Hi, Robert, how did you meassured the power draw?
I mean, I guess its not the value of the whole system, so is that numbers the difference in the Kill-a-Watt values with the card working vs. idle or working vs. the system without the GPUs?
i can't speak for Robert, but that's exactly how i do it. obviously you can calculate a GPU's idle power consumption the same way - measure the entire system's power draw from the wall while the GPU is idling. then shut down, uninstall the GPU, restart, and measure the entire system's power draw again. then subtract that value from the former, and there you have your GPU's idle power draw.
i recently acquired both a Zotac GTX 580 3GB and a Gigabyte WindForce GTX 670, and i've been performing multiple experiments with them (and with the Gigabyte WindForce GTX 560 Ti i already own). when i'm done i'll be able to fill in the GTX 670 section of the BRP run times chart in the CUDA and OpenCL benchmarks thread, as well as glean some very useful information about the effects of PCIe bandwidth as regards Einstein@Home...
by the way, i went ahead and did that for the GTX 560 Ti (factory OCed to 900MHz), the reference GTX 580 3GB, and the GTX 670 (supposedly factory OCed to 980MHz and boosts to 1058MHz, though it constantly sits at 1110MHz under load...go figure lol). the GTX 560 Ti draws 25W at idle and 179W at 96% load, the GTX 580 3GB draws 41W at idle and 227W at 96% load, the GTX 670 draws 25W at idle and 163W at 95% load, and the 580 and 670 together draw 57W at idle and 316 under load, in case anyone's curious. obviously the load i'm talking about is Einstein@Home BRP work...
the GTX 580 3GB draws 41W at idle and 227W at 96% load, the GTX 670 draws 25W at idle and 163W at 95% load, and the 580 and 670 together draw 57W at idle and 316 under load, in case anyone's curious. obviously the load i'm talking about is Einstein@Home BRP work...
Hi Sunny. The load figures escape me. Are the GPUs still running at 95%+? While the power saving at idle should perhaps be a little less due to one GPU not being used for system graphics, there shouldn't be much difference at load. How does 163w + 227W = 316W? Are all the other processes the same? Are there perhaps fewer CPU WUs running? When I add GPUs to a system I'm not seeing this kind of behavior at all.
I have been running a second PSU for graphics via one of my systems and was able to capture power draw for just the graphics card. My GTX 680 draws between 100-120w in Linux while running 1 to 3-tasks at once. My 7970 draws between 155 and 163w in Linux running 3-tasks at once and 96% GPU load. I know the 680 can draw upwards of 200w while running a GPU stress test but 120w is about the highest I can get for power draw with BRP4. Idle power consumption is similar between both cards at around 20w. Fan speed also has an effect on power draw and the NVIDIA card I run at 80% and the AMD card at 70%.
I have been running a second PSU for graphics via one of my systems and was able to capture power draw for just the graphics card. My GTX 680 draws between 100-120w in Linux while running 1 to 3-tasks at once. My 7970 draws between 155 and 163w in Linux running 3-tasks at once and 96% GPU load. I know the 680 can draw upwards of 200w while running a GPU stress test but 120w is about the highest I can get for power draw with BRP4. Idle power consumption is similar between both cards at around 20w. Fan speed also has an effect on power draw and the NVIDIA card I run at 80% and the AMD card at 70%.
If you're only looking at the power being drawn by the second PSU you're overlooking the up to 75W coming from the mobo itself.
If you're only looking at the power being drawn by the second PSU you're overlooking the up to 75W coming from the mobo itself.
I completely forgot about the power draw on the PCI-E slot when typing out my original post. Based on power draw I have with the board, CPU, accessories, and three running AMD cards via the first PSU, I am estimating an additional 20-25w consumption per card via the slots. I will have to do some more testing to get a more exact number. I am currently going off what I remember the motherboard and CPU power draw was when I previously ran the system headless.
RE: Hi, Robert, how did you
)
i can't speak for Robert, but that's exactly how i do it. obviously you can calculate a GPU's idle power consumption the same way - measure the entire system's power draw from the wall while the GPU is idling. then shut down, uninstall the GPU, restart, and measure the entire system's power draw again. then subtract that value from the former, and there you have your GPU's idle power draw.
i recently acquired both a Zotac GTX 580 3GB and a Gigabyte WindForce GTX 670, and i've been performing multiple experiments with them (and with the Gigabyte WindForce GTX 560 Ti i already own). when i'm done i'll be able to fill in the GTX 670 section of the BRP run times chart in the CUDA and OpenCL benchmarks thread, as well as glean some very useful information about the effects of PCIe bandwidth as regards Einstein@Home...
by the way, i went ahead and
)
by the way, i went ahead and did that for the GTX 560 Ti (factory OCed to 900MHz), the reference GTX 580 3GB, and the GTX 670 (supposedly factory OCed to 980MHz and boosts to 1058MHz, though it constantly sits at 1110MHz under load...go figure lol). the GTX 560 Ti draws 25W at idle and 179W at 96% load, the GTX 580 3GB draws 41W at idle and 227W at 96% load, the GTX 670 draws 25W at idle and 163W at 95% load, and the 580 and 670 together draw 57W at idle and 316 under load, in case anyone's curious. obviously the load i'm talking about is Einstein@Home BRP work...
Sunny129 wrote: obviously the
)
Interesting numbers--thanks. Could you add to the %load detail the additional detail of number of simultaneous BRP jobs in each case?
RE: the GTX 580 3GB draws
)
Hi Sunny. The load figures escape me. Are the GPUs still running at 95%+? While the power saving at idle should perhaps be a little less due to one GPU not being used for system graphics, there shouldn't be much difference at load. How does 163w + 227W = 316W? Are all the other processes the same? Are there perhaps fewer CPU WUs running? When I add GPUs to a system I'm not seeing this kind of behavior at all.
I'll let you know. I just
)
I'll let you know. I just ordered a sapphire 7790 for running Einstein.
Cheers!
Thanks, I see 20% difference
)
Thanks, I see 20% difference - Capeverde vs Bonaire(aka 7790) in your results. What is your "Capeverde" card - 7770 with GDDR5?
I have been running a second
)
I have been running a second PSU for graphics via one of my systems and was able to capture power draw for just the graphics card. My GTX 680 draws between 100-120w in Linux while running 1 to 3-tasks at once. My 7970 draws between 155 and 163w in Linux running 3-tasks at once and 96% GPU load. I know the 680 can draw upwards of 200w while running a GPU stress test but 120w is about the highest I can get for power draw with BRP4. Idle power consumption is similar between both cards at around 20w. Fan speed also has an effect on power draw and the NVIDIA card I run at 80% and the AMD card at 70%.
RE: I have been running a
)
If you're only looking at the power being drawn by the second PSU you're overlooking the up to 75W coming from the mobo itself.
RE: If you're only looking
)
I completely forgot about the power draw on the PCI-E slot when typing out my original post. Based on power draw I have with the board, CPU, accessories, and three running AMD cards via the first PSU, I am estimating an additional 20-25w consumption per card via the slots. I will have to do some more testing to get a more exact number. I am currently going off what I remember the motherboard and CPU power draw was when I previously ran the system headless.