I recently replaced my EVGA GTX 750 Ti with an EVGA GTX 970 FTW. This is my first experience with a factory-overclocked card. I've been trying to use the debug option in the NVIDIA Control Panel to temporarily downclock to reference stock speeds while troubleshooting some TDR errors in some games. Here's what I'm expecting (and what used to happen):
 
- When I first got the card and tried this, I would check the system information and note the graphics clock speed being at 1215 MHz under details.
- Turn on debug mode.
- Now the graphics clock speed should drop to 1050 MHz under system information.
- Using Precision X 16 should also show the white arrow move down to around 1050 MHz and a -165 MHz offset for the GPU clock (and 1215 MHz and 0 MHz offset when at factory overclock).
 
Here's what is happening now:
 
- Check system information and see 1215 MHz under details.
- Turn on debug mode.
- Still see 1215 MHz under details (even after closing and re-opening NVCP).
- Check Precision X 16 and I see the -165 MHz offset, but it still shows the speeds being at factory overclock.
- When I try to put load on the GPU, it still appears to hit the higher speeds in HWMonitor (need to do more elaborate testing to confirm though).
 
I work in IT, so one of the first questions I usually ask a user is "well, what has changed?" Unfortunately, a few things have changed here:
 
- I was a moron (albeit an excited one) when I first got the card and booted up with the old drivers from the 750 Ti installed still. Not a clean setup by any means, but things did work. Anyways, between the time I got the expected behavior and what is happening now, I did an uninstall of the NVIDIA drivers, wiped things with DDU, and then did a clean install of the latest drivers. So, technically, that's two changes: clean driver install and an upgrade to the next version which is the current stable release from last week, 368.81.
- Getting the new card led to me upgrading my monitor to an Acer XB271HU. The key highlights here are changing from a 1920x1080 at 60 Hz monitor to a 2560x1440 at 144 Hz monitor with G-Sync.
 
Things I've tried since then:
 
- Uninstalling my old copies of Precision X 16 (along with deleting old profile files) and OC Scanner from last year. Rebooting. Installing new versions that are current as of today.
- I read in a few places that the 144 Hz refresh rate combined with 1440p can cause issues with clock speeds? Downed it to 120 Hz just in case.
- Tried doing a clean boot of Windows with most startup programs disabled, since some programs like Razer Synapse peg the GPU just by running in the background.
 
Other things I've noticed:
 
- Last night, I found a bunch of games were performing horribly as I was going through and tweaking settings for the new monitor. Finally popped HWMonitor open and found that the GPU clock was stuck at 405 MHz. Rebooting fixed the problem, but it was pretty strange.
- I seem to get a lot of TDR errors when games switch back and forth between windowed/fullscreen modes and when going back to the Windows desktop. Never have problems in the games themselves. Games that run in 4K with DSR seem to be more frequent offenders. I've considered the power supply being at fault, but why would the system run GPU stress tests stable for an hour without any errors if there are issues with unstable power on the 12V rail? Also, I never had these errors with the 750 Ti, but I realize this is a more demanding card and the games are now using higher settings.
- I'd also like to note that I'm using the 2nd BIOS via the toggle switch on the back of the card. I wasn't wanting to do much with overclocking, but I wanted the more aggressive fan curve that I read was included in that BIOS.
 
Aside from more nuclear options like re-building the OS from scratch and/or RMAing the card, any suggestions for things I can try? I guess my next step could be another clean driver removal followed by an install of the previous driver I was using? This isn't really a dealbreaker at this point. I really just want to understand why the GPU isn't responding to requests to adjust the target clock speeds (and with NVIDIA's own debug function no less).
 
Thanks for any assistance I can get! I humbly defer to this community's expertise for help with this problem. (Also, sorry for the huge wall of text. Sheesh, can I ramble or what?) :)
 
Travis

So it looks like the current Precision X 16 actually puts a load on the GPU just by being open (didn't notice that in old version). I tried playing with lowering the GPU clock speeds by various increments. It seems like not only does the clock speed set not update, but the scaling of the effect of the change is skewed as well. For example, I set the offset to -317 MHz, but the current speed drops by 115 MHz to 1100 MHz (from 1215 MHz). Again, the clock speed reported by NVCP and Precision X is still 1215 MHz, despite the change to 1100 MHz. This explains why the drop of 165 MHz set by debug mode doesn't even register a change in performance. This answer only leads to more questions...

I'll give it a try (it has been brought up very frequently in my research online)! Any problems with leaving Precision X installed as long as it isn't running?

MSI Afterburner shows the same behavior as Precision X did (having to drop 317 MHz to see an actual drop of 115 MHz in GPU clock speed). The whole thing is bizarre. The best part? When I run something like Furmark to push into boost speeds, I swear it actually went down in clock speed instead of up. This almost makes sense if you consider that the GPU is actually supposed to be offset by -317 MHz. If boost is an increase of 125-150 MHz, it could possibly see a boost up to a -165 MHz offset, which is still a decrease from the actual -115 MHz drop I was getting for non-boost speeds. In other words, changing the regular speed is skewed, but boost is still technically going by the right offset numbers. This is all theoretical of course. I really have no idea what is going on. Moving forward with driver cleanup/rollback.

Time to send it home.

No problem. Hope the windows re-install resolves the issue for you.

twalls
Wow, so even the Windows re-install had no effect. Going to try swapping out the FTW for the SC again for another fresh install of Windows. My roommate has a 750 Ti FTW. Want to see how the debug mode behaves on his to do a sanity check at this point.
 
On a side note, another strange issue has cropped up. I've seen this on the FTW card, but not the SC one (but that just means it hasn't happened yet). I sometimes get a line of pixels from the middle of the screen that end up on the left or right side of the screen. I've seen it in games, while at the desktop, and even during booting. It tends to go away when switching between one or the other (appears in game but goes away when quitting, appears on boot screen but goes away once Windows loads, etc.) So far, turning off the monitor and turning it back on makes it go away each time. Is this likely/definitely a monitor issue, cable issue, or an issue with the DisplayPort port on the card? Trying to rule the video card out first. I can now say it doesn't appear likely that it is software-related.

Sounds like a display cable problem.

Sorry, link deleted, just try this and reboot.  (just the option 1)
w w . tenforums.c o m /tutorials/4189-fast-startup-turn-off-windows-10-a.html
Remeove the spaces