Advanced search

Message boards : Graphics cards (GPUs) : Quadro 600

Author Message
Saro
Send message
Joined: 9 Mar 10
Posts: 7
Credit: 883,933
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 21102 - Posted: 29 Apr 2011 | 12:00:30 UTC

Hi guys,
I've installed the PNY Quadro 600 graphic card on my desktop. Unfortunaly the CPU is still a E2160, but I'm feady for an upgrade.
Have you any information about its (Quadro 600) performance with GPUGRID.

Many Thanks,
Docsaro

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 21106 - Posted: 29 Apr 2011 | 15:16:37 UTC - in response to Message 21102.

I take it this is the new Quadro 600 (GF108GL, 40nm).
Assuming it is then its similar to a GT430, but slightly downclocked. Expect a real-life performance of about 164 GFlops peak. As you can see from this table of performance, it would take about 4 times as long to finish a task as a GTX550 (an entry level GPU for crunching at GPUGrid). You Quadro 600 might be able to crunch the normal length tasks in a useful time if the system is optimized properly and running 24/7, but you would need to test it to know for sure. Basically it's not a recommended card, but you might get some use out of it.

What does Boinc Messages say about the card (about the 13th line down)?

Saro
Send message
Joined: 9 Mar 10
Posts: 7
Credit: 883,933
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 21136 - Posted: 2 May 2011 | 15:06:01 UTC - in response to Message 21106.

Thank you for your reply and sorry for the delay in mine.
The result of the performance is as surprising as negative.
I took a Quadro 600 graphic card, although an entry-level, thinking with performing Cuda calculations and I would not imagine a so low performance value.
I would like to deal with DNA alignment of microbial genomes and wanted to equip a minimal resource to get into CUDA.
Thanks anyway for the comparison table. I guess I will buy the alternative graphic card equipped with the GTX550.
Docsaro

Saro
Send message
Joined: 9 Mar 10
Posts: 7
Credit: 883,933
RAC: 0
Level
Gly
Scientific publications
watwatwatwatwat
Message 21137 - Posted: 2 May 2011 | 15:07:22 UTC - in response to Message 21136.

The line you asked for:

02/05/2011 12:10:32 NVIDIA GPU 0: Quadro 600 (driver version 27061, CUDA version 4000, compute capability 2.1, 962MB, 164 GFLOPS peak)

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 21140 - Posted: 2 May 2011 | 20:09:36 UTC - in response to Message 21136.

The result of the performance is as surprising as negative.
I took a Quadro 600 graphic card, although an entry-level, thinking with performing Cuda calculations and I would not imagine a so low performance value.


Well, if you buy a Quadro you actually buy the software (different driver) and support. Everything else is pretty much stock, maybe except for more memory. If a Quadro is also fast you probably wouldn't want to afford one ;)

I would like to deal with DNA alignment of microbial genomes and wanted to equip a minimal resource to get into CUDA.


Ehm, is that already available software or do you want to develop it? If it's the former, then you should take a look at its requirements and then decide which card to get. If it's the latter.. well, you're going to invest a lot of time into this, even with CUDA. I don't know if you're doing this for a living or just for fun. In any way you'll want to be efficient, right?

Generally it only makes sense to go for GPU computing if your calculations take painfully long on CPUs. That means your GPU can probably never be fast enough either: even if you achieve a speedup of 10 or 20 times the CPU speed, you'll still be waiting for the results at some point.

I don't know how serious you are about this, but if it's your job you should probably be serious ;) And in that case I'd consider a higher end card. The price increase will probably pale in comparison to what someone has to pay you to make it faster. If you go for a GTX560 instead of a GTX550 you should get about twice as much performance for 70€ more.

And depending on whether the CC 2.1 shaders can be used well by your code (which you probably can't tell by now) the CC 2.0 cards (GTX470, GTX570, GTX580) may also be very attractive solutions, since here you're guaranteed to be able to use them all, whereas the worst case for CC 2.1 cards is just 2/3 of all shaders at any one time.

Oh, and if you need double precision you're really f*cked. CC 2.1 cards support it at 1/12th the single precision performance. The CC 2.0 chips could do it at 1/2, but it's throttled to (I think..) 1/8th on consumer cards. Not sure about the Quadros, but in (very expensive) Teslas it's unlocked.

Regards, MrS
____________
Scanning for our furry friends since Jan 2002

Post to thread

Message boards : Graphics cards (GPUs) : Quadro 600

//