I have the same results:
GTX 560 TI => 1x ~1115s, 2x ~2000s
(WCG CPU tasks running in all CPU cores.)
Try leaving one cpu core free and see if the gpu time goes down, your gpu's COULD be waiting for cpu time.
Just finished testing. Leaving 1 core or 4 cores free while running 1 or 2 tasks simultaneously has absolutely no effect on completion times, they always average to 1x 1110s and 2x 2000s.
I have the same results:
GTX 560 TI => 1x ~1115s, 2x ~2000s
(WCG CPU tasks running in all CPU cores.)
Try leaving one cpu core free and see if the gpu time goes down, your gpu's COULD be waiting for cpu time.
Just finished testing. Leaving 1 core or 4 cores free while running 1 or 2 tasks simultaneously has absolutely no effect on completion times, they always average to 1x 1110s and 2x 2000s.
Hmmm I have just returned to Einstein for a while and also have two Nvidia 560Ti's, but they are in different machines. Both are doing two units at once but have vastly different times to complete, one is 6,000 seconds while the other is only taking half that. One is taking 6,174.67(gpu) 1,774.57(cpu) whjle the other is taking 3,141.64(gpu) 1,052.54(cpu). The slow one is NOT using the dual core cpu for crunching at all while the faster other is a quad core and is using 3 of the 4 cores for crunching. I am guessing it is the clock speed of the cpu and the system ram that is the big difference. They are both Boinc ONLY machines. The slow one is: Intel(R) Pentium(R) D CPU 2.66GHz.
while the faster one is: Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz.
They are both running 64 bit Win7 Ultimate but one has 4gb of ram while the slower one only has 3gb of ram. So neither is the newest but both do okay, I guess after Christmas I will have to consider decommissioning the dual core and going with another 6 or 8 core. My newer 6 and 8 core machines have AMD gpu's in them, crunching elsewhere, but also have 16gb of ram.
Hmmm I have just returned to Einstein for a while and also have two Nvidia 560Ti's, but they are in different machines. Both are doing two units at once but have vastly different times to complete, one is 6,000 seconds while the other is only taking half that. One is taking 6,174.67(gpu) 1,774.57(cpu) whjle the other is taking 3,141.64(gpu) 1,052.54(cpu). The slow one is NOT using the dual core cpu for crunching at all while the faster other is a quad core and is using 3 of the 4 cores for crunching. I am guessing it is the clock speed of the cpu and the system ram that is the big difference. They are both Boinc ONLY machines. The slow one is: Intel(R) Pentium(R) D CPU 2.66GHz.
while the faster one is: Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz.
They are both running 64 bit Win7 Ultimate but one has 4gb of ram while the slower one only has 3gb of ram. So neither is the newest but both do okay, I guess after Christmas I will have to consider decommissioning the dual core and going with another 6 or 8 core. My newer 6 and 8 core machines have AMD gpu's in them, crunching elsewhere, but also have 16gb of ram.
That doesn't sound right; I get 1400s on a GTX460/Core2 E6300, and 1600s on a GTX260/Pentium 4 3Ghz (e.g. this host which I've just set up).
Could it be related to PCIe? It does seem PCIe 2.0 x16 is needed to really use the GPUs; I was getting 2,500s with the same GTX260 in a PCIe 1.0 slot.
Hmmm I have just returned to Einstein for a while and also have two Nvidia 560Ti's, but they are in different machines. Both are doing two units at once but have vastly different times to complete, one is 6,000 seconds while the other is only taking half that. One is taking 6,174.67(gpu) 1,774.57(cpu) whjle the other is taking 3,141.64(gpu) 1,052.54(cpu). The slow one is NOT using the dual core cpu for crunching at all while the faster other is a quad core and is using 3 of the 4 cores for crunching. I am guessing it is the clock speed of the cpu and the system ram that is the big difference. They are both Boinc ONLY machines. The slow one is: Intel(R) Pentium(R) D CPU 2.66GHz.
while the faster one is: Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz.
They are both running 64 bit Win7 Ultimate but one has 4gb of ram while the slower one only has 3gb of ram. So neither is the newest but both do okay, I guess after Christmas I will have to consider decommissioning the dual core and going with another 6 or 8 core. My newer 6 and 8 core machines have AMD gpu's in them, crunching elsewhere, but also have 16gb of ram.
That doesn't sound right; I get 1400s on a GTX460/Core2 E6300, and 1600s on a GTX260/Pentium 4 3Ghz (e.g. this host which I've just set up).
Could it be related to PCIe? It does seem PCIe 2.0 x16 is needed to really use the GPUs; I was getting 2,500s with the same GTX260 in a PCIe 1.0 slot.
i was thinking the same thing...i can't imagine even an ancient CPU would double a GTX 560 Ti's crunch times. we have to keep in mind that even when comparing two different platforms, one with a CPU only half as powerful as the other, a bulk of the computation is done on the GPU, not the CPU.
i was thinking the same thing...i can't imagine even an ancient CPU would double a GTX 560 Ti's crunch times. we have to keep in mind that even when comparing two different platforms, one with a CPU only half as powerful as the other, a bulk of the computation is done on the GPU, not the CPU.
When I had that GTX 560 TI in Phenom IIX6 1055 I got 1x ~1420s times (now ~1110s in i7-3770).
