Latest GeForce WHQL drivers

[MiTuS-2]

i'm actually runnning this project: http://fah-web.stanford.edu/cgi-bin/fah ... ed?p=11217 and in the description i can read BETA TESTS. The system stability is perfect, i checked for artifacts with 3 different tools.i don't see any problems.

[MtM]

3d stability != folding stability.

Please, stop your client and start reading about the client and project first. You obviously do no know what you're talking about.

Btw 11217 is not a BETA wu, but maybe you could start reading here -> viewtopic.php?f=16&t=8

That's a thread where the real beta testers have to sign up. The fact that the description says beta might have to do with the fact that the client is in beta stages. Again, it is not the same.

Edit: to excuse the harsh tone, this guy made another thread which is now gone, I'm not being harsh because of his posts in this thread alone.

[Flathead74]

pI11217 was released to -advmethods, announcement is here:
http://foldingforum.org/viewtopic.php?p=157778#p157778

-advmethods is also considered as open, or late stage beta testing.

This can often cause confusion.

[erekfah]

are the newly released (10-18-2010) 260.89 WHQL NVIDIA GEFORCE drivers recommended for folding?

[uncle fuzzy]

Since these were just released, I don't expect to see an immediate response. The folks who watch drivers closely usually say use the newest WHQL unless problems have been reported. I follow the "keep using the old drivers until I hear about a benefit from the new ones" school.

[JimF]

I am trying them now with my GT 240 (Win7 64-bit), and am seeing a phenomenon that I thought had been left behind a couple of drivers ago. That is, the PPD varies over time, from a low of 3674 to a high of 4178 (on P5765) over a 15 minute period. Now it is on the way down again, and the whole cycle will take about 30 minutes. Whether this is a real effect or just a measurement anomaly (HFM.NET 1.5.1) I have no idea, but I will probably go back to 258.96 tomorrow after some more study. I did a really clean install by the way, removing all of the Nvidia drivers, control panel and audio from Programs and Features, then a reboot followed by a further clean with Driver Sweeper 2.5.0 and another reboot. Then I installed 260.89 using the "clean install" option. That should do it.

[bruce]

PPD predictions are based estimates from past performance. HFM does not measure ACTUAL PPD so it's really difficult to believe that those estimates can change that much that quickly.

Try it yourself. Calculate the time between the current % and the previous % and the one before that, etc. Now based on those average times, figure out how long it will take to finish the current WU and add in however long you've already been running to get a total estimated time for that WU. (Based on that time and the points per WU you can now calculate an estimated PPD.)

You'll also notice that HFM allows you to choose different methods of projecting. Is the time for just the most recent percent more accurate than the average over the past three percent? ... or since the WU started? ... or over all WUs you've done? Obviously part of that depends on what else is running in your computer that would be making FAH run slower than if it was the only thing running, but they're all just estimates and you have to decide which one works best for your system.

You may be seeing an actual anomaly, though. Remember that most WUs are checkpointed every 15 minutes. That means that for 14 minutes, the FAHcore is calculating continuously. Then the FAHcore writes checkpoint data to disk. Depending on how your disk cache works and how big those files are, the time to complete each percent may be lengthened while your computer writes the data to disk. You can adjust your checkpoint interval to some other value and see if the 15 minute cycle that you've observed changes too. It's really unlikely that it has anything to do with drivesrs.

[OzPapaSmurf]

As always, I will be guided by what more experienced Folders report about drivers.

[JimF]

It's really unlikely that it has anything to do with drivesrs.

That may very well be the case, since the variation is a lot less with the present work unit (P10515), which varies from 3962 down to 3931. I have seen that type of variation before with other drivers, although this is a dedicated GT 240 that has nothing else to do but fold. But for some reason I have never seen the larger variation with the previous driver (258.96), though that may just be a coincidence.

[sswilson]

These seem to be running fine under XP 32bit w/ a GTS450.

[seanego]

I have a problem with GTX275 on 260.89 WHQL in Win XP SP3. Noticed that if I want to watch a movie while running F@H my GTX275 (666/1548/1161) enters "Low-Power 3D" state (400/800/300) and then won't go faster until reboot. Over- or downclocking doesn't affect this. So now I'm back to 258.96 and problem's gone.

[db597]

Seems fine - I've successfully completed 3 WUs on my GTX460 with 260.89.

[Liqmaticus]

Installed new 260.89 drivers running on an EVGA GTX 295 CO-OP FTW multi-GPU card. Now when I run the GPU clients both GPUs display 99% GPU being used, but after about 5 minutes one of the GPUs goes down to about 25-35% from there on even if I restart the client. I have to wipe out the folder and restart fresh and the process restarts itself. 99% for a few minutes and then way down to 25% to 35% usage. Reverting back to 258.96 where problem disappears.

[Liqmaticus]

Nevermind. 258.96 having same issue and if I try to disable multi-gpu on my 295 one GPU client comes back UNSTABLE_MACHINE yet disabling multi-gpu on my other machine with a 9800GX2 worked fine to get both GPUs working at 99%. Odd.

[Liqmaticus]

So it appears my problem is when I have my four cores on my CPU crunching away it brings down one of my GPUs to about 25% usage. If I release one of my CPU cores the effected GPU goes up about 20% to 25% where as the other GPU stays at a constant 99%. If all four of my CPU cores are released both my GPUs are at 99%. Anyone heard of CPU cores effecting GPU cores like this?