Folding Forum

is this about the max i will get out of my pc. current spec

core i7
3gb ddr3
2x 9800gt's
1 gpu client on each card
and 2x vmware clients running on my i7

re The 9800GT's, no. I run 9800GT's at present and get 5765PPD on them on project 576X's . You could overclock the shaders on them via rivatuner or similar. Having said that they are the best projects at present for ppd on my system. Other projects will run as low as 3500ppd, I figure I average 4500ppd over all projects.
Your shaders will start at a default 1500, depending on your cooling you should be able to move them up closer to 1800.
Have got much higher than that but sometimes will get a string of immediate failures (EUE's) so best to be a bit conservative as EUE's not only annoying but bad for science.
You don't say what speed your cpu is running at, but at stock those numbers seem OK, you will see some discussion as to whether the i7 can run more than 2 x vmSMPs but I'm not buying into that. For comparison my Phenom II 940 @ 3.44 GHz does a little better than that with 2 x notfred's vm SMP's.
kiore.

Addit. look at what I am doing with a similar set of projects and the same config.
http://foldingforum.org/viewtopic.php?f ... 10#p101707

well i have oc'ed my cpu to 3.7ghz and my gfx cards to 650mhz core and 1750mhz shader. but my gfx cards are hitting about 95 degrees sometimes so didnt want to push them much more !!!
also i have read alot about more than 2 vmsmp's and it seems like if i have 4 running i'll get about the same total ppd as i would having just the 2. also i'd like my cpu not to be sitting at 100% usage all the time as i still have to use it lol

OK then, just overclock the shaders though, 1750 sounds good. If you're using it for other things and the the 2x vmSMPs don't slow it down too much that looks good. Hey, nearly 16k ppd is good stuff! Other people do more, so what, keep that effort up this is a marathon not a sprint
kiore.

The i7 would do much better under native linux with HT on ... I guess you can expect 7 000 PPD at least. There's a long thread about i7 performances here : viewtopic.php?f=44&t=10033

kiore wrote: look at what I am doing with a similar set of projects and the same config.
http://foldingforum.org/viewtopic.php?f ... 10#p101707

You may want to take a look at the numbers FahMon is reporting the don't add up. According to FahMon it takes you aprox 7hrs to complete a 1920 point WU = 6582 PPD but it also says you are 80% complete and still have 3hrs to go, if you look at the total time for the unit 412 min. that means the WU is 57% complete. Looks like FahMon has some serious calculation errors.

Grandpa_01 wrote:

kiore wrote: look at what I am doing with a similar set of projects and the same config.
http://foldingforum.org/viewtopic.php?f ... 10#p101707

You may want to take a look at the numbers FahMon is reporting the don't add up. According to FahMon it takes you aprox 7hrs to complete a 1920 point WU = 6582 PPD but it also says you are 80% complete and still have 3hrs to go, if you look at the total time for the unit 412 min. that means the WU is 57% complete. Looks like FahMon has some serious calculation errors.

That the asynchronous clocks business, especially noticeable on vm ware they actually take about 15 hrs. Wish they were doing 6.5k each.
kiore.

toTOW wrote:The i7 would do much better under native linux with HT on ... I guess you can expect 7 000 PPD at least. There's a long thread about i7 performances here : viewtopic.php?f=44&t=10033

Have to agree with ToTow, but if you want to run GPU on Linux, its a bit more complicated. But not impossible.

Has anyone ever tried to run the GPU client within a VM in a Linux host/Windows guest?

alpha754293 wrote:Has anyone ever tried to run the GPU client within a VM in a Linux host/Windows guest?

You don't need a VM to run the GPU2 client on a Linux host for NVIDIA cards. Wine generally works.

When i tried native linux SMP with GPU client through wine, i ran into a single but critical inconvenience. I wasnt able, with any tool i could find, to unlock the core and shader clocks while overclocking. Is there ANY tool for linux that can do that? And since my core clocks become unstable half the way than where the shaders can go, I'd lose MUCH more on GPU than what I'd gain with native linux SMP.

What do you use to overclock your NV GPU under Linux ? I know NV Clock, but I don't know if it can unlock shaders : http://www.linuxhardware.org/nvclock/

^w^ing wrote:When i tried native linux SMP with GPU client through wine, i ran into a single but critical inconvenience. I wasnt able, with any tool i could find, to unlock the core and shader clocks while overclocking. Is there ANY tool for linux that can do that? And since my core clocks become unstable half the way than where the shaders can go, I'd lose MUCH more on GPU than what I'd gain with native linux SMP.

I think the only way you can do this with Linux is to edit the BIOS itself. I don't think Coolbits under Linux lets you adjust the shaders alone. There are some BIOS editing guides available on other sites.

Yes nvclock was the most useful (if not the only) tool i could find and it couldnt unlock the shaders.

Ah well editing the bios isnt something i would resort to.