| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Suggestions / Feedback
Thread: What a difference a GPU makes |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 16
|
|
| Author |
|
|
cw64
Advanced Cruncher Joined: Oct 6, 2007 Post Count: 120 Status: Offline Project Badges: 
|
Surely this has been suggested before. However since I have just seen first hand the potential of GPU processing I would like to see it in WCG projects ASAP.
----------------------------------------  |
||
|
|
toss
Senior Cruncher New Zealand Joined: Jan 3, 2007 Post Count: 220 Status: Offline Project Badges: 
|
hi cw64,
There are already a stack of threads covering this matter. Bottom line is that it won't happen any time soon. Cheers |
||
|
|
cw64
Advanced Cruncher Joined: Oct 6, 2007 Post Count: 120 Status: Offline Project Badges:
|
Seems pretty clear to me that it would a waste of time to not pursue GPU apps. Are no projects suited for GPU processing?
----------------------------------------[Edit 1 times, last edit by cw64 at Aug 26, 2010 2:55:45 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Yes, it has been suggested several times. But don't take those numbers too seriously. The dramatic increase in boinc credit is the result of some projects that give way too much credit, not equivalent to the increase in processing power.
I also do some Collatz from time to time, with the only GPU capable that I have, a GTX260M in a gaming laptop. Since its a laptop, I don't want to use it for 24x7 operation, just a few WU at night when it's a little bit colder (the GPU jumps to ~85°C running this project, 90°C+ when playing games). Now look at the RAC in my signature (currently 1299 for WCG, 1435 for Collatz). This is for 7-9 days per day in WCG vs 2-4 hours per day in Collatz! Of course GPUs have great potential, and it would be fantastic if WCG supported them, but don't be fooled by Collatz (and probably MilkyWay too) boinc credits. |
||
|
|
nasher
Veteran Cruncher USA Joined: Dec 2, 2005 Post Count: 1423 Status: Offline Project Badges: 
|
as far as i can tell the GPU units are all given way too much credit for hour...
----------------------------------------and yes they have been discusses here and will happen in the future sometime but not in the next few months. yes i would love it if they could use the power of the GPU's here but right now they cant so we just do what we can and if you have suitable GPU's then use them on other projects  |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
A lot of people knock Collatz so in it's defence I would like to add that I found it to be a very useful training ground for gpu crunching.
Slicker runs a very tight ship over there and the forums are very helpful for beginners  Found the tip 'turn on the gpu' under preferences' especially useful  Credit wise Did check this out using a gtx275 a WU took 26 mins and produced between650/800 credits depending on wingman. Did one of these on the CPU it took 15 hours so 48 credits a hour which is a bit on the high side compared to others thatI have crunched. Using a 2.66 duo running windows XP32 I get: Qmc@home .. 30 credits an hour WCG ... 10/20 credits an hour Malaria ..16/18 credits an hour VP the last lot took 20 mins a wu at 15 credits so 45 credits an hour Climate change.. 25 credits an hour. Going back to the gpu after using Callatz for the experience had the courage to move on to longer wus over at gpu grid.( about 1500 credits an hour on the 275). |
||
|
|
roundup
Veteran Cruncher Switzerland Joined: Jul 25, 2006 Post Count: 843 Status: Offline Project Badges: 
|
hi cw64, There are already a stack of threads covering this matter. Bottom line is that it won't happen any time soon. Cheers Seems pretty clear to me that it would a waste of time to not pursue GPU apps. Are no projects suited for GPU processing? Yes, this has been discussed several hundred times. However, maybe C4CW could be a potential entry point for GPU@WCG. The reason is, that the science engine of C4CW called LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) is also available for CUDA (NVidia). A briefing on this matter can be found here: http://lammps.sandia.gov/workshops/Feb10/Mike_Brown/gpu_tut.pdf Furthermore the size of the project (maybe more than a million CPU years as explained here) would justify the additional development efforts to utilize the members' GPU resources. |
||
|
|
cw64
Advanced Cruncher Joined: Oct 6, 2007 Post Count: 120 Status: Offline Project Badges:
|
Darn* that graph looked impressive yesterday. Today
---------------------------------------- Milkyway tasks taking ~1.5 mins on GPU or ~4.25 hours on CPU* So really, massive amounts of time and effort are not being put into GPU apps for WCG because.........? * To put that into some context the same CPU has been chewing through the last splash of DDDT C types at 1.62-1.37 hours per p WU and 0.59-0.53 per s WU *edited to appropriate forum language - ErikaT [Edit 2 times, last edit by ErikaT at Aug 27, 2010 1:47:29 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
How much is that? Some 230K boinc credits in one day??? Really impressive. So, you'll be doing in 5 days more credit than me in 5 years at WCG. I don't know which and how many graphics cards you have, but it has to be something very serious.
But, as I already said, don't use credits as a measure of actual processing power. The difference on processing time in Milkyway that you point out is more valid and very interesting, a 170x increase I think. P.S.: @Nasher: I just realized that you joined WCG one day before me  |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges:
|
So really, massive amounts of time and effort are not being put into GPU apps for WCG because.........? You'll find all the answers there (there are multiple reasons):Search for CUDA OR GPU |
||
|
|
|