Welcome to MilkyWay@home

GPU/CUDA Milkyway

Message boards : Application Code Discussion : GPU/CUDA Milkyway
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · Next

AuthorMessage
Brickhead
Avatar

Send message
Joined: 20 Mar 08
Posts: 108
Credit: 2,607,924,860
RAC: 0
Message 18420 - Posted: 12 Apr 2009, 22:00:23 UTC - in response to Message 18404.  

Surprise, surprise.
Cuda capabale cards are listed here.

http://www.nvidia.com/object/cuda_learn_products.html

As long as double precision capability is required, mere CUDA capability is irrelevant. And resistance is futile. (Sorry, couldn't resist:)
ID: 18420 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 18449 - Posted: 12 Apr 2009, 23:36:05 UTC - in response to Message 18420.  
Last modified: 12 Apr 2009, 23:37:10 UTC

Surprise, surprise.
Cuda capabale cards are listed here.

http://www.nvidia.com/object/cuda_learn_products.html

As long as double precision capability is required, mere CUDA capability is irrelevant. And resistance is futile. (Sorry, couldn't resist:)


Yes! As jedirock and Brickhead say, only the nVidia GeForce cards capable of double precision are relevant. That leaves the top 4 or 5 GeForce cards in the left column as relevant - GTX 265 - 295, with some doubt o the 260.

Nice list of CUDA compliant graphics cards, pity about the double precision.

I suspect it could be argued that the top end of the Tesla cards may also be OK?
Go away, I was asleep


ID: 18449 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile verstapp
Avatar

Send message
Joined: 26 Jan 09
Posts: 589
Credit: 497,834,261
RAC: 0
Message 18495 - Posted: 13 Apr 2009, 5:13:06 UTC
Last modified: 13 Apr 2009, 5:55:47 UTC

In which case, how long until the Radeon code is available? :)

Not that it really matters... it just means that I continue running gpugrid [once their %$#%$#$#$!$@ server is back up] on my 9800s for a while longer and continue running out of work for my 3800s and 4800 using 0.19e with the non-GPU client.

Edit: And now that I've looked again at the MW-GPU thread, I can see that someone has discovered what i thought was bleedingly obvious - you can do double precision math on a single precision processor [16bit on an 8bit cpu] just as long as you write the routine and can stand the performance hit. Mentioned here:
[one of these days the forum will be upgraded to recognise text that looks like a url... or I'll learn how to correctly format a url for this board]

http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=785&nowrap=true#18458

Keep up the good work guys.
Cheers,

PeterV

.
ID: 18495 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Paul D. Buck

Send message
Joined: 12 Apr 08
Posts: 621
Credit: 161,934,067
RAC: 0
Message 18499 - Posted: 13 Apr 2009, 7:30:10 UTC - in response to Message 18495.  

[one of these days the forum will be upgraded to recognise text that looks like a url... or I'll learn how to correctly format a url for this board]

http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=785&nowrap=true#18458

{url=http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=785&nowrap=true#18458}The url[{/url}

Just change the curly brackets to square and:

The url
ID: 18499 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Cluster Physik

Send message
Joined: 26 Jul 08
Posts: 627
Credit: 94,940,203
RAC: 0
Message 18505 - Posted: 13 Apr 2009, 7:51:07 UTC - in response to Message 18495.  

In which case, how long until the Radeon code is available? :)

If Travis would have posted the CPU part for the new app capable of handling the longer WUs for the GPU project, as he promised, maybe the porting would be finished already.

Edit: And now that I've looked again at the MW-GPU thread, I can see that someone has discovered what i thought was bleedingly obvious - you can do double precision math on a single precision processor [16bit on an 8bit cpu] just as long as you write the routine and can stand the performance hit.

With floating point it is a bit harder than for integers ;)
And frankly, I doubt it would sense to take the effort if an old X2 is faster than a 8800GT, isn't?

The GTX200 series is the way to go if you want double precision on nvidia cards.
ID: 18505 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile verstapp
Avatar

Send message
Joined: 26 Jan 09
Posts: 589
Credit: 497,834,261
RAC: 0
Message 18516 - Posted: 13 Apr 2009, 9:54:22 UTC

Paul, thanks for the url.
Cheers,

PeterV

.
ID: 18516 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Emanuel

Send message
Joined: 18 Nov 07
Posts: 280
Credit: 2,442,757
RAC: 0
Message 18610 - Posted: 13 Apr 2009, 23:36:42 UTC

Could single precision cards emulating double precision be used for the normal project? Since they're so much slower, they shouldn't put too much load on the server, and they might be on par with CPUs. I think there's a lot of people out there who'd like to use their single precision cards..
ID: 18610 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile verstapp
Avatar

Send message
Joined: 26 Jan 09
Posts: 589
Credit: 497,834,261
RAC: 0
Message 18632 - Posted: 14 Apr 2009, 4:04:05 UTC

Well I have 3 x 9800s that I'd much prefer to be using on MW than GPUGrid...

