Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by ManuelG, Jun 6, 2015.
right now he's probably hiding under a table somewhere
because he's probably getting bombed on
Yeah looks like it, guess they changed some refresh P-state polling rate inside r352/r353 branch and its not so compatible
Or something with nvapi in general if some get it no matter what, like back with r280.xx - 290.xx drivers and GF104/114 TDR (web browsing, flash) after specific time period, while some GF110 chips glitched by 3d switching and stayed longer in 3d P-state..
275.xx or lower was fine and nvidia later said they tweaked P-state polling rate for gpu voltage/frequency to react faster or something in those lines and its going to continue with this new approach 280.xx+.
Here it glitched only light load Firefox hw on (64bit variant) and only by 2 specific cases (forums, and video flash content) and specific time period ~ 2-3days with normal power off at night.
Although Im fine now with 353.12, day 7, minimal monitoring sw usage, only for gaming and its set to 4000ms.
And I updated flash to v18.xx beta so maybe its something with that too, possible conflict with official flash v17.xx and newer driver..
My two cents. I'm running the hotfix 353.12 with MSI AB running with 'typical' settings. Disabling H/W acceleration in Firefox has eliminated crashes. With 352.86 I had black screen crashes just sitting on the desktop with no apps active whether MSI AB was running or not. 353.06 was a bit more stable but still had desktop crashes.
Zero driver timeouts with Opera after stopping Afterburner. Something in the recent drivers doesn't like being polled I guess.
Going by the majority it does look to be a problem with drivers and polling, this isn't just related to MSI Afterburner as even EVGA Precision does it. Perhaps Unwinder and Nvidia can look into it rather than shrugging their drivers/software off as perfection
Been running Afterburner since forever, never caused me any crashes
Installed 353.06 - PC crashes regularly
Installed 347.88 - PC stopped crashing
Installed 353.12 - PC has crashed once
It's not Afterburner, it's the drivers
I am using the v353.06 drivers along with MSI Afterburner v4.1.1, GeForce Experience and PhysX (latest versions which came with the driver). I also use Chrome v43.0 as my main browser and have Hardware Acceleration enabled. However, I do not usually leave the browser open while gaming.
I seem to have fixed the initial TDR problem I was experiencing with Project CARS and Far Cry 4 on my EVGA GTX 980 SC by doing a completely clean install of the driver using DDU v15.3. Despite the TDRs though, all the other games I tested seemed to run fine, including The Witcher 3 which I played for 7 hours. I have NEVER experienced a TDR while just browsing, they have only occurred while gaming.
Now, all my games seem to run without the TDR issue, though I have had a couple of them unexpectedly exit to the desktop without an error or Even log entry (e.g. The Witcher 3, albeit it has only happened once in over 60 hours of playing the game and I have had no other crashes).
I always uncheck the two 3D and Miracast options when I install the driver. I also disabled the NVIDIA HDMI audio devices as well since I do not use them. Finally, I disabled the pointless (for me) NVIDIA Streamer service since I do not use ShadowPlay nor do I have a SHIELD device. I played Project CARS for over an hour when prior to the clean reinstall of the graphics driver it would always within 5 minutes of starting a race. I've ran numerous benchmarks and played lots more games since and they all seem to run fine. I have not seen a TDR error in fact in three days since I did the driver reinstall.
When I installed 352.86 I had one "Display Driver stopped responding" error but since then it's been working without any problems. I've do run Chrome with HW acceleration enabled and I do use MSI AB (when I play Witcher 3).
After that I've both installed the 353.06 and .12 Hotfix and still no crash.
I totally agree, ive ran with and without AB and any driver above 350.12 i get freeze in chrome and driver stopped responding. So yeah its the drivers on my machine anyhow and tbh i very rarely get any problems with drivers. Nvidia is at fault i think.
There are some TDR timeouts that can be changed, I guess most of you guys tried that already?
