nvidia 375 Drivers

J

jfb9301

[H]ard DCOTM February 2016
Joined
Jan 11, 2005
Messages
1,652
Are there any versioni 375 drivers that work with folding? I've been holding out with 372 series drivers, but my son just got Battlefield 1 for PC, and he cannot launch the game due to an error that 375.62 drivers are required to play the game. This is not quite the big issue with points as his machine only has a GTX960 in it right now, he and I had to RMA the SEAHAWK that was in the machine. When that SEAHAWK comes back, it will be much more important to folding progress.
 
Last I heard was at 375.95 was still bad for folding. Some people have stopped folding on their cards to play games.
 
373 is the newest driver I've found to work with folding on the new cards. Not sure if it's new enough for BF1
 
I'm seeing Core_21 is 0% stable on the latest 375.95 driver (which supposedly includes a hotfix for a memory bug). Core_18 seems to complete units correctly. I'm hoping 373 fixes it. I'll give it a shot.
 
I'm running 373.xx on my 1080 without issues except that I realized I overloaded the breaker sometime this morning and over half my D went offline until I woke up.
 
Nothing else? That suxors.... maybe they will get this sorted out, but I'm not holding my breath. Of course Stanford will insist that it is a nVidia issue, and nVidia probably insists that the issue is on Stanford's end. Please keep me posted if anything else changes. I have to get back to studying.
 
Just got a 1080- I can verify the latest 376.09 is a no go. I'm trying 373.06, I heard it works...

Oh-
From Vijay's blog:
Core 21 (OpenMM GPU) issues with new Windows NVIDIA drivers 375.57 (released 10/20) and 375.63 (released 10/23)

We have received a number of reports that the recent release of Windows NVIDIA drivers 375.57 (released 10/20) and 375.63 (released 10/23) is causing many core 21 (OpenMM) projects to fail. We don’t yet know the cause of this problem, but are actively investigating it.

In the meantime, we strongly encourage you to stick with or downgrade to the last known working Windows NVIDIA driver (372.70) if you would like to continue folding core 21 (OpenMM) GPU projects.

If you have already upgraded to 375.57 or 375.63 and are experiencing issues, we encourage you to report your issues to the NVIDIA Driver Team:
 
I hear they have done nothing to fix it in 376.16, heck it's not even on the list of issues in the driver documentation. People at folding forums, and nvidia forums are getting a bit terse with nvidia.
 
Yeah this is frustrating. This bug has existed for multiple driver revisions for months.

Are they lazy or is this a sign of a bigger problem?
 
I tried 376.33 for the last few days on two GTX 1050s and it's been "stable" from what I've seen.
They would not fold on the last few drivers.
 
When I get home from xmas vacation, I'll have to look into this. Early reports were that .33 did not fold, but there was to be a hot fix to solve the issue. Looking good, thanks for the heads up.
 
It appears I am only partially right about 376.33.
The 1050s appear to drop to a low clock (~680MHz instead of boost of 1750MHz)
when WUs of a certain core start and no way to get them to boost again until that WU finishes
and I power off/on the PC with another more stable WU running. Then boost clocks return.
I thought it was just a one-time event but it happens regularly when the sequence above is repeated.
 
Weird- I'm new to nVidia, what do I look for in boost clocks tonight?
The new driver did fold, but I didn't see a finished unit (but didn't check before bed).
 
I would expect 1080s to be well above 1700MHz when in boost mode.
It will depend on your model of course.
The "crippled" mode will be way below that. Most likely near the same 680MHz I am seeing.
 
Core32 has it right.
Check for those frequencies with a program like nvidia inspector or GPUz. I'm sure you could see it in afterburner or others, but I use simpler utilities. I don't recall off the top of my head what a 1000 series card defaults to on error, but a 900 series defaults to 405 MHz, and will not go higher than that for any reason until you reboot the boxen.
 
Ah- I just pulled up the EVGA overclocker, it looked good (2100Mhz), but I'll grab nVidia inspector tonight and double check.
 
Whoops- I can game fine at that Mhz, but F@H is failing. Trying again at stock speed...

Ok, at near stock speed I can fold fine on a 1080 with these drivers. Bonus!
 
Last edited:
I loaded the 376.48 driver Skilllz mentions onto the 1050s late last night and no frequency drops have occurred with either since.
More than one project type has been completed as well.
Hopefully this allows the 10xx series boards to fold solidly.

Update: Still completely stable since last Friday nights install.
 
Last edited:
Loaded the new 378.49 driver last night and both 1050s ran over night just fine folding.
 
Did PG fix it on their end or is this fix from Nvidia?
 
No idea. I simply updated the Nvidia driver when it notified me it was available.
I've not seen any type of notice from PG of a fix, but they aren't known for giving a crap about the donors so take that for what it's worth.
 
At this stage its mainly a NV fix, PG have an updated core in testing but it may not be out for a couple of weeks yet
 
If I wasn't about to be going out of town again I'd load them up, but when my boxen goes down again they won't be restarted until March so I can't take that chance.
 
I'll second that the new drivers fold fine- no tweak needed on my end.
 
The updated core should be out on Monday, PPD may be a bit worse - not sure if this is down to the driver or the new core.

Core 22 will be entering testing soon, promising to be quicker
 
Will that require a new F@H install package, or is that shipped in the work units?
 
Gosh I hope not. I don't want to update all my boxen. That'll take forever.
 
DWolvin said:
Will that require a new F@H install package, or is that shipped in the work units?
Click to expand...

Not sure, a core update or new core does not usually require a new version of the client - however a newer version of the client is also being tested so they may take the opportunity to force an update.

I would say the core 21 update will not require an update to the client but core 22 might.
 
DWolvin said:
Will that require a new F@H install package, or is that shipped in the work units?
Click to expand...

the Core is not shipped within the work units; though a work unit might require a certain version of a core and will trigger the required download of it from the Stanford servers. Actually I got recently a 0x21 v0.0.18 downloaded from some beta units.
 
Nathan_P said:
the new version of core 21 has been released to full F@H- More info here

https://foldingforum.org/viewtopic.php?f=24&t=29633
Click to expand...
Personally not going to try a post-373.06 drive until they quit talking hotfix. Did my 1080 fold with the "HotFix" - Yupe it did, but points/day were down substantially and then started having hard Win10 crashes, and these crashes appeared each time associated Folding unit crash.

Backed up to 373.06 --> No more Win10 crashes and not having a point reduction anymore. Folding without any pauses, except occ a unit download delay on my old MSI 780 Lightening

Gary Smith
 
Yeah the core fix and the hotfix driver don't always play nice together. Hopefully NV remove the hotfix soon.
 
You must log in or register to reply here.
Back
Top