Cheers,

PeterV

.
ID: 18632 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
zpm

Send message
Joined: 27 Feb 09
Posts: 41
Credit: 123,828
RAC: 0
Message 18856 - Posted: 15 Apr 2009, 20:25:06 UTC - in response to Message 18632.  
Last modified: 15 Apr 2009, 20:26:52 UTC

least my 260 has work, tons of it. about 200 seti mb waiting, and 1 gpugrid going right now, it's almost done and those seti's get knocked out in 20 minutes. i guess the seti server loves my turn around time. it loaded my pc up. my bonic list is as long as my arm, a quad-core running cpdn, docking, roseeta, and others right now.
ID: 18856 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
zpm

Send message
Joined: 27 Feb 09
Posts: 41
Credit: 123,828
RAC: 0
Message 18857 - Posted: 15 Apr 2009, 20:29:13 UTC - in response to Message 18610.  

Could single precision cards emulating double precision be used for the normal project? Since they're so much slower, they shouldn't put too much load on the server, and they might be on par with CPUs. I think there's a lot of people out there who'd like to use their single precision cards..


yea, that would be cool. i have a ati card sitting (9250) that's taking space on a shelf.
ID: 18857 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Dan T. Morris
Avatar

Send message
Joined: 17 Mar 08
Posts: 165
Credit: 410,228,216
RAC: 0
Message 18915 - Posted: 16 Apr 2009, 1:36:33 UTC

Ok wheres the beef as an old saying goes. Has anyone got the gpu app yet for nvidia or ATI?

DD,

ID: 18915 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Bruce
Avatar

Send message
Joined: 28 Apr 08
Posts: 1415
Credit: 2,716,428
RAC: 0
Message 18919 - Posted: 16 Apr 2009, 1:55:05 UTC - in response to Message 18915.  
Last modified: 16 Apr 2009, 1:58:37 UTC

Waiting to see this new GPU stuff, still waiting,etc. ;-P Most likely found a bug in program or something to that effect. Hope it doesn't take too much longer. Not even sure my video card will be suitable for it, but I'm wiiling to try.
ID: 18919 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Dave Przybylo
Avatar

Send message
Joined: 5 Feb 08
Posts: 236
Credit: 49,648
RAC: 0
Message 18920 - Posted: 16 Apr 2009, 1:58:24 UTC - in response to Message 18919.  

I just talked to Travis and he said he was going to post what he had soon.
Dave Przybylo
MilkyWay@home Developer
Department of Computer Science
Rensselaer Polytechnic Institute
ID: 18920 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Bruce
Avatar

Send message
Joined: 28 Apr 08
Posts: 1415
Credit: 2,716,428
RAC: 0
Message 18921 - Posted: 16 Apr 2009, 1:59:10 UTC - in response to Message 18920.  

Thanx Dave that was QUICK!!
ID: 18921 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Dave Przybylo
Avatar

Send message
Joined: 5 Feb 08
Posts: 236
Credit: 49,648
RAC: 0
Message 18922 - Posted: 16 Apr 2009, 2:03:22 UTC - in response to Message 18921.  

:)

I'm usually always on the forums. I just post when I feel it will help.
Dave Przybylo
MilkyWay@home Developer
Department of Computer Science
Rensselaer Polytechnic Institute
ID: 18922 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Misfit
Avatar

Send message
Joined: 27 Aug 07
Posts: 915
Credit: 1,503,319
RAC: 0
Message 18939 - Posted: 16 Apr 2009, 6:09:51 UTC - in response to Message 18922.  

I'm usually always on the forums. I just post when I feel it will help.

...and then some. ;)
me@rescam.org
ID: 18939 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Bruce
Avatar

Send message
Joined: 28 Apr 08
Posts: 1415
Credit: 2,716,428
RAC: 0
Message 18971 - Posted: 16 Apr 2009, 12:09:39 UTC - in response to Message 18920.  

I just talked to Travis and he said he was going to post what he had soon.


I guess Travis found another or more "glitches", still waiting patiently, but getting restless!! ;-P
ID: 18971 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile GalaxyIce
Avatar

Send message
Joined: 6 Apr 08
Posts: 2018
Credit: 100,142,856
RAC: 0
Message 18977 - Posted: 16 Apr 2009, 13:13:08 UTC - in response to Message 18920.  

I just talked to Travis and he said he was going to post what he had soon.

I hope he's going to post a few WUs for those waiting patiently for them ;)


ID: 18977 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Simplex0
Avatar

Send message
Joined: 11 Nov 07
Posts: 232
Credit: 178,229,009
RAC: 0
Message 19059 - Posted: 16 Apr 2009, 18:21:18 UTC

Maybe they have gone to Stanford and beging for help.
It seams to be the only guys who can make programs both
for ATI and NVIDIA.;)
ID: 19059 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile GalaxyIce
Avatar

Send message
Joined: 6 Apr 08
Posts: 2018
Credit: 100,142,856
RAC: 0
Message 19064 - Posted: 16 Apr 2009, 18:34:23 UTC - in response to Message 19059.  
Last modified: 16 Apr 2009, 18:34:32 UTC

Maybe they have gone to Stanford and beging for help.
It seams to be the only guys who can make programs both
for ATI and NVIDIA.;)

That's OK by me since I have both ;)

ID: 19064 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · Next

Message boards : Application Code Discussion : GPU/CUDA Milkyway

©2024 Astroinformatics Group