Page 3 of 3 FirstFirst 123
Results 21 to 23 of 23

Thread: MW WUs not using all of GPU power?

  1. #21
    Gold Member
    Slicker's Avatar
    Join Date
    October 25th, 2010
    Location
    South of Cheeseland
    Posts
    1,253

    Re: MW WUs not using all of GPU power?

    Quote Originally Posted by Fire$torm View Post
    I just checked MW's app page and the ATI app is still listed as ATI14 which uses Stream code. Take a look at the MW wu's, are they tagged as ATI14?
    When I first started crunching MW 2+ years ago, one could run two wu's simultaneously but that was using CCC 10.10. When ATI/AMD started adding openCL to the driver's it fubared the ability to do this, at least for my 4850s. You may wish to try an app_info file like this one (if ATI14)
    The driver never really had anything to do with running two WUs at a time on the same GPU. That was because of the way the app was written. A GPU can only run one function from one app at a time. However, ATI allows a second app to be loading and unloading data into GPU memory at the same time the first one is actually running. While the apps Gipsel (a.k.a. Crystal Physik on MW) wrote were really good, a little more performance could be gained by running them two at a time. I think Gipsel turned over his code to MW and is no longer writing the apps. If that's the case, Travis and company probably have no clue what a mutex or semaphore is or why it is useful to have in the MW ATI app and just got rid of the code -- and the ability to run two WUs at a time.

    Also, after Gipsel would compile the GPU code, he would tweak the GPU assembly language because he could get another 15-30% more performance that what the compiler could do. So, if Travis is making changes to the ATI code and Gipsel isn't helping, I would expect both performance to drop and the abililty to crunch 2 at a time to cease.

    As far as OpenCL vs CAL, it won't be a choice for very long. ATI has announced they are dropping CAL and that OpenCL will be the ONLY option in the future. That sucks because, even if the app is written to have 99% utilization, it sill runs 30-50% slower on OpenCL -- or at least the Collatz app does. nVidia's OpenCL version isn't any better but since they have a much larger CUDA following, they haven't said anything about dropping CUDA.
    Spring 2008 Race: (1st Place)

  2. #22
    Past Administrator
    DrPop's Avatar
    Join Date
    October 13th, 2010
    Location
    SoCal, USA
    Posts
    7,635

    Re: MW WUs not using all of GPU power?

    Quote Originally Posted by Slicker View Post
    ... So, if Travis is making changes to the ATI code and Gipsel isn't helping, I would expect both performance to drop and the abililty to crunch 2 at a time to cease.

    As far as OpenCL vs CAL, it won't be a choice for very long. ATI has announced they are dropping CAL and that OpenCL will be the ONLY option in the future. That sucks because, even if the app is written to have 99% utilization, it sill runs 30-50% slower on OpenCL -- or at least the Collatz app does. nVidia's OpenCL version isn't any better but since they have a much larger CUDA following, they haven't said anything about dropping CUDA.
    Wow. So that totally explains the drop in production when running MW. I thought I was going crazy, but I used to get at least 20K+ more credits / day on the 4870 there a year and a half ago.

    Also on your last note - this is cold water for ATI/AMD GPU crunchers. Does that mean one should not invest in any more AMD GPUs for crunching? I mean, if all projects are forced to go either OpenCL or CUDA, then what is the future for awesome crunching cards like HD 5970? Do we expect a performance hit of up to 50% lower credits? Doesn't AMD realize this is the death knell for their GPUs to ever compete in the high-end computing / server arena?

  3. #23
    Silver Member

    Join Date
    October 9th, 2011
    Location
    Charlotte, NC
    Posts
    323

    Re: MW WUs not using all of GPU power?

    My guess is ati just wants to concentrate on gaming, but if they stick to pure open cl, Science, etc will flock to nv cuda

    I've never been an ati fan, but now I'd never buy one for boinc. It would be a waste IMO as you'd get better bang for ur buck for cuda

    I'm sure boinc projects will support ati, but only with crippled open cl

    Anything cross platform never does 100%

Page 3 of 3 FirstFirst 123

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •