Is there a linux CUDA BRP app yet? Or is it just ignoring me because of something I broke?
Thanks!
Hi!
There is no Linux and no Mac OSX CUDA BRP app yet. The reason is that on these platforms, the BRP app runs into some error condition with the latest driver generation, NVIDIA is currently looking into this matter.
If I understand you correctly, what you are saying is, in my case, I have a 6 core processor, only run 5 on CPU tasks, leaving the 'spare' one to handle the GPU tasks and that will boost my throughput to the GPU?
Im willing to try anything here to get them to ramp up to a decent workload w/o my intervention and not sluff off every task completion.
I have setup all my machines with min 1 CPU unused for CPU crunching (use at most processors setting <100%). Means using max 7 CPU units on the I/ machines and only one on the dual core machines. Nothing changed on the priorities. The remaining CPU is used from the CUDA tasks. I have stable high workload on all GPU's
Whenever Use at most processors setting is set to 100%, regardless the setting of % processor usage, the GPU load varies from 10% to values above. Seems to be waitstates for the task switching on the needed CPU. Crunching time is ~ 40% longer
There is no Linux and no Mac OSX CUDA BRP app yet. The reason is that on these platforms, the BRP app runs into some error condition with the latest driver generation, NVIDIA is currently looking into this matter.
Ok - thanks. I will be patient then!
I would much rather find pulsars than run collatz numbers :-)
There is no Linux and no Mac OSX CUDA BRP app yet. The reason is that on these platforms, the BRP app runs into some error condition with the latest driver generation, NVIDIA is currently looking into this matter.
Ok - thanks. I will be patient then!
I would much rather find pulsars than run collatz numbers :-)
Does Milkyway support linux/mac GPUs? It's real science, not a credit factory.
It almost seems as if ATIs were made for Milkyway, whereas nVidias are quite unefficient over there (it's a hardware issue: low or castrated double precision performance). As an ATI user I'd rather see the green cards do other worthwhile science which the ATIs can not do (Einstein, GPU-Grid, SETI).
But if the alternative was Collatz, DNETC or PrimeGrid - well, maybe in that case Milkyway would be the better alternative despite the relatively low efficiency.
Does Milkyway support linux/mac GPUs? It's real science, not a credit factory.
Milkyway has a Linux CUDA app. The ATI app is in the works for Linux but not available currently that I know of.
I should have specified that I run 32bit - I searched for a CUDA MW app but didn't find any. Everything I ran across said windows or 64bit linux. I don't have any need of a 64bit operating system yet and everything else is still easier to run in 32bit (maybe for another year or two). If anyone has a link to a 32bit app, I would be grateful!
Yes, I am not that keen on Collatz, but Primegrid has driver issues with Linux CUDA for me. Einstein is my priority but my GPU is sitting idle if no-one else writes Linux apps for it. If anyone has suggestions of 'real science' for Linux 32bit CUDA I am happy to listen...
Is there a linux CUDA BRP app
)
Is there a linux CUDA BRP app yet? Or is it just ignoring me because of something I broke?
Thanks!
RE: Is there a linux CUDA
)
Hi!
There is no Linux and no Mac OSX CUDA BRP app yet. The reason is that on these platforms, the BRP app runs into some error condition with the latest driver generation, NVIDIA is currently looking into this matter.
CU
HBE
RE: If I understand you
)
I'm using http://www.bitsum.com/prolasso.php on my i7-970 @ 4 GHz, 8 tasks at "idle", plus BRP3_1.04 on my GTX 460 in "realtime".
Has reduced WU duration from 10 hours to 50 minutes.
The great thing is that it automatically runs all BRP tasks at the chosen priority, I don't have to "fed the baby" every 50 minutes...
HTH
dunx
P.S. I hope to add a 2nd GTX 460 next week, and another next year.
I have setup all my machines
)
I have setup all my machines with min 1 CPU unused for CPU crunching (use at most processors setting <100%). Means using max 7 CPU units on the I/ machines and only one on the dual core machines. Nothing changed on the priorities. The remaining CPU is used from the CUDA tasks. I have stable high workload on all GPU's
CPU GPU GPU Usage Cruncing time
I7 870 GTX 460 52% 55 minutes
I7 860 GTS240 58% 89 minutes
Dualcore GT240 77% 170 minutes
DualcoreM Quadro FX770M 74% 205 minutes
Whenever Use at most processors setting is set to 100%, regardless the setting of % processor usage, the GPU load varies from 10% to values above. Seems to be waitstates for the task switching on the needed CPU. Crunching time is ~ 40% longer
RE: There is no Linux and
)
Ok - thanks. I will be patient then!
I would much rather find pulsars than run collatz numbers :-)
My max is 45% usage (GTX460
)
My max is 45% usage (GTX460 SC and FTW)on E@H with the aid of PL. Second project is PG on GPU.
The cores (Phenom's II X4 945) are crunching LC.
RE: RE: There is no Linux
)
Does Milkyway support linux/mac GPUs? It's real science, not a credit factory.
RE: Does Milkyway support
)
Milkyway has a Linux CUDA app. The ATI app is in the works for Linux but not available currently that I know of.
It almost seems as if ATIs
)
It almost seems as if ATIs were made for Milkyway, whereas nVidias are quite unefficient over there (it's a hardware issue: low or castrated double precision performance). As an ATI user I'd rather see the green cards do other worthwhile science which the ATIs can not do (Einstein, GPU-Grid, SETI).
But if the alternative was Collatz, DNETC or PrimeGrid - well, maybe in that case Milkyway would be the better alternative despite the relatively low efficiency.
MrS
Scanning for our furry friends since Jan 2002
RE: RE: Does Milkyway
)
I should have specified that I run 32bit - I searched for a CUDA MW app but didn't find any. Everything I ran across said windows or 64bit linux. I don't have any need of a 64bit operating system yet and everything else is still easier to run in 32bit (maybe for another year or two). If anyone has a link to a 32bit app, I would be grateful!
Yes, I am not that keen on Collatz, but Primegrid has driver issues with Linux CUDA for me. Einstein is my priority but my GPU is sitting idle if no-one else writes Linux apps for it. If anyone has suggestions of 'real science' for Linux 32bit CUDA I am happy to listen...