Then some new data for a card that is still missing from the database.
GTX 660 factory over clocked card: 1x ~1200s, 2x ~2020s
GTX 660 factory over clocked card: 1x ~1200s, 2x ~2020s
Now that more tasks have completed the average of the last 30 tasks has settled to 2x ~2010s. The card seems to be identical to GTX 560 TI in terms of number crunching power. I have to say that Asus DirectCU II is pretty awesome cooler, the new card is barely audible even under full load.
I tested using the Application Configuration, or app_config.xml file in BOINC 7.0.42 on my HD6850 - 2GB, still using Catalysts 12.4, even though some think that those drivers don't have OpenCL support under Windows 7. ;-)
Set it up for Einstein only, using the following entries:
[pre]
einsteinbinary_BRP4
2
0.50
0.45
[/pre]
Outcome: task one ran for 3,826.45 seconds. task two ran for 3,820.36 seconds.
Single task run time is about 2,300 seconds, so running two has a higher efficiency. Checking with GPU-Z temperature was only about 2-3 degrees higher than on single runs, 66-67C against 64C. GPU load on a single task is around 80%, on two tasks around 96%.
But since I only have one videocard and no immediate need to blow this one up, I'll go back to running one at a time. :)
RE: GTX 560 Ti ---->
)
I have the same results:
GTX 560 TI => 1x ~1115s, 2x ~2000s
(WCG CPU tasks running in all CPU cores.)
RE: RE: GTX 560 Ti ---->
)
Try leaving one cpu core free and see if the gpu time goes down, your gpu's COULD be waiting for cpu time.
RE: RE: RE: GTX 560 Ti
)
Just finished testing. Leaving 1 core or 4 cores free while running 1 or 2 tasks simultaneously has absolutely no effect on completion times, they always average to 1x 1110s and 2x 2000s.
RE: RE: RE: RE: GTX
)
Hmmm I have just returned to Einstein for a while and also have two Nvidia 560Ti's, but they are in different machines. Both are doing two units at once but have vastly different times to complete, one is 6,000 seconds while the other is only taking half that. One is taking 6,174.67(gpu) 1,774.57(cpu) whjle the other is taking 3,141.64(gpu) 1,052.54(cpu). The slow one is NOT using the dual core cpu for crunching at all while the faster other is a quad core and is using 3 of the 4 cores for crunching. I am guessing it is the clock speed of the cpu and the system ram that is the big difference. They are both Boinc ONLY machines. The slow one is: Intel(R) Pentium(R) D CPU 2.66GHz.
while the faster one is: Intel(R) Core(TM)2 Quad CPU Q8200 @ 2.33GHz.
They are both running 64 bit Win7 Ultimate but one has 4gb of ram while the slower one only has 3gb of ram. So neither is the newest but both do okay, I guess after Christmas I will have to consider decommissioning the dual core and going with another 6 or 8 core. My newer 6 and 8 core machines have AMD gpu's in them, crunching elsewhere, but also have 16gb of ram.
RE: Hmmm I have just
)
That doesn't sound right; I get 1400s on a GTX460/Core2 E6300, and 1600s on a GTX260/Pentium 4 3Ghz (e.g. this host which I've just set up).
Could it be related to PCIe? It does seem PCIe 2.0 x16 is needed to really use the GPUs; I was getting 2,500s with the same GTX260 in a PCIe 1.0 slot.
Only for notice, all relevant
)
Only for notice, all relevant last values updated.
http://www.dskag.at/images/Research/EinsteinGPUperformancelist.pdf
DSKAG Austria Research Team: [LINK]http://www.research.dskag.at[/LINK]
RE: RE: Hmmm I have just
)
i was thinking the same thing...i can't imagine even an ancient CPU would double a GTX 560 Ti's crunch times. we have to keep in mind that even when comparing two different platforms, one with a CPU only half as powerful as the other, a bulk of the computation is done on the GPU, not the CPU.
RE: i was thinking the same
)
When I had that GTX 560 TI in Phenom IIX6 1055 I got 1x ~1420s times (now ~1110s in i7-3770).
Then some new data for a card that is still missing from the database.
GTX 660 factory over clocked card: 1x ~1200s, 2x ~2020s
RE: GTX 660 factory over
)
Now that more tasks have completed the average of the last 30 tasks has settled to 2x ~2010s. The card seems to be identical to GTX 560 TI in terms of number crunching power. I have to say that Asus DirectCU II is pretty awesome cooler, the new card is barely audible even under full load.
I tested using the
)
I tested using the Application Configuration, or app_config.xml file in BOINC 7.0.42 on my HD6850 - 2GB, still using Catalysts 12.4, even though some think that those drivers don't have OpenCL support under Windows 7. ;-)
Set it up for Einstein only, using the following entries:
[pre]
einsteinbinary_BRP4
2
0.50
0.45
[/pre]
Outcome:
task one ran for 3,826.45 seconds.
task two ran for 3,820.36 seconds.
Single task run time is about 2,300 seconds, so running two has a higher efficiency. Checking with GPU-Z temperature was only about 2-3 degrees higher than on single runs, 66-67C against 64C. GPU load on a single task is around 80%, on two tasks around 96%.
But since I only have one videocard and no immediate need to blow this one up, I'll go back to running one at a time. :)