PC performance loss after NVIDIA driver update

[Bacono10]

General decrease in PC performance

So yesterday I updated my NVIDIA drivers from GeForce Experience

(driver version 355.60)

and ever since that my PC's performance has dropped considerably.

Things are generally slower, but it is most most noticeable in games.

before the driver update I would get solid 60+ FPS in Battlefield 4 at Ultra settings, but now the average is 30-40 FPS on medium settings. Same situation for all my other games.

I have tried uninstalling the driver and reinstalling it, and installing an older driver to no avail.

any suggestions are appreciated

PC SPECS:
Strix GTX 970

AMD FX-8350

8GB RAM

600 WATT PSU



Nothing has been overclocked.

Thanks in advance.

 

[damric]

CMOS resetting does not take ten minutes on any system known to me, which is the majority of them. I seriously doubt ANY system. Residual power is depleted within seconds, or at least no more than a minute. Generally, probably as soon as the battery connection through the CMOS batter is broken in reality.

Never in my life have I heard of letting the system "cool" before removing the cmos battery. Not in school, in practice or from any other tech, ever.

I've actually seen worse. Not even that long ago I had an ASUS 790X board that would literally take 3 days to drain out the capacitor for the BIOS IC.

 

[Darkbreeze]

Yeah... 10 minutes may be so much, but I've seen some systems keeping energy after arround 30 seconds without the battery and if you say 10 minutes you make sure the MOBO will get a full minute without the battery.
If a sensor doesn't detect the temperature changes well, it could be because it either keeps the last temperature state when powered down or because it takes the first temperature when powering on, both are terrible designs, i have seen them both. Altough the first case should anyway be fixed when removing the battery, one never knows.

That's total nonsense you're spouting. The system doesn't "keep" anything from the readings on any prior session. I don't even know what you're talking about. I don't think you do either. I don't like telling anybody not to post, but if you're going to post misinformation, don't bother to post at all.

If a sensor " doesn't detect the temperature changes well", it's because there's a faulty sensor on the motherboard or a problem with the bios tables, no other reason, ever.

 

[Darkbreeze]

CMOS resetting does not take ten minutes on any system known to me, which is the majority of them. I seriously doubt ANY system. Residual power is depleted within seconds, or at least no more than a minute. Generally, probably as soon as the battery connection through the CMOS batter is broken in reality.

Never in my life have I heard of letting the system "cool" before removing the cmos battery. Not in school, in practice or from any other tech, ever.

I've actually seen worse. Not even that long ago I had an ASUS 790X board that would literally take 3 days to drain out the capacitor for the BIOS IC.

This would be an extreme exception to the rule though I think. I've never personally seen. I guess it's not a bad idea to go ahead and pull the power cable from the wall, remove the cmos battery and then push and hold the power button for a minute or two, which I suppose would discharge the stored charge. I know it does on laptops, but I've never seen a cmos not reset after pulling the cmos battery for 30 seconds. I certainly can't argue that it's impossible though. Just that I never encountered it. Then again, there's a lot of things I'm sure I haven't encountered that have definitely been known to happen.

 

[Bacono10]

When I entered the BIOS after I clearing the CMOS, it said it had been reset and all of the settings were set to default, so I think it worked for me in this case.

 

[Darkbreeze]

So, any change on the cpu ratio/speeds?

 

[Darkbreeze]

Here's what you want to look at. This is from the very first HWinfo screen you posted. Notice the speeds in green, which are at 4100mhz. That would be the "normal" reading for the chip at it's full speed. The reading in red are what it was currently at when you took the screenshot. Green was the max reading it picked up. The screenshots you posted after that first one never showed the cpu going to it's max normal clock of 4ghz.

 

[Bacono10]

Here is a screenshot after letting prime95 run for longer, I saw the speed jump up for a second but then go back down.

