Win 10 pushed 388.13 driver

[JonasTheMovie]

Noticed that my rig had not delivered any points the last 18 hours, opened the VNC, lo and behold, fah_21 had crashed and a reboot splash was covering the screen.
After the reboot the rest of the WU shows a quarter of the normal ppd, but I guess that has to do with the slow return time due to the crash.

But, whats up with Win 10 pushing a driver update even thou I have turned all auto updates off? Why even push a Nvidia update?
Anyone know anything if 388.13 is worth to keep or should I revert back to an older?

[JimboPalmer]

Nvidia has 390.65 out, I would always use the Nvidia version. (or AMD for an AMD card)

https://www.geforce.com/drivers/results/129081

[bruce]

When you run WIn10, you really can't revert to an older driver. They seem to push NV drivers soon after NV releases them. In many cases, they have REMOVED the OpenCL drivers, which crashes FAHCore_21.

I have conflicting information about whether they're now updating the OpenCL drivers at the same time or not. Personally, I'd appreciate reports about whether they did or not.

I try to update my drivers from NV before Win10 pushes the new ones so that OpenCL gets updated, while Core_21 isn't running.

[JimboPalmer]

I would 'finish' the current work Unit then; I would uninstall every nvidia 'feature' but the driver, then do a clean install of just the driver, if it was me.

[foldy]

@TAMUmpower: The first problem with crashed work unit is because you need to pause folding before nvidia driver update and resume folding after.

The second problem with lower PPD I guess is because you got a new work unit and there are big work units with higher PPD and small work units with lower PPD.

Did you try to reboot your PC? Is the GPU clock at ~1900 Mhz again?

[TAMUmpower]

Ah thanks. Yea I've been pausing it now and it saves the WO. I've rebooted a ton of times trying different drivers and settings. The clock has remained the same its just the utilization that seems low. It's just odd because I saw it able to pull 92-96% on the GPU already so I know there isnt a hardware issue. Realistically it should be able to just sit at 100% like all the other systems I've folded on.

[foldy]

It is still a bug in FAH or Windows 10 that is push installs a GPU driver while GPU application is running. Maybe Windows 10 only has a check for DirectX games and not for OpenCL apps. Or FAH needs to set a magic Windows flag to signal GPU is used.

[bruce]

I have WindowsUpdate configured to ask me what I want to install. (On Win10-home, that doesn't prevent them from forcing the driver update, but it does delay it). I saw the planned update to 388.13 but I was already on a later version from NV. WindowsUpdate still wanted to install 388.13 Anyway, it took several manual changes including pausing/resuming the active WU, but it's happy now.

I would 'finish' the current work Unit then; I would uninstall every nvidia 'feature' but the driver, then do a clean install of just the driver, if it was me.

That's what I do, too, unless my sound is using the speakers on the monitor.

[JonasTheMovie]

So far the 388.13 has proved to be more productive than the earlier "best" driver I used. PPM has been increased 7-9% .

[bruce]

A project=to-project variation of +-10% is not uncommon. Unless you're seeing +7-9% ON THE SAME PROJECT, you really don't know what you think you know.

[JonasTheMovie]

So, after months of steady levels of project returns the available projects just happened to changed at the same time as the drivers updated...
Looking at multiple GPU users an increase of ppd can been seen a couple of days before my win10 pushed the update to my computer.
Since the stat server was down just those days it´s hard to tell anything more from the overall user stats.

Never the less, ppd´s have increased at the same time as my driver updated, not at the same time as it increased for other people.

[JimboPalmer]

It does happen, when I first got my GTX 1050ti last April, I was getting about 150,000. Now I average 170,000. so new has been faster for me.