New PCI/PCIe based GT520 cards

mickydl*
mickydl*
Joined: 7 Oct 08
Posts: 39
Credit: 197,778,345
RAC: 70,897

OK, for those who are

OK, for those who are interested, I am currently testing the GT430 PCI. Here are a few numbers:

Project............Runtime
Einstein...........14h 14m (51280s)
DNETC............1h 26m (5171s)
PPS sieve........1h 25m (5117s)
Cullan/Woodall..0h 14m (856s)

The test machine is a Phenom IIx4 965 with 64 Bit Linux.
As you can see, for memory intense applications like Einstein the card is practically useless unless you have a very old PC. For projects that have less memory access it might be worth trying.

Michael

dskagcommunity
dskagcommunity
Joined: 16 Mar 11
Posts: 87
Credit: 949,978,246
RAC: 392,865

Wow thx for your test! I had

Wow thx for your test! I had planned such a card or more cards for Einstein or Seti. But Einstein was prefered and happy now that i dont buyed it :/ Only Seti would be interessing now, how fast this card can compute one of them.

On another thread you can see 430PCIe(!) with, i think 1 hour 40 per WU, runtime. That really hurts PCI Version needs over 14 hours O.o

DSKAG Austria Research Team: [LINK]http://www.research.dskag.at[/LINK]

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,625
Credit: 89,660,889,859
RAC: 59,582,667

If you look through the

If you look through the taskID output for that task on the website, there appears to be about 9.5 hours where time was elapsing but no crunching was being done. Perhaps your card was running another project or doing something else during that period. If you add up the actual crunching time (where checkpoints were being committed) it's much less than the run time you quote. Maybe you should try a couple more tasks where nothing else is interfering. The card is probably a lot better than you think.

I don't have any GPU crunching experience so I don't know why the 9.5 hour gap is there.

Cheers,
Gary.

mickydl*
mickydl*
Joined: 7 Oct 08
Posts: 39
Credit: 197,778,345
RAC: 70,897

RE: If you look through the

Quote:

If you look through the taskID output for that task on the website, there appears to be about 9.5 hours where time was elapsing but no crunching was being done. Perhaps your card was running another project or doing something else during that period. If you add up the actual crunching time (where checkpoints were being committed) it's much less than the run time you quote. Maybe you should try a couple more tasks where nothing else is interfering. The card is probably a lot better than you think.

I don't have any GPU crunching experience so I don't know why the 9.5 hour gap is there.

The 9 hour gap is probably the night time when the machines are switched of.
I don't run them 24/7. I didn't take the time to actually go through the log file and add up all times. The reported times are the run times as reported by the project.

I remember calculating the estimated run time after the first 1.5 hours into the WU. Based on the progress bar and the run time at that point a came up with a total estimated run time of over 14 hours. So it appears to be right. However, I will run a couple of more WUs to test the card.

Michael

Claggy
Claggy
Joined: 29 Dec 06
Posts: 560
Credit: 2,631,841
RAC: 31

I got my PCI GT430 a couple

I got my PCI GT430 a couple of days ago, just done a couple of Seti@home bench runs with it,
then i set up a Second Boinc Installation on my E8500, and set it running Seti Cuda only (using an Alpha app):

http://setiathome.berkeley.edu/show_host_detail.php?hostid=6199212

A mid AR task there is taking around 4,400 seconds, compare it to my PCI 8400 GS that takes around 14,500 secs (using the Stock app at the moment):

http://setiathome.berkeley.edu/show_host_detail.php?hostid=5306154

Claggy

mickydl*
mickydl*
Joined: 7 Oct 08
Posts: 39
Credit: 197,778,345
RAC: 70,897

I now have made a second test

I now have made a second test and the result stays the same. BRP4 WUs take 14h14m with the GT 430. I have also checked my first test WU and the result is also 14h14m as stated in one of my previous mails.

So my conclusion doesn't change. For Einstein the card is practically useless unless you have a very low-end CPU.

Michael

Amauri
Amauri
Joined: 12 Jul 11
Posts: 7
Credit: 30,749,249
RAC: 10,530

My GT520 PCIe (Linux) takes

My GT520 PCIe (Linux) takes 2h39m to run an E@H WU. And I Love its low power consumption...

microchip
microchip
Joined: 10 Jun 06
Posts: 50
Credit: 98,074,127
RAC: 0

My PCIe GT 440 does an E@H WU

My PCIe GT 440 does an E@H WU in 90 minutes. I originally planned to buy the GT 520 but after I read some reviews and saw the performance being so bad, I went for the GT 440. Some reviews said the GT 520 is so bad, it shouldn't even exist :)

Amauri
Amauri
Joined: 12 Jul 11
Posts: 7
Credit: 30,749,249
RAC: 10,530

RE: My PCIe GT 440 does an

Quote:
My PCIe GT 440 does an E@H WU in 90 minutes. I originally planned to buy the GT 520 but after I read some reviews and saw the performance being so bad, I went for the GT 440. Some reviews said the GT 520 is so bad, it shouldn't even exist :)

GT 440 = 311 GFlops / 65 W = 4.78 GFlops/W
GT 520 = 155 GFlops / 29 W = 5.34 GFlops/W

When you have a low power consumption requirement, GT 520 is better than GT 440. I don't need a very good performance, just want to run some WU's without heating my computer.

Cheers,
Amauri

dskagcommunity
dskagcommunity
Joined: 16 Mar 11
Posts: 87
Credit: 949,978,246
RAC: 392,865

RE: I now have made a

Quote:

I now have made a second test and the result stays the same. BRP4 WUs take 14h14m with the GT 430. I have also checked my first test WU and the result is also 14h14m as stated in one of my previous mails.

So my conclusion doesn't change. For Einstein the card is practically useless unless you have a very low-end CPU.

Michael

Ok thx for the test again, very bad result :/

DSKAG Austria Research Team: [LINK]http://www.research.dskag.at[/LINK]

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.