Message boards :
Number crunching :
Similiar perfomance replacement suggestions for GTX295 Co-op
Message board moderation
Previous · 1 · 2
Author | Message |
---|---|
Al Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 |
I didn't realize that the GPU's put such a strain on the CPU, I thought that once the tasks were running, the CPU was pretty much idling, but appearantly not. Would it make a difference if you had a multi-core'd CPU like a X5690, or even duals of them in a server board? I know there is a setting that you can set to reserve a certain amount of CPU for the GPU, would they need the whole CPU to keep up with the video cards? I know I've got a lot to learn, that's why I ask lots of questions :) |
.clair. Send message Joined: 4 Nov 04 Posts: 1300 Credit: 55,390,408 RAC: 69 |
Reserving one core or thread if you have an intel with HT is enugh for most systems, once you get to building monster crunchers the amount of cpu time used depends on the components that it is built with. You do not need an expensive fast cpu to feed the gpu`s, just something that `will do the job`. At the moment my main cruncher is driven by a P4 (3.6ghz prescot660 with HT) with one ati7970 and room for two more when i can afford them, the cpu does FreeHal nci work (when the project has any work to do). BOINC mangler is set to use 99% of cpu, so it will only use one core/thread for cpu work and seti is set to only send MB gpu work, it work`s out ok for me and this crunching box. |
Paul D Harris Send message Joined: 1 Dec 99 Posts: 1122 Credit: 33,600,005 RAC: 0 |
I run my i7 920 and 2 460 at 100% of cpu no core reserved for feeding my gpu all 8 core crunching with no problem at all. |
zoom3+1=4 Send message Joined: 30 Nov 03 Posts: 65690 Credit: 55,293,173 RAC: 49 |
Of course I want a serious response! :-) I have no idea if there's a limitation to the number of GPU's one machine can support in BOINC. I suppose if you use the Hydro version of the 295, and install 8 of them into a board like my SR-2 or 4-way SLI, you could technically have up to 16. That is of course if you have the money, and if the software would support it. Is there a limit? Imagine 8 Hydros working in one machine. Man! A 2000+ watt machine easy. Only 2 GTX295 cards will work properly in Vista or Windows 7, so that leaves XP or XP64 or maybe W2K... The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's |
Al Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 |
I am running XP64 in all of my rigs, it seems to work pretty well for me. I am looking for a deal on some OEM licenses again. I have 10, 2 packs of 5 with the CD which I bought on ebay a couple years ago, but they are getting harder to find. Any idea why the limit on the newer OS's? You'd think they would make it better not worse in terms of multi-GPU as OS's progress. |
Wembley Send message Joined: 16 Sep 09 Posts: 429 Credit: 1,844,293 RAC: 0 |
I am running XP64 in all of my rigs, it seems to work pretty well for me. I am looking for a deal on some OEM licenses again. I have 10, 2 packs of 5 with the CD which I bought on ebay a couple years ago, but they are getting harder to find. Any idea why the limit on the newer OS's? You'd think they would make it better not worse in terms of multi-GPU as OS's progress. Bill Gates is trying to force more and more hobbyists into buy the server versions of his software. |
Al Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 |
Server versions support more GPU's? That seems strange, since the GPU is the last thing one would think about for a server, most have a simple VGA connection, or are headless, at least when I was in it a few years back. You'd think that the Ultimate version would be the one that is unlimited, at least in terms of this type of GPU support, that would make sense (and dollars for M$), but a server version? I dunno... |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Server versions support more GPU's? That seems strange, since the GPU is the last thing one would think about for a server, most have a simple VGA connection, or are headless, at least when I was in it a few years back. You'd think that the Ultimate version would be the one that is unlimited, at least in terms of this type of GPU support, that would make sense (and dollars for M$), but a server version? I dunno... From various other project sites I have read that in Windows 7 you can get 4 ATI GPUs or 8 NVIDIA GPUs to work. They were specifying newer cards like 590's and 6900's for these configurations. The newer hardware may explain why they worked more so than the 295's. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Martin Chartrand Send message Joined: 24 Apr 04 Posts: 8 Credit: 5,122,693 RAC: 0 |
Hello everyone. Since this thread is about GPU I thought I'd throw at you some questions just for fun. I have a GTX 285 and when there is work i probably crunch between 1000 to 5000 Units MAX. With CUDA of course. How could I compare that to a GTX 680 or 690? My GTX 285 has 240 CUDA core while the 680 has 1534 and the 690 over 3000. Is it the CUDA core that determine how much one crunches? I am not that savvy and was just wondering if any of you would have a vague idea. Martin Chartrand |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13715 Credit: 208,696,464 RAC: 304 |
Is it the CUDA core that determine how much one crunches? Partly. All other things being the same, the GPU with the most cores will process more work per hour. However- memory speed, clock speed for the various different parts of the GPU itself & architecture all have a big impact on performance. As good at cruning WUs as the new Kepler GPUs are (GTX670, 680, 690) Nvidia has gone down a different path to previous video cards. In the past their highend video cards & high performance computing Tesla card were pretty much the same hardware (with only a few relatively minor differences). This time there will be 2 distict types of Kepler card- the present ones for use as video cards, and a much more powerfull version for data processing. As fast as the new cards are at crunching Seti WUs, the new Kepler based Tesla cards are expected to be significanlty faster. And have prices to match. Grant Darwin NT |
©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.