solved! GTX 1070ti (ASUS ROG STRIX) stuck at 139MHz

[RebelKhaotic]

Totally 100% actually SOLVED this time.

 

TL;DR:

If your card will not boost, always shows perfcap of power, and MAYBE boosts when slammed with a massive challenge. It's probably a power delivery issue, and if it's not your PSU or cable, odds are it will take some soldering that if you are not VERY comfortable with, I wouldn't recommend.

/TL;DR

 

I've seen this same issue across the internet with no solution, now it appears we have one. 

 

So, after a lot of head scratching and moving things around, I noticed that there always seemed to be a pwr perfcap reason with this card. For a time I could reliably get it to boost the clock if hit with enough load. I already had a 2070 super on the way and intended to give this card to my wife. 

When I moved it to her machine, things seemed to go back to the worse state where I could almost never get this thing to operate properly. 

In a last ditch effort to save the card (and my wallet) I decided to take the whole thing apart again. I replaced all of the thermal pads with Arctic MX-4 compound, cleaned everything AGAIN. And resoldered the power connections and a few of the components that looked a little "less than brand new" all on the power delivery side. 

Now the card operates properly, the power delivery side of the card must have been somewhat degraded, and is now basically new.

 

OLD POSTS AND UPDATES FOLLOW:

 

 

Not solved:

 

Appears I jumped the gun a bit. The card functions properly now but it appears only in synthetic benchmarks and some games...
As it turns out, if you don't slam the GPU with quite a lot, it won't boost at all, even up to base clock... Older games or lower settings will keep it running at 139MHz and sometimes you'll get abysmal FPS. Other games, or going to absolute maximum settings, everything seems to end up okay.

Any insight is appreciated

 

SOLVED:
everything I did is listed. last thing MAY be what fixed it, I can't be sure at this point.

My ASUS ROG Strix 1070ti started behaving strangely about two weeks ago now, just overnight.

I load up a game (Overwatch in that case) and the game runs at about 12FPS
I started investigating and noticed that my GPU clocks were not going above 139MHz 
I decided to try to test this, and noticed that when I launched furmark (I have an old version 1.19.1.0) that just launching the application (not the test) the GPU would start boosting assuming there was a game or another benchmark running in the background. After searching around I found that some people had success "locking" the clock of their cards in this situation using MSI Afterburner (I only had ASUS GPU tweak installed) so I attemped to do that. Locking the clock speed there didn't do anything, until once again, I launched this version of furmark. MSI Kombuster, btw, has the same issues the games do. If I launch furmark, the KOMBUSTER test starts working properly, but it doesn't do the same thing as furmark does and force the GPU to boost. It seems odd that one application will force the GPU to behave properly while others won't and seems to point me to some sort of software system issue rather than an actual defect with the card (although I did do some investigation there)

Actions taken thus far:
re-flashed the BIOS on the card
used DDU to uninstall drivers
Rolled back to an october release of Nvidia drivers
uninstalled the latest "quality update" from MS
Disassembled the card and cleaned the power contacts etc.
Placed the card in my wife's computer (does the same thing with the latest nvidia drivers and windows updates)
Completely uninstalled the drivers AGAIN, and reflashed AGAIN, currently running on windows provided nvidia drivers as they appear to be an even older version than I can find from Nvidia.

it should be noted that once I do force the card to a higher clock successfully everything appears to run fine, the GPU doesn't over heat, everything appears to be fine reported by GPU-z and HWmonitor
I'll record a video of what happens to try to avoid any confusion shortly and provide a link.

Edit:
Video Link as promised.

UPDATE:

"Reset" windows 10 with the option to keep my files, reinstalled drivers, not much else, got afterburner and kombuster for testing. (MSI's tools are much better than those from ASUS)
After I did that there was still no change, except that I noticed it seemed that every so often when I would ALT+Tab while running the stress test, it would just start working. I checked GPU-Z to see if anything there would help and noticed a constant PerfCap for power... so I decided to play with that some more. I did the typical up the power limits to 120% and that didn't seem to help. I'd read somewhere that playing with the voltage dial on 10 series cards didn't do anything really, so I decided not to play with it. UNTIL I DID, I enabled that in MSI Afterburner and set it to +20%.... Not sure what that did, but my card just started clocking normally again.

Naturally, because I don't know when enough is enough, I started working through overclocking the thing and crashed everything
this of course, reset all my overclock settings when I rebooted... HOWEVER
EVERYTHING STILL WORKS

not sure if somehow the voltage limits got screwy somehow in the firmware and flashing the firmware didn't help, but that seemed to jog it's Memory

I did see another post about this same issue with this card on this forum, however there were no solutions. Maybe now we have one!

testing with MSI Kombuster is fixed, but appears not much else. Games still don't force this GPU to come out of idle speeds....

things that work:
MSI Kombustor
Unigine Super Position (OpenGL/DirectX game and benchmark modes)
The Witcher 2 
Metro Last Light Redux
Overwatch on ultra at 150% render scale (still get over 100FPS, but this should NOT be needed, lower settings should provide higher framerate, not lower)

things that DON'T:
Overwatch at sensible settings
Dishonored
L4D2

 

[Noah0302]

I can relate to something just stop working overnight, like my PC just would not boot with the OC Ive had for months. Probably a degraded chip...

So I was wondering if maybe something on your card also degraded, be that the chip itself, memory, VRMs...

 

Sadly I have no idea on how to check those things, but I wish you good luck man.

[RebelKhaotic]

13 minutes ago, Noah0302 said:

I can relate to something just stop working overnight, like my PC just would not boot with the OC Ive had for months. Probably a degraded chip...

So I was wondering if maybe something on your card also degraded, be that the chip itself, memory, VRMs...

 

Sadly I have no idea on how to check those things, but I wish you good luck man.

I would agree with the degradation perspective IF the card was unable to boost at all and showed signs of problems when I forced it to. I don't even have to set the clock manually, the issue is literally that some applications (old furmark) and others (games, MSI Kombuster) don't seem to be able to the tell the GPU "HEY! WAKE UP!"
I did also remove the heatsink and backplate, clean everything up super nice, considered re-soldering the power connections but they looked fine once I cleaned them up...
I'm an RF Engineer by trade, so I'm not afraid of a little component level work especially minor soldering, but the card appears to be 100% fine, seems to be a communications glitch somewhere that I just can't seem to find. 

Would have figured flashing the firmware would be a bit like a "factory reset" but it doesn't appear to be so...

thanks for the well wishes though Noah!
 

[RebelKhaotic]

update, it appears about 85% of the time, just alt+tab'ing to another program seems to get things going... not every time, but MOST

[RebelKhaotic]

On