Question Help with underclocking GPU that is probably failing

[pmacarthur]

Hi there,

I have been troubleshooting an issue with my system for several months where it instantly restarts randomly when playing more demanding games - nothing out of the ordinary though. I can play LOL and minecraft no problem, but if i load something like stelaris or PUBG up, I will eventually see the restart issue. Long story short: started eliminating a bunch of components, reinstalled windows, and did a whole bunch of troubleshooting with temps that kept leading me to dead ends. I even RMA'ed my PSU (corsair AX760 platnum) and the issue persisted.

I eventually got to the graphics card and happened to have an old gtx 580 that i could swap in to test. I noted that the system didn't shut down with the 580 in (albeit ran a lot slower). I spoke with EVGA they basically told me to pound sand because my warranty expired 6 months before I "officially" reported the issue to them.

I decided to try and figure out what specifically is wrong with the card because frankly; i don't want to spend the better part of a grand on a new one right now. I installed EVGA precision. and started looking at temps. Now, I'm no expert with overclocking. I could probably follow some guides and find some success, but i don't really understand a lot of the basic concepts around overclocking GPU's. Because of this I have always ran my cards with stock settings.

I have no baseline for how my card behaved originally so I have been relying on the spec sheet for the card to tell me what the clock speed should be. When i started monitoring the card with precision xoc i noticed that the GPU clock has wild fluctuations in its clock speed. i will see the card sit at like 800ish MHZ at system idle or when playing LOL. However, when i play more demanding games i see it shoot up to around 1300mhz. this seems to be outside what the spec sheet says for the card, so im wondering if there is something out of whack. I am thinking i could lower the clock speed on the card a bit and stop the restarts from happening.

I don't think the shutdowns are being caused by a thermal issue. I tried cranking the fans up to 100 and just had the issue happen while the card was close to 60c. I understand i can turn on Kboost to force the card to the max clock. i have done that and will now test to see if i can get a reboot to happen while playing LOL.

If anyone has any input on this I would appreciate it. I dont really know how to decrease my clock speed.

EVGA GTX 980ti SC
1328MHZ GPU clock
3505MHZ mem clock
1.162 voltage
36c

 

[anotherdrew]

So, if this is your card, then I am surprised you are getting a GPU clock of 1328. That seems to be beyond the specs. That could easily cause a crash without overheating the GPU.

The GPU clock is normally controlled through firmware on the card (sometimes called VGA BIOS). If you have ever messed with this, there is the potential you have loaded an incompatible one. Secondarily, the clock can be controlled by software (like the Precision software you installed).

I think it is more likely a software issue, so I would first look for and remove any overclocking software. You are getting drivers from EVGA or nVidia directly? If not, I would download and install the newest drivers. Lastly, you can check your card specs and VGA BIOS version with something like GPU-Z. Here is a list of BIOS ... your version should match the one for your card. If something seems wrong, proceed with caution. Loading the wrong BIOS can brick your card.

 

[knickle]

^^^THIS^^^

1328Mhz is a 20% overclock from factory, in which case you don't need an underclock. You need to return to the factory settings.

Are you sure you quoted the correct numbers?

 

[anotherdrew]

Stock clock speeds for the 980 Ti is 1000/1075 ... the SC+ is clocked at 1102/1190 (a 10% overclock from factory) ... unless I am reading something wrong.

 

[pmacarthur]

I will initially get the drivers from EVGA, but then i install geforce experience and will get new releases from nvidia direct. I have the most current driver installed, and have just loaded GPUZ. I also uninstalled precision, but the clock is still over the stock OC. How would i go about lowering the clock speed back to factory if precision has the offset at 0 already?

I haven't done anything to the firmware on the card. i do seem to have a version listed on the site you linked. is this something i should try flashing on the card to correct the issue? are these bios updates typically something that gets published by evga/nvidia?

Screens:
GPU1 & 2 are the settings before i uninstalled percision. GPU 3 & 4 is after i uninstalled precision

 

[hang-the-9]

Click the Default button, that should bring system to stock speeds.

 

[pmacarthur]

That's the thing... its at the default settings in precision. the clock still goes past the listed specs. I know pascal cards can do this, but i didn't think the 980 could. What would be the best way to underclock it? I can change the clock offset slider to negative, but that only goes to -90 mhz which still isn't within spec.

 

[knickle]

Is it possible that you have a second overclocking utility that you forgot about that might be overriding the xvga precision settings? I would start looking at installed programs and see if there are tools that you forgot about. MSI afterburner for example.

 

[pmacarthur]

Unfortunately no, I just clean installed windows to eliminate anything like that. the only thing that was installed was precision. Its also important to note that i have never overclocked this card. the only thing i used precision for was fan speed.

 

[knickle]

In the short term I would try lowering your temperature target to something around 70ºC, and hit apply. That might help limit the clock speed and make it more stable. I would then hop on EVGAs support forum and post your issue. There's probably more than one expert there that can help you with this. https://forums.evga.com/

 

[pmacarthur]

Thanks. should i turn off kboost, or leave that on?

 

[knickle]

I assumed you turned on K-Boost in order to show what the frequencies are when you were taking screen shots from windows. As far as I know, K-Boost will run the card at MAX 24/7, wasting more energy.

 

[pmacarthur]

will kboost keep the clock up at 1300 if I lower the temp target? I'm not really concerned with energy consumption.

 

[knickle]

will kboost keep the clock up at 1300 if I lower the temp target? I'm not really concerned with energy consumption.

Sorry, I don't have the answer. My suggestion is to experiment.