ABP2 CUDA applications
7 Jan 2010 18:04:06 UTC
Topic 194710
(moderation:
Quote:
Any news on ABP2 CUDA? or a 'possible' ATI client?
The initial ABP2 CUDA App will be essentially the same as the ABP1 one, i.e. it will just run the FFT on the GPU and the other calculations on the CPU. A custom kernel to bring these to the GPU, too is still being worked on.
If there is a similar library to cuFFT for ATI one could use the same approach, but with the same drawback (lots of computation still done on the CPU).
Currently it looks like we'll finish the CUDA kernel for ABP2, the port this to an OpenCL version that would support both GPUs (and then some computing devices).
BM
ABP2 CUDA applications
)
ABP1 is now replaced by ABP2, also featuring a CUDA app for Windows and Linux.
This thread is for discussion of the ABP2 cuda apps.
Bikeman
I'm still getting 3.13, ABP1
)
I'm still getting 3.13, ABP1
Something is very strange.
)
Something is very strange. The BOINC Manager is requesting work for CPU and GPU but only gets S5R6 tasks. This repeats every minute. There will be no ABP2 work. Because of this behaviour I ran yesterday and today into the mayium daily quote. I am now full of S5R6 tasks, the manager is in high priority mode and I hope to get the work finished befor the deadline.
RE: Something is very
)
See Bernd's post.
Michael
Team Linux Users Everywhere
RE: Something is very
)
Are you saying that BOINC receives new CPU tasks when asking for GPU ones? That would be a severe BOINC problem.
You do have plenty of tasks though. Is that due to your cache settings?
Gruß,
Gundolf
[edit]BOINC shouldn't request any new tasks if in high priority mode.[/edit]
Computer sind nicht alles im Leben. (Kleiner Scherz)
RE: RE: Something is very
)
BOINC is in high priority since I ran into the dayli quota. My normal cache is set to 3 days. It requested work for both: "Requesting new tasks for CPU and GPU" every minute and got always a S5R6 task.
Arn´t the ABP2 task not in the pipe yet?
If you have a look at the tasks of this host, I think, there would be the same problem.
RE: Arn´t the ABP2 task
)
The ABP2 app is out, but tasks are only generated in very small batches at the moment. So a few hundred WUs or so are generated, the returned results are inspected manually by the scientists, and if everything is ok the pipeline is turned wide open, so to speak.
Flooded caches: I think the "duration correction factor", a metric that is used by BOINC to guesstimate runtime, is only managed by BOINC on a per project basis, and not "per application". This could result in severe over- or underestimation of runtime if the mix of WUs for the different applications changes (as now).
If you feel like cancelling some results because you think there are just too many in the queue now to have a realistic chance to finish before the deadline, just cancel them and maybe set E@H to "no more work" for one or two days to prevent excessive cache filling. It's completely legitimate.
CU
Bikeman
I got 16 ABP2 CUDA WUs and my
)
I got 16 ABP2 CUDA WUs and my 260 GTX is crunching the first one right now.
According to GPU-Z the GPU load ist @ 14%. The ABP1 CUDA Abb only used 5%. The WU will take approximately 40 min to finish.
Sascha
RE: Something is very
)
I discovered the same problem yesterday on one of my systems. I told it not to request more work (until the cache runs down a bit).
Seti Classic Final Total: 11446 WU.
My i5 750 paired with a gtx
)
My i5 750 paired with a gtx 260 is plowing through the abp2 cuda tasks in 30-35 minutes. That compares to roughly 280 minutes for abp1 tasks. 8.8x +/- speedup. Very nice. :)