By chance seen this thread, I hope someone might still be interested in this 'fresh' data from my own experience (especially because my times are LOWER/SHORTER than ALL I saw posted here).
My times (I ALWAYS crunch only 1 GPU task at time, no matter which project!!!):
1 WU: 1620 s(+/(-) around 30-50 seconds at MAX on average, but if you will a rough number, which everyone can check out, I'd say it is an average of around 1630 s - )
I use the latest NVIDIA drivers (also BETA), but in last months NONE of them had significant effect on crunching time.
I'd also mention, that 99,99% of my WUs are VALID and almost ALWAYS I crunch together with 2 or 3 (mostly) WCG CPU WUs.
Most of other relevant software and hardware info can be checked out at looking at the specs of my computer I use (if more specific info is needed, I'll provide also that).
Regards.
Denis
Thank You for the link for the benchmarking software, I downloaded it already. :)
* this card is FACTORY OC on 970 MHz and I NEVER 'tamper' with even higher frequencies (as I know for sure, that some people do, and I'd strongly advise against it), on the contrary, when I crunch GPUGRID (long) tasks, I MUST decrease the clock to 900 MHz to be successful!
“A little knowledge is a dangerous thing. So is a lot.” - Albert EINSTEIN
i have several results too: GTX 560 Ti ----> 1x~1,100, 2x 2,000, 4x 4,100, 5x 5,200
GTS 450 ----> 1x~2,200, 2x 4,200
GT 240 ----> 1x~5,400, 2x 10,500
The GTX 560 TI and GTS 450 values look good to me, but the GT 240 value is a bit too pessimistic. I use one in a Linux box of mine and I get around 3500 sec per WU (quite impressive for a GPU that old with just 96 cores).
HD 7950 ------> 1x 1,145
GTX 560 Ti ----> 1x~1,100
7950 outperforms 560 ti by 2 times counting in tflops, but has the same WU's crunch time.
Wherefrom do you get these crunching-times?
My HD6950 crunches 2wu's in 62 min win7 64 i7 two cores left free 5632 GFLOPS peak
My GTX550ti crunches 2 wu's in 53min win7 64 i3 two cores left free 891 GFLOPS peak
In general I agree, nVidia crunches faster in many projects, but the CNG-Cards (HD7xxx series) should perform much better than your crunching-times show.
Maybe you have a setup-problem?
Alex, Astrocrab,
The HD 7950 (or any card for that matter) results posted can be misleading. It depends on the PCIe bus and paired CPU. The 1145 seconds for the HD 7950 was my number sent in based on an older system (see the list below). I think they perform quite well on Einstein. I use the AMD cards because of the DP performance on MW where I also contribute. Here are my current HD7950 (3GB memory) times for Einstein (no overclocking):
Runtimes at 1x vary from 960 to 1860 seconds depending on the support for the GPU. The HD 7950 is a PCIe 3.0 card. Maybe when running 2x the two tasks don't interface well with PCIe 1.1 and 2.0.
By the way System-1 is now running an NVidia GTX 560Ti (PCIe 2.0). Runtimes are
The GTX 560Ti is rock solid running 2x on System-1 (PCIe 2.0)
The HD7950s are rock solid running 2x on System-2 (PCIe 3.0), but fail miserably at 2x on the 1.1 and 2.0 systems!
The HD7950 on the right system is considerably faster than GTX 560Ti.
Hi! By chance seen this
)
Hi!
By chance seen this thread, I hope someone might still be interested in this 'fresh' data from my own experience (especially because my times are LOWER/SHORTER than ALL I saw posted here).
I'm talking about the Gigabyte NVIDIA GeForce GTX 550 Ti* (official site: http://www.gigabyte.eu/products/product-page.aspx?pid=4070#ov) and the results achieved when crunching the Arecibo GPU tasks:
My times (I ALWAYS crunch only 1 GPU task at time, no matter which project!!!):
1 WU: 1620 s(+/(-) around 30-50 seconds at MAX on average, but if you will a rough number, which everyone can check out, I'd say it is an average of around 1630 s - )
I use the latest NVIDIA drivers (also BETA), but in last months NONE of them had significant effect on crunching time.
