WARNING Do not upgrade to 375/376.xx drivers (for xx<48)

[Calico]

This driver has been replaced by nvidia with 375.63

Still fubar here on my 1070. Core reaches only the base clock (1594mhz) for a moment and then down clocks to idle. My screen went black for a few seconds and came back on with my monitor flickering until I rebooted the computer

[MikeNSantee]

Well the 375.63 drivers seem to have fixed my problem as well.

[Calico]

Working now after a clean 375.63 installation(clean first with DDU) but noticing different GPU behaviour. Power limit only around 70% and GPU usage is 90% instead of 100%. Before these values would be maxed out at 100%. Guess I wont complain because thermals are better.

Am hearing issue behind 375.57 was missing OpenCL.dll file

Edit: No I'm afraid it's still very much broken. I was able to complete one work unit. When it reset to start the next one, the GPU core/mem frequencies did not boost properly and my driver crashed

[Breach]

Same here, no boost beyond the base clock. With 375.57 and 373.06 it would at least go to P2 (never P0 though).

[toTOW]

This driver has been replaced by nvidia with 375.63 at http://www.nvidia.com/drivers .

If you do not have a specific reason to upgrade and your folding is working I recommend not upgrading at this time, at least until there's some assurance that nv has released a valid update. If you already did and folding is not working please reinstall the older drivers that worked.

It's broken too ... only a few small projects might be able to start, but most of core 21 projects won't ...

If you don't need to upgrade, avoid any driver of the 375.xx branch ...

[rightarmman]

375.63 latest driver Still effecting FAH ... Can someone post last stable release?
Thank you.
rightarmman
GTX1070 375.63
Win10Pro 64Gb

Code: Select all

04:02:56:WARNING:WU00:FS01:FahCore returned: BAD_WORK_UNIT (114 = 0x72)
04:03:43:WU02:FS01:0x21:ERROR:exception: Error downloading array interactionCount: clEnqueueReadBuffer (-5)
04:03:43:WARNING:WU02:FS01:FahCore returned: BAD_WORK_UNIT (114 = 0x72)
04:03:53:WU00:FS01:0x21:ERROR:exception: Error downloading array interactionCount: clEnqueueReadBuffer (-5)
04:03:54:WARNING:WU00:FS01:FahCore returned: BAD_WORK_UNIT (114 = 0x72)
04:04:09:WU00:FS01:0x21:ERROR:exception: Error downloading array interactionCount: clEnqueueReadBuffer (-5)
04:04:09:WARNING:WU00:FS01:FahCore returned: BAD_WORK_UNIT (114 = 0x72)
******************************* Date: 2016-10-26 *******************************

[Breach]

This driver has been replaced by nvidia with 375.63 at http://www.nvidia.com/drivers .

If you do not have a specific reason to upgrade and your folding is working I recommend not upgrading at this time, at least until there's some assurance that nv has released a valid update. If you already did and folding is not working please reinstall the older drivers that worked.

It's broken too ... only a few small projects might be able to start, but most of core 21 projects won't ...

If you don't need to upgrade, avoid any driver of the 375.xx branch ...

Has anyone been able to get through to Nvidia??? I've a feeling every release will give a problem until someone works with Nvidia to give us a "friendly" release.
In the meantime, I've gone back to a driver before the 375.xx releases, this seems to solve the problem for now. But, I've lost an entire day of production. No telling how many people aren't even aware their GPU slot is freezing up!

Yes, there's an update from nVidia: https://forums.geforce.com/default/topi ... -/#5003286

Work units reporting wrong values in Folding@Home - Under investigation

[Hannibal13]

Hopefully this gets resolved soon. Maybe FAH can send out an email since this must effecting a large number of machines if they have been using Nvidia GPUs. I waited untill Monday to update the driver and I have not been able to fold since. The second thing that has messed up my folding performance in the past week.(This and the internet outage on Saturday.)

[Calico]

373.06 is latest stable for me. The weird part is I thought Nvidia works directly with FAH programmers to optimize their drivers for fahcores.

With 375.57 first not working, 375.63 a couple days later was a hot fix for an outlook email app or something....I'm sure a F@H fix will be implemented in the next one. But until then it's 373.06 for me especially with EVGA time zone challenge coming up in a few days

[7im]

373.06 is latest stable for me. The weird part is I thought Nvidia works directly with FAH programmers to optimize their drivers for fahcores.

Nvidia does not work directly with FAH like that. If they did, we wouldn't see occasional bugs in the drivers like this.

People submit reports to NV about bugs, and then NV fixes them, eventually. They publish updated drivers, we install them. That's about it.

[Calico]

New driver. Folding apparently still under investigation

https://forums.geforce.com/default/topi ... 10-28-16-/

[bruce]

New driver. Folding apparently still under investigation

https://forums.geforce.com/default/topi ... 10-28-16-/

That's not surprising. The stated goal is to make a driver that is "game ready" and they've probably have a backlog of game problems that have been in their queue longer that FAH has been. We can continue to make problem reports ... and be patient.

[Breach]

New driver. Folding apparently still under investigation.

https://forums.geforce.com/default/topi ... 10-28-16-/

"Still under investigation." So this new driver, 375.70, still has the malfunction? Has anyone tried it yet?

Yes. Yes. Don't update:

Code: Select all

19:37:40:WU00:FS01:0x21:ERROR:exception: Error downloading array interactionCount: clEnqueueReadBuffer (-5)

[bruce]

Thanks! I hope F@H will consider putting out an urgent e-mail to all of its members to not update to any 375.xx driver.

FAH does not maintain a broadcast list of Donor emails and has never sent one out. They do post urgent messages on the blog which they have done.

[bruce]

It probably already has impacted production, though not as much as you might think. The number of GPUs with the faulty drivers is somewhat limited and FAH is designed to recover from errors by quickly assigning the WU to somebody else.

Personally, once I find a driver version that's dependable, I don't upgrade spontaneously and I never choose a "game ready" update since they don't provide anything interesting to me.