Message boards :
Number crunching :
Energy efficient GPU for servers?
Message board moderation
Author | Message |
---|---|
Bil Send message Joined: 27 Jan 01 Posts: 76 Credit: 1,887,795 RAC: 0 |
Hello! Anyone investigating questins about low level videocards like as geforce 210, start range ati and so on - say, with energy consumption up to 20 watts, 30 watts, 40 and 50? i think, there around is lot of various 24/7 servers, where videocard not very important, and there be purchased a cheap videocards - but what is better, and what card no? |
ivan Send message Joined: 5 Mar 01 Posts: 783 Credit: 348,560,338 RAC: 223 |
Hello! Currently the Rev. 2 DDR5 GT640 (1046 MHz) is cheap(ish) and has one of the higher Flops/Watt ratings on the Wiki page (49 W, 16.39 GFlops/W). It doesn't need auxiliary power leads, the PCI-e slot provides enough. I got one (used) for £50, from memory they are about £80 or so new. |
Bil Send message Joined: 27 Jan 01 Posts: 76 Credit: 1,887,795 RAC: 0 |
thanks, thats good info. it prices starts about 70 eur there. before that i slowly thinking about 550ti for about 100eur.... |
ivan Send message Joined: 5 Mar 01 Posts: 783 Credit: 348,560,338 RAC: 223 |
thanks, thats good info. it prices starts about 70 eur there. before that i slowly thinking about 550ti for about 100eur.... Oh, and the good thing about mine is that it is on a single-width card and comes with a mounting bracket for a low-profile case. Make sure you're looking at the 1046 MHz version though, there are several models all called GT 640. |
Bil Send message Joined: 27 Jan 01 Posts: 76 Credit: 1,887,795 RAC: 0 |
ok, thank you. sheduled from next salary. http://aio.lv/lv/pc-components/graphic-cards/asus-90yv04a0-m0na00/ that is right one ( rev 2.0) ? P.S. 630 rev 2.0 looks even more efficiency and good - for price less 50 eur, peak perf there is about 690 gflops, power consumption about 25 watt, and there is 27 gflops per watt vs 16 on geforce 640! looks like lot more efficient. i think, i buy one 630 from asus first. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
A PCIe x16 slot is designed to deliver up to 75w. So you can go with a slightly more powerful card if you wish. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
ivan Send message Joined: 5 Mar 01 Posts: 783 Credit: 348,560,338 RAC: 223 |
ok, thank you. sheduled from next salary. That's the business! Looks like it's formatted for single-width low-profile too, like my Gigabyte. As for which way to go, it depends on where your priority lies - cost, power consumption, or processing speed. As the old adage goes, "choose any two"! :-) |
Frosch Send message Joined: 13 Apr 01 Posts: 3 Credit: 1,467,450 RAC: 0 |
I had both Cards in my System (GT630 + GT640). The GT640 is in practise 25 % faster than the GT630, measured in RAC on a Core2Duo 2.80 Ghz System with Einstein@home. |
Bil Send message Joined: 27 Jan 01 Posts: 76 Credit: 1,887,795 RAC: 0 |
good. i ordered 630, and get it in next week, i think. if differnce is only 25%, 630 looks better - 630 TDP is 25 watts, 540 - 49 watt or so - i e 640 consume alnmsot x2 power, and get only +25%, not +100% :: ok, i get it and bit test, and then it, maybe, go to my small home server. and then i do think, who put on my home desktop - maybe any bit faster, for some gaming ( nexuiz?)? i not be very big gamer novadays ( not like when i was young, pentium 166mmx and quake and unreal 1 go...:) ) on another hand -i get some years on nvidia geforce 7600 GS on my desktop, and now i have geforce 9400GT - i think GT630 is better and significantly more performance than any of that... thanks all for an info... PS yes, my target on 24/7 systems at home is better performance with minimal energy consumption, as electricity grow in prices every few years. not 1 kwh is about 0.16 eur - it becomes 100 kwh = 16 eur. in a year bills perspective it going to be very expensive, a get a 200w or even more, on videocard... |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
One more thing to consider is non-linear speed improvement with increase of CU number in GPU. It's definitely true for all OpenCL variants, it was true for original CUDA app and I believe it's still true for current CUDA app also (just didn't look the code some time). The reason in some limit in size of data chunk to process at once. This data chunk is limited by algorithm used and though some apps like AP can shift size in some degree, that shift adds some overhead. Hence, if one use GPU with let say twice more number of CUs (and twice more TDP/twice cost) one never got twice performance on it. (EDIT: good news that twice CUs maybe will not consume twice more though ;) ) SETI apps news We're not gonna fight them. We're gonna transcend them. |
Frosch Send message Joined: 13 Apr 01 Posts: 3 Credit: 1,467,450 RAC: 0 |
TDP is not exactly what you are looking fore... Both the GT630 and GT640 deactivate parts of the Chip if they are not needed, a GT630 running CUDA with 99% load, but "sleeping" Graphic device runs under 15 Watt (Hard to measure...), overclocked to 1050 Mhz but with 41° Celsius. As soon as Graphic is activated (showing only a "standing" Desktop) the temperature of the Chip jumps to 53° Celsius. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.