| Author |
Message |
|---|
 Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13513
ID: 53948
Credit: 237,712,514
RAC: 0
                           
|
|
When I installed the GTX460 last week, WUs were taking about 830 seconds and the GPU was running at 90%. More recently, WUs are taking about 750 seconds and the GPU is running at about 98%.
Did something changed in the application or the command line parameters?
____________
My lucky number is 75898524288+1 |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
                
|
|
Michael did you see my post in the thread App_info file be created for 2 Wu at once with Gtx 460 Just like Seti@home?
~750sec for one unit
~1100sec for two units
Must be a changed command line parameter because http://www.primegrid.com/download/?C=M;O=D lists no newer cuda-apps.
____________
Best wishes. Knowledge is power. by jjwhalen
 |
|
|
|
|
|
On my GTX465, WUs are still finishing in the range 480~520 sec, as they have since Rytis broke the Fermi barrier. (I recently up-clocked the GPU slightly, so they are now at the lower end of that range.) I confirm that the GPU is running at an astonishing 98%.
What I have noticed is that the more recent 514nnn & 515nnn series WUs are downloading with ETC estimates that are much higher, in the neighborhood of 1.5 hrs oo) But they are still finishing in the same ~8.5 min. Go figure.
____________
  |
|
|
rroonnaallddVolunteer developer
Volunteer tester
Send message
Joined: 3 Jul 09
Posts: 1213
ID: 42893
Credit: 34,634,263
RAC: 0
|
|
I think the estimates are raised to solve the problem with "exceeded elapsed time" on ATI-cards.
____________
Best wishes. Knowledge is power. by jjwhalen
|
|
|
|
|
I think the estimates are raised to solve the problem with "exceeded elapsed time" on ATI-cards.
Makes sense I guess. I'm receiving an awful lot of resends that have errored out multiple times on ATI cards before getting to me. |
|
|
|
|
|
Tried 2 WUs/GPU on a few different cards including a GTX 460. If running 2 at a time is any faster at all it's negligible on my machines. Same with both Collatz and the newer MW WUs...
|
|
|
|
|
Tried 2 WUs/GPU on a few different cards including a GTX 460. If running 2 at a time is any faster at all it's negligible on my machines. Same with both Collatz and the newer MW WUs...
It was faster with the initial short WUs (ppsieve, 1G, 1M exponent range) because BOINC required 4-5 seconds to start a new WU after a WU completed. Since the runtimes were in the 135 seconds range on my card (GTX 460) the throughput increased by 3%-4% by eliminating the idle times this way.
Today the WUs are longer (tpsieve, 3G, 3M exponent range) and the gain in throughput you can expect from this is below 1% and simply not worth the hassle.
Note: I don't know how ATI cards behave in this regard (I'm not eager to put my 4770 into the box just to check this). The OpenCL app has changed significantly since the last time I've conducted experiments with ATI cards.
____________
|
|
|
|
|
Tried 2 WUs/GPU on a few different cards including a GTX 460. If running 2 at a time is any faster at all it's negligible on my machines. Same with both Collatz and the newer MW WUs...
It was faster with the initial short WUs (ppsieve, 1G, 1M exponent range) because BOINC required 4-5 seconds to start a new WU after a WU completed. Since the runtimes were in the 135 seconds range on my card (GTX 460) the throughput increased by 3%-4% by eliminating the idle times this way.
Today the WUs are longer (tpsieve, 3G, 3M exponent range) and the gain in throughput you can expect from this is below 1% and simply not worth the hassle.
Pretty much the same result that I had with my GTX 460. Interestingly my GTX 260 and GT 240 cards were several percentage points slower when running 2 WUs at a time.
|
|
|