Windows Registry Editor Version 5.00
(save the above as a .reg file and double-click it; reboot)
Windows Registry Editor Version 5.00
One time this kept happening with me (TDR due to hardware acceleration in browsers), and it turned out my power supply was failing, and that was the only sign of it going bad. Replaced the power supply and it stopped.
My TDR issues got out of hand around the release of the 344.xx drivers.For me they were caused by temp throttling.Got rid of that and my cards' felt brand new ever since.One bluescreen when I installed 350.12 on release but none since and I've tried every driver afterwards with great results.Currently running 353.12 hotfix.
Just got my first TDR in Firefox after installing these drivers yesterday Back to 347.88 I go I guess. And no I am not going to turn off hardware acceleration in Firefox, I didn't pay $530 for this card to have to turn stuff off. What is happening to nVidia, they used to be known for the quality of their drivers.
damn this drivers are greatly unstable, i was browsing perfectly in firefox (this one never crased before sice 353.06 release), i installed new flash 18 tested it and boom, screen goes black and "nvidia kernel module" or whathever error shows, supposely my crashes were only on chrome but now even firefox crashes, and i crashed twice, i was browsing on youtube how crysis 1 run on a gtx 980, and the second while writing this.
Two days now since upping the polling rate to 5000/6000 in MSI Afterburner has met with complete success, not just that but there are a large amount on people on OCUK also reporting this success http://forums.overclockers.co.uk/showthread.php?t=18675135
So Unwinder, Nvidia and those that have developed other monitoring software does need to look into this.
Drivers crashing on my 780 GTX SLI and crashing on my new 980 TI SLI
Nobody besides NV should look into driver related issues. Neither me in AB, nor w1zzard in GPU-Z, ASUS in GPU tweak, Gigabyte in OC Guru or EVGA in Precision (and each of tools is reported as the "route" of crashing) use anything but NVIDIA's own official NVAPI to display graphics card state in realtime. And NVAPI is build special to provide graphics card state reporting to vendor specific software.
The only thing you achieve by increasing polling rate in AB and similar tools from 1000ms to 6000ms is reducing the amount of access to NVIDIA driver via NVAPI with 6x factor. The amount of NVAPI calls during 10 minutes of testing with 1000ms polling period is equal to 1 hour of testing with 6000ms polling. If there is a problem randomly occurring in driver once per let's say 10000 NVAPI calls and you needed roughly 3 hours to reproduce it with 1000ms polling rate, then you'll need roughly 18 hours with 6000ms period to have comparable chances to reproduce it. It is not a fix by any means, it is a placebo that makes the problem harder to reproduce. If those who follow the guide of "fixing" it via altering polling rate don't get it - that's not my problem.
And that thread in OCUK is perfect example why I was unhappy, that's what is happening when data is being interpreted by not to knowledgeable users.
We just want less TDRs, man. I don't care who is responsible but if setting the polling rate higher makes my end user experience better then that's what I'll do.
Yes, we are aware the problem is on Nvidia's end. THIS VERY THREAD IS NVIDIA ASKING US TO HELP THEM FIX THEIR ****. We know that upgrading the latest drivers is what prompted this. That doesn't mean that we can't ****ing put 2 and 2 together to help Nvidia track down what's causing the issue in their driver.
Maybe we should try using smoke signals so no one's offended by the mention of any particular piece of software.
Incidentally, polling rate to 6000 and a clean install of the hotfix drivers hasn't produced a TDR in over 24 hrs. There is likely still a conflict here but at least it's tolerable.
And I just want to focus on more important things and development instead of being forced to comment fake statements like "AB is the root of problem" or "So Unwinder does need to look into this.." a few times per day. Especially when I'm mentioned as the person responsible for fixing it.
I don't think that you have a chance to understand me until you create something yourself, share it with a few million community then start feeling on your skin what does it mean to be bombarded by users who just want to play without TDRs and don't care about anything else.
Credit where credit is due, Unwinder is replying, ManualG has run off