 

[Darkbreeze]

I don't know. Clearly there is an issue. The CPU clock speed should stay at 4000-ish Ghz with the processor power management at 100% and Cool N Quiet off. Even with those at 5% minimum on power management and Cool N Quiet on, the CPU should at least be capable of running at it's top speed under a load and staying there until there is no demand. I'm going to go out on a limb and say it's the motherboard since the CPU temperature does not seem to indicate a thermal condition and we see the CPU IS capable of attaining it's max speed if only briefly. It could be the cpu though. You might want to pull the CPU again and check to see that there are no bent pins or pin sockets on the CPU and motherboard.

It could also be the CPU. Make sure as well that you have the 4 or 8 pin CPU power cable connected to the motherboard in addition to the 24 pin connector.

 

[Bacono10]

The 8 pin CPU as well as the 24 pin connector are both in.

I guess tomorrow I will check the CPU and mobo.

 

[Darkbreeze]

I think there's a good chance that perhaps the motherboard VRMs are causing the throttling. I've seen this on several 970 chipset boards where the CPU core and package temps were fine but the system was still throttling due to either faulty VRMs or VRMs that were getting hot. Motherboard temps don't look suspicious though, but that may not mean anything. How old are your motherboard and cpu?

 

[Bacono10]

They are both less than 3 months old and were installed at the same time.

 

[Darkbreeze]

Go into the OC screen in your bios, and post a screenshot of it like you did before.

 

[Bacono10]

 

[Darkbreeze]

Your CPU is running at 4Ghz when it's in the BIOS, which is as it SHOULD be. I think maybe this is a windows problem and I'd recommend doing a clean install.



Prior to installing as outlined by another of our members:


1) Check that the software and hardware you mostly use are comptible with Windows 10 and have drivers available


2) Make sure to update all necessary drivers and Windows updates


3) Some devices might need a specific driver to operate with Windows 10, it might be a good idea to download these Windows 10 compatible versions prior to upgrading to have them ready


4) Backup all and any important files from your computer to a cloud service or external harddrive


5) Create recovery media for your current Windows install, if you have a OEM pc/laptop refer to the user manual or check the website of the manufacturer for how to create the recovery media for your device (some people have been unable to use the rollback feature after upgrading to Windows 10)



How to do a clean install


Dowload the ISO here: https://www.microsoft.com/en-us/software-download/windows10


Use this tool to create USB or DVD installation media using the downloaded ISO: https://www.microsoft.com/en-us/download/windows-usb-dvd-download-tool


Windows 10 clean install tutorial:

http://www.tenforums.com/tutorials/1950-windows-10-clean-install.html

 

[Bacono10]

Someone else suggested to to set the voltage to 1.375 and see if that helps. Would that be worth a try?

 

[damric]

It's throttling for sure. Probably VRMs.

 

[Darkbreeze]

Increasing voltage only helps with system stability, and creates further heat. I seriously doubt voltage is your problem, especially since it was at 1.4v at one point if I recall, and the problem was still present. There's little doubt in my mind that you should need over 1.35v for it to operate at it's stock 4Ghz frequency, unless you received the worst FX-8350 chip ever sold, and probably not even then.

Since it is showing as CURRENTLY RUNNING at 4Ghz, actively, in the BIOS, but will not run at higher than 1400mhz for more than a second or two when in windows, I'm fairly certain that it's either a motherboard VRM or Windows issue. I would highly suggest trying a clean install first, followed by a RMA of the motherboard if that fails to clear up the issue.

 

[quangtran]

i just dont get it, your CPU temp and coreclock seem normal, max load still 4Ghz. GPU is OK. I suggest recording all the numbers in realtime again while you are playing games, then double check the log file

 

[Darkbreeze]

Only in the BIOS do those numbers seem right. In windows, a much difference scenario is painted.

Download and run Core Temp, and run Prime again, take screenshots of the Core temp window while Prime is running on small FFT so we can verify that it's not just a misreporting problem in HWinfo. I doubt it is, but lets be sure.

 

[Bacono10]

At first it said in Core Temp and HWinfo that it was at 4Ghz so I started up a game and tested it, my performance was back to normal for maybe 3 minutes and then started to decrease. When I alt tabbed to check the monitors, it was back to 1.4 again.

