Advanced search

Message boards : Number crunching : Unified Memory in CUDA 6

Author Message
Profile Zarck
Send message
Joined: 16 Aug 08
Posts: 145
Credit: 328,473,995
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 35571 - Posted: 9 Mar 2014 | 12:57:56 UTC
Last modified: 9 Mar 2014 | 12:58:48 UTC

http://devblogs.nvidia.com/parallelforall/unified-memory-in-cuda-6/

Soon available in GPUGRID ?

Bientôt disponible dans GPUGRID ?

@+
*_*
____________

ExtraTerrestrial Apes
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 17 Aug 08
Posts: 2705
Credit: 1,311,122,549
RAC: 0
Level
Met
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 35578 - Posted: 9 Mar 2014 | 20:55:42 UTC - in response to Message 35571.

Probably not. The memory is "only" unified logically but not physically, i.e. accessing memory over the PCIe bus is still as costly as it was before. The feature just hides the complexity of doing so from the programmer. Which matters when writing new programs (and the automatism works well), but is of no benefit to a program already using careful manual memory management. Which I suppose GPU-grid is doing, because it's performing quite well.

MrS
____________
Scanning for our furry friends since Jan 2002

Profile MJH
Project administrator
Project developer
Project scientist
Send message
Joined: 12 Nov 07
Posts: 696
Credit: 27,266,655
RAC: 0
Level
Val
Scientific publications
watwat
Message 35580 - Posted: 10 Mar 2014 | 0:28:30 UTC - in response to Message 35571.

Nope, because we don't need it. A very handy feature for newcomers to CUDA, to be sure, but not of much relevance to anyone who has an existing CUDA app.

Matt

Profile skgiven
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 23 Apr 09
Posts: 3968
Credit: 1,995,359,260
RAC: 0
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 35602 - Posted: 10 Mar 2014 | 21:44:42 UTC - in response to Message 35580.

I think the importance is that an application can use the GDDR from several GPU's; GDDR5 being faster than system RAM :)
____________
FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Post to thread

Message boards : Number crunching : Unified Memory in CUDA 6

//