GTX 970 FTW GPU Clock Issues, stuck at low frequencies ~100 - 500 MHz, horrible frame rate

gkraju747

Prominent
Jan 23, 2018
3
0
510
Hello,

My gtx970 is running with PSU 1100W. So I know there is more than enough juice for it. Its been working fine on all games for past couple of years but now all of a sudden, the frame rates have dropped across all games which were totally fine until now... For example, crysis 3 used to play at 50-60 FPS but now with same settings is between 10-20 FPS!. I have installed PrecisionX and have been monitoring the GPU clock, when its working fine, it used to be around 1000-1200MHz for the most part which would explain the normal FPS but I can clearly see the low FPS is directly associated with that clock throttling back down almost in a min of starting the game down to 200 MHz. I am not sure why is this happening with nothing changed on my system.

I have gone through lot of posts and tried lot of stuff and even had the Card replaced with no luck. I suspected it is PSU but again how can 1100W PSU be not enough all of a sudden. Restarted the PC, disabled and enabled the drivers, reinstalled diff versions of drivers, removed secondary drives to reduce the load on PSU.

I am completely lost on how to get this thing working again and not sure where and why its getting bogged down. It randomly works fine for few mins after a restart. Is this somekind of a known bug with the 970's or its drivers?.. Any help or suggestions pls?
 
Solution
Good to hear another psu fixes the problems.:) Make sure you let them know that. When (if,but i don't see why not) getting another psu from the rma would i probably prefer that in the system over the CX,but that's up to you.

gkraju747

Prominent
Jan 23, 2018
3
0
510
PSU - Corasir HX1050
Yes tried 10 diff old drivers already, same behavoiur
No option to check on another system, but already got this via RMA just 3 months back because the old one started to exhibit same issue!!

Actually when I got the replacement, it had same issue for few days but after removing some devices and doing chassis cleaning, it started work fine. Now its back again on this new one as well..

Its kind of obvious at 1st glance that its PSU or a buggy GPU model itself. I can rule out any windows or driver issue at this point.

Can a PSU be working just fine but not giving enough volts to the GPU?... What I observed via Precision is, when the GPU Load is 100% thats when the GPU Clock hits rock bottom. But when the GPU Load is around 50% ( games that are couple of yrs old ), then GPU clock is working fine in 900-1200Mhz range as normal. Its only when I playing the latest games, I am seeing this which were fine as well.. If the PSU is not able to give enough watts to GPU, it should not even power up right?

One more thing is, when I do complete shutdown and power off from PSU and then start the system, thats when the 1st latest game I load, starts working fine in full GPU clock ( as if the issue just resolved itself ) only for a couple of mins and then the GPU clock starts coming down to 135 Mhz and thats it. Seems like the GPU is not able to sustain the load for some reason.. either its not getting enough extra volts it needs for 100% load or going to some kind preserving mode...

Temp's are definitely not the issue I know.

What do you think?
 

Vic 40

Titan
Ambassador
Maybe that a log can tell more,

How to make a log,

download hwinfo32,
install and open it,
check "sensors-only",
click "run",
at the bottom of the window click "logging start",
choose a name and place for the log like "hwinfolog" at the "desktop",
this log can be viewed/opened with either excel from Microsoft or libre office calc.
you can upload it onto a site which facilitates this like SaberCatHost or DropBox,these i like since not so much (or any) spam comes with these.
Just make one until this happens like you described.


There's more info in these lately than i like,but voltages should appear which is what i wil look primarely at besides maybe cpu/gpu temps and such info. You can as described above take a look yourself of course. ;)
 

gkraju747

Prominent
Jan 23, 2018
3
0
510
The issue is now resolved. Turned out to be the PSU as suspected. I went and bought a CX750 and replaced it. Now everything is working fine!.. I could see the main difference was the GPU voltage was constantly fluctuating between 1v and 1.05v with the old PSU where as the card could take 1.2V under load .

Now after I replaced it and monitoring the same, the Voltage never drops from its designated 1.2V.

PSU was 4 yrs old, time to RMA.

So the GPU was not getting the voltage it required for smooth operation and whatever it was getting was also unreliable.
 

Vic 40

Titan
Ambassador
Good to hear another psu fixes the problems.:) Make sure you let them know that. When (if,but i don't see why not) getting another psu from the rma would i probably prefer that in the system over the CX,but that's up to you.
 
Solution