I'd also mention, that 99,99% of my WUs are VALID and almost ALWAYS I crunch together with 2 or 3 (mostly) WCG CPU WUs.
Most of other relevant software and hardware info can be checked out at looking at the specs of my computer I use (if more specific info is needed, I'll provide also that).
Regards.
Denis
Thank You for the link for the benchmarking software, I downloaded it already. :)
* this card is FACTORY OC on 970 MHz and I NEVER 'tamper' with even higher frequencies (as I know for sure, that some people do, and I'd strongly advise against it), on the contrary, when I crunch GPUGRID (long) tasks, I MUST decrease the clock to 900 MHz to be successful!
“A little knowledge is a dangerous thing. So is a lot.” - Albert EINSTEIN
RE: i have several results
)
The GTX 560 TI and GTS 450 values look good to me, but the GT 240 value is a bit too pessimistic. I use one in a Linux box of mine and I get around 3500 sec per WU (quite impressive for a GPU that old with just 96 cores).
http://einsteinathome.org/task/313823581
Cheers HB
I have a new value for your
)
I have a new value for your list. (non opencl)
260GTX = 1*2200 secs, with 2 wus exact the double time. (WinXP)
and for opencl:
560TI (384) = 2*2150secs (winXP)
DSKAG Austria Research Team: [LINK]http://www.research.dskag.at[/LINK]
hm cant edit anymore.. 2*2030
)
hm cant edit anymore.. 2*2030 secs is the right value for the 560TI (384). Missread something ^^
DSKAG Austria Research Team: [LINK]http://www.research.dskag.at[/LINK]
RE: I have a new value for
)
wow, so there's no improvement in efficiency from running multiple tasks simultaneously on a GTX 260, eh?
not really. they run @95-96%
)
not really. they run @95-96% . the 560 gets a little improvment from 92 to 95 or so when i remmeber correct.
DSKAG Austria Research Team: [LINK]http://www.research.dskag.at[/LINK]
RE: RE: HD 7950 ------>
)
Alex, Astrocrab,
The HD 7950 (or any card for that matter) results posted can be misleading. It depends on the PCIe bus and paired CPU. The 1145 seconds for the HD 7950 was my number sent in based on an older system (see the list below). I think they perform quite well on Einstein. I use the AMD cards because of the DP performance on MW where I also contribute. Here are my current HD7950 (3GB memory) times for Einstein (no overclocking):
System-1 i5-2500K, PCIe 2.0 x16 1145 seconds running 1x (errors at 2x)
System-2 i7-3770K, PCIe 3.0 x8 (2cards) 960 seconds each card running 1x
System-2 i7-3770K, PCIe 3.0 x8 (2cards) 1390 seconds each card running 2x
System-4 Pentium E5300, PCIe 1.1 x16 1860 seconds running 1x (errors at 2x)
Runtimes at 1x vary from 960 to 1860 seconds depending on the support for the GPU. The HD 7950 is a PCIe 3.0 card. Maybe when running 2x the two tasks don't interface well with PCIe 1.1 and 2.0.
By the way System-1 is now running an NVidia GTX 560Ti (PCIe 2.0). Runtimes are
System-1 i5-2500K, PCIe 2.0 x16 1360 seconds running 1x
System-1 i5-2500K, PCIe 2.0 x16 2135 seconds running 2x
The GTX 560Ti is rock solid running 2x on System-1 (PCIe 2.0)
The HD7950s are rock solid running 2x on System-2 (PCIe 3.0), but fail miserably at 2x on the 1.1 and 2.0 systems!
The HD7950 on the right system is considerably faster than GTX 560Ti.
Happy crunching,
Joe
Edit: These are 1.28 times.
some new results: HD 7970
)
some new results:
HD 7970 ----> 1x~650, 2x~950, 4x~1,800, 5x~2,200
RE: some new results: HD
)
Wow! Nice times! It smokes my HD7950s.
Hummm.. I know what I want for Christmas!
Joe
Edit: HD 7970 Gigahertz edition?
yep, this one
)
yep,
this one http://www.gigabyte.com/products/product-page.aspx?pid=4290#ov
even under 98% load it is not even hot, just very warm. aticonfig reports 56C°
for example, 560Ti under heavy load come up to 67C°. and it's hot indeed.