That's when I took the screenshot.


I haven't had time to do a clean windows install yet by the way.

 

[Britton]

Hello,

Don't mean to interrupt this thread or confuse the main topic but I to have seen this same issue with my Windows 10 install. It's why I'm here reading.

In my case though the problem exists on a mobile Nvidia product, the Quadro FX380m. Windows 10 apparently does not work right with any of the drivers from Nvidia or from Windows update. The issue appears as locked core(s), shader, and memory speeds. My chip is no where near as advanced as the one in the screen shots but the problem is the same. I have tried all the drivers, rolled back, used the basic one, and clean driver installs. The only thing I haven't tried yet is a fresh install of Windows 10, which still may not fix this issue. I keep the laptop in High Performance mode all the time and changing the settings to prefer maximum performance has no effect.

Currently the only work around for this problem is to simply reset the machine, not power off and on again. That's what does it, the powering off and back on. In other words a cold boot locked the speeds at about half of there normal values. This "fix" may not work on a desktop model. I noticed the issue on my desktop as well, running an Nvidia card, but that seems to have fixed itself after updating drivers to version 355.60. I also feel as though this unusual behavior is directly related to Windows 10 itself. After resetting the laptop speeds return to normal and the chip throttles up and down as usual. Notably if i pull the power adapter from the back and put it right back in the chip will throttle back down and remain locked at half speed until I reset (not power off) the system.

Side note on my desktop though, it runs the Nvidia GTX460 and installing the current driver directly from Windows update stopped that problem and now the card drives both monitors at full speed as expected. I'm hoping Nvidia can do the same for my laptop, or MS fixes it. If after a few more weeks this problem still persist I'll be forced to try a fresh install, and if that won't fix it go back to Windows 7. I guess this scenario is about the same as old XP systems trying to update to Windows Vista/7. Some hardware was left unsupported to the point of becoming obsolete. Which sucks for those who can not afford to go buy the latest and greatest.

 

[ERIC J]

Thank you for your reply but I followed the instructions exactly and it did not help the problem. I'm still getting horrible performance.

I would be happy to test any other ideas you have.

There has been a problem lately with Nvidia drivers, my suggestion is to "roll back" to a previous driver that was giving you better performance.
Thats what i did when my GTX 980 classified was getting a lower score in bench tests and it fixed the issue.
I am using 353.62 driver and it is 100% stable with no performance drops.

 

[Darkbreeze]

I do not recommend doing a reset, refresh or restore on any system that was "upgraded", as I've already seen too many instances of it causing an unresolvable issue with booting the OS or total system corruption.

In any case, a "reset" can only be done if there has been a system image recorded or saved. Or if there is still an OEM system partition but that would require reverting to your previous OS. I recommend that you back everything up and do a clean install as I already outlined. Alternatively, you can download and run one of the portable Linux versions to see if the CPU behavior is duplicated in Linux. This would not affect your current configuration and could be run from USB. If the CPU works fine, with all cores clocking the way they should be, you know it's a windows issue. If they don't, you know it's hardware.

 

[ZelmoQuad]

Thank you for your reply but I followed the instructions exactly and it did not help the problem. I'm still getting horrible performance.

I would be happy to test any other ideas you have.

Roll it back and wait for the next future driver.

 

[Britton]

In case it helps I was able to fix my gpu clock speeds by turning off the "Turn on fast startup" option from within Control Panel>Hardware and Sound>Power Options. On the left hand side select "Choose what the power buttons do" and the option is in there. You may need to click the blue security link at the top to edit the grayed out options at the very bottom. I unchecked them all and have not had anymore cold boot issues with the mobile gpu staying throttled back in a lower power state.

Sadly I did a fresh install to find out if that was the problem. So in a last ditch effort before downgrading I happened to stumble upon the options above. More than likely my reinstall wasn't even necessary, just turning off that one option seems to have resolved my slow 3D/Video performance issues and restored the gpu clocks to full speed. Even after a powered off boot up. Good luck.