleslaw at Apr 22, 2014 8:55:15 PM] |
||
|
|
noderaser
Senior Cruncher United States Joined: Jun 6, 2006 Post Count: 297 Status: Offline Project Badges: 
|
WUprop has some data comparing GPUs: http://wuprop.boinc-af.org/results/compar_gpu.py
----------------------------------------The best way to determine which GPU you should buy for computing is to check at a particular project, since not all projects support all GPUs and there are variations in programming that may favor some cards over others. |
||
|
|
Coleslaw
Veteran Cruncher USA Joined: Mar 29, 2007 Post Count: 1343 Status: Offline Project Badges:
|
Even WUProp's numbers will be unreliable. It doesn't provide info on if the users were running multiple work units per card ect.. as I mentioned above either. It can however give you a very rough guestimate.
---------------------------------------- |
||
|
|
|