this job will be too easy with GPU please make it cuda enabled.
As I am sure you are aware all projects are considering this as a viable option. The problem right now is that CUDA is only an Nvidia standard, ATI uses something totally different. Meaning that if any project makes their project work for CUDA cards it won't then work for other brands of video cards, unless they go thru the same process for each. Word/hope is standards may be coming that involve several manufacturers, how close is unknown though. There are really big wigs pushing for it though, and not just for us crunchers either.
SETI has WU's for Cuda-enabled systems and also WU's for non Cuda-Systems.
MilkyWay does the same.
Believe it or not, it is more than 10 times faster. On my quad-core machine i have sometimes 11 WU's active, a WU from Milkyway takes 0.1 processors.
Make both types o WU's and everybody will be happy!
Make both types o WU's and everybody will be happy!
It isn't a problem about making type of tasks. All tasks run on either CPU or GPU are the same. It's a problem to make the application. You have to port it over into a format that the GPU understands, which isn't that easy.
One correction on your post, though. Milkyway doesn't run on CUDA. It runs on ATI 38x0 and 48x0 cards only.
Now, for more information as to how far Einstein is with their GPU application, read this thread.
Make both types o WU's and everybody will be happy!
It isn't a problem about making type of tasks. All tasks run on either CPU or GPU are the same. It's a problem to make the application. You have to port it over into a format that the GPU understands, which isn't that easy.
One correction on your post, though. Milkyway doesn't run on CUDA. It runs on ATI 38x0 and 48x0 cards only.
Now, for more information as to how far Einstein is with their GPU application, read this thread.
One correction on your post, though. Milkyway doesn't run on CUDA. It runs on ATI 38x0 and 48x0 cards only.
MilkyWay needs a double precision GPU, meaning an ATI HD 38xx, 47xx, 48xx, 58xx or 5970,
OR
a Nvidia card with compute capability 1.3 or higher, so a GTX260 or better and all GTXs after it, and some of the more recent GTS and even GT models. (GT 420/30/40, GTS 450)
GPU BASED CLIENT
)
As I am sure you are aware all projects are considering this as a viable option. The problem right now is that CUDA is only an Nvidia standard, ATI uses something totally different. Meaning that if any project makes their project work for CUDA cards it won't then work for other brands of video cards, unless they go thru the same process for each. Word/hope is standards may be coming that involve several manufacturers, how close is unknown though. There are really big wigs pushing for it though, and not just for us crunchers either.
Hi, SETI has WU's for
)
Hi,
SETI has WU's for Cuda-enabled systems and also WU's for non Cuda-Systems.
MilkyWay does the same.
Believe it or not, it is more than 10 times faster. On my quad-core machine i have sometimes 11 WU's active, a WU from Milkyway takes 0.1 processors.
Make both types o WU's and everybody will be happy!
RE: Make both types o WU's
)
It isn't a problem about making type of tasks. All tasks run on either CPU or GPU are the same. It's a problem to make the application. You have to port it over into a format that the GPU understands, which isn't that easy.
One correction on your post, though. Milkyway doesn't run on CUDA. It runs on ATI 38x0 and 48x0 cards only.
Now, for more information as to how far Einstein is with their GPU application, read this thread.
RE: RE: Make both types o
)
not only HD48xx and 38xx, with HD4770 so.
Hi, is any progress with GPU
)
Hi, is any progress with GPU computing?
RE: Hi, is any progress
)
ABP2 has been available on the GPU for a while now. On Nvidia GPUs, that is. Not on ATI GPUs, such as you have in your machine.
Is there any progress on GPU
)
Is there any progress on GPU clients ?
My 1 TFLOPS (dual precision!) Radeon 5970 GPU is currently wasting processing time on Collatz Conjecture.
My Cambridge friends at QMC
)
My Cambridge friends at QMC and QuantumFIRE say they are working on GPU versions. Stay tuned.
Tullio
RE: My Cambridge friends at
)
That's really great! What's the progress?
RE: One correction on your
)
MilkyWay needs a double precision GPU, meaning an ATI HD 38xx, 47xx, 48xx, 58xx or 5970,
OR
a Nvidia card with compute capability 1.3 or higher, so a GTX260 or better and all GTXs after it, and some of the more recent GTS and even GT models. (GT 420/30/40, GTS 450)
https://boincstats.com/signature/5/user/2177/project/sig.png