Question Making my GTX 970 survive through underclock

[Munizx]

I've got an old EVGA GTX 970 that is dying. It started with crashes in high load, now it crashes just after the system starts. Made many tests and now I'm convinced it's a hw problem. I can't buy a new graphics card right now, so I'm trying to make the most of this one.

I've been using EVGA Precision to underclock. Initially, I would only use -100MHz and would be ok, but now I have to use -750MHz to be able to work. Once I do underclock, I have no problems and can get things done.

The EVGA Precision doesn't save my underclock, so I have to set it manually every time (am I missing sth here?). But sometimes it's not fast enough, and takes too long to load. So I started using MSI Afterburner, that saves my underclock and loads much faster. The problem is, the Afterburner only goes as far as -502MHz...

So...

1. Is there a way to make the EVGA precision to load faster AND save the underclock?
2. Is there a way to underclock to -750MHz in the Afterburner? (preferred)

Thanks in advance.

 

[Eximo]

Have you taken the card apart and put it back together with new thermal compound, that would be the first thing I would do. Stability could be bad due to a dry spot on the GPU itself. Enough heat cycles can harden the thermal compound until it flakes off. GPU temp sensor might not catch a small part of the die getting too warm. Underclocking would solve something like that.

Same for a memory chip, but that doesn't sound like the problem.

You might try re-flashing the vBIOS on the card. Somewhat risky, but it might bring it back.

But also, you could make a custom vBIOS with your stable settings and have it boot that way naturally.

Maxwell II BIOS Tweaker v1.36 Download

Maxwell BIOS Editor lets you modify the VBIOS of your NVIDIA Maxwell-based GeForce graphics card from the GeForce GTX 900 Series, like GTX 980 Ti, GTX

www.techpowerup.com

Do ALOT of reading before making any attempts with the BIOS.

 

[hotaru.hino]

I take it you're using the slider. You should instead use the V-F curve editor. In MSI Afterburner, press "CTRL + F" and it'll bring up another window. From there, set ALL of the points to the same value as the first one. You can use a combination of the arrow keys, tab, shift, and CTRL to manipulate the points if you don't want to do it via mouse.

This tells the video card to not even bother going higher with clock speed beyond the lowest the VF curve allows.

As an example, I have my video card set to this:

This keeps the video card from clocking higher than 1605MHz (I use this profile whenever I don't really need performance)

 

[Munizx]

Have you taken the card apart and put it back together with new thermal compound, that would be the first thing I would do. Stability could be bad due to a dry spot on the GPU itself. Enough heat cycles can harden the thermal compound until it flakes off. GPU temp sensor might not catch a small part of the die getting too warm. Underclocking would solve something like that.

Hi Eximo,
Thanks, I'll try the thermal compound, it might be the problem (hope so!). Going to leave the vBIOS option as a last resort, as I'm not really secure how to do it.

I take it you're using the slider. You should instead use the V-F curve editor. In MSI Afterburner, press "CTRL + F" and it'll bring up another window.

Hi Hotatu,
I liked this option! But when I tried the control+f, nothing happened. Read somewhere that the board must be compatible (I guess mine isn't) :/

 

[Munizx]

Have you taken the card apart and put it back together with new thermal compound, that would be the first thing I would do. Stability could be bad due to a dry spot on the GPU itself. Enough heat cycles can harden the thermal compound until it flakes off. GPU temp sensor might not catch a small part of the die getting too warm. Underclocking would solve something like that.

Took the card apart, and the thermal compound wasn't on the best shape. Removed, reapplied, reboot, with all the care in the world, but to no avail. Tried twice, but both times the underclock of -750Mhz is still needed to keep the PC running... It was a good try, I had high hopes =/
Maybe running so long with an old thermal paste damaged the chip somehow?

Now I'm starting the PC with MSI Afterburner with -502, then I close it, open EVGA Precision to -750, and then the PC is ready and mantains stable for the rest of the day...

 

[Eximo]

All that is left is to try a vBIOS flash and see if that brings it back.

If you want to avoid software tools, then the vBIOS tweaking tools I listed before is an option. But trying a stock vBIOS flash first might work, you never know.

Just find your exact card:

TechPowerUp

Extensive repository of graphics card BIOS image files. Our database covers submissions categorized by GPU vendor, type, and board partner variant.

www.techpowerup.com

TechPowerUp

www.techpowerup.com