AMD Athlon 64 3500+ system with PCI-e

Message boards : Number crunching : AMD Athlon 64 3500+ system with PCI-e
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3

AuthorMessage
Profile jason_gee
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 24 Nov 06
Posts: 7489
Credit: 91,093,184
RAC: 0
Australia
Message 1523436 - Posted: 2 Jun 2014, 4:26:18 UTC - in response to Message 1523431.  
Last modified: 2 Jun 2014, 4:28:53 UTC

Why aren't AP Units run on CUDA?


Simply because no-one wrote a Cuda application for AP yet :) (well that I know of, perhaps they exist).

I've chosen to focus on Multibeam with Cuda for the time being for several reasons:
- relatively limited or sporadic AP task availability
- a lot of questions about AP validation and blanking to be resolved.
- focus on refining tools and techniques rather than immediate credit, for later use by any application (AP, GBT, other projects etc)
- The OpenCL variants developed by others appear to be doing fine (while when I started on Cuda multibeam it was not doing 'fine')
- Gives me as many field proven tools as possible when the time is right ( If done now, Cuda hosts would basically chew through all the issued AP in hours...), leading to,
- ... Significant credit issues would need to be resolved first.

That doesn't stop anyone making a Cuda enabled AP app, and AP integration is planned in x42 phase 2 (of 3), but for me that's lower priority than pulling ahead of the Cuda architecture release cycles for MB, and perfecting higher level tools/techniques.
"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.
ID: 1523436 · Report as offensive
Profile TimeLord04
Volunteer tester
Avatar

Send message
Joined: 9 Mar 06
Posts: 21140
Credit: 33,933,039
RAC: 23
United States
Message 1523439 - Posted: 2 Jun 2014, 4:56:47 UTC - in response to Message 1523436.  

Why aren't AP Units run on CUDA?


Simply because no-one wrote a Cuda application for AP yet :) (well that I know of, perhaps they exist).

I've chosen to focus on Multibeam with Cuda for the time being for several reasons:
- relatively limited or sporadic AP task availability
- a lot of questions about AP validation and blanking to be resolved.
- focus on refining tools and techniques rather than immediate credit, for later use by any application (AP, GBT, other projects etc)
- The OpenCL variants developed by others appear to be doing fine (while when I started on Cuda multibeam it was not doing 'fine')
- Gives me as many field proven tools as possible when the time is right ( If done now, Cuda hosts would basically chew through all the issued AP in hours...), leading to,
- ... Significant credit issues would need to be resolved first.

That doesn't stop anyone making a Cuda enabled AP app, and AP integration is planned in x42 phase 2 (of 3), but for me that's lower priority than pulling ahead of the Cuda architecture release cycles for MB, and perfecting higher level tools/techniques.


Thanks Jason. I was just curious; being brand new to GPU crunching...

I've also just re-attached to Einstein@Home. WOW over 4 Hrs to complete a WU there, (estimated), on my new card. I've given Einstein a low priority; so, only one WU for now. We'll see what happens with SETI Main and Beta first.
TimeLord04
Have TARDIS, will travel...
Come along K-9!
Join Calm Chaos
ID: 1523439 · Report as offensive
Previous · 1 · 2 · 3

Message boards : Number crunching : AMD Athlon 64 3500+ system with PCI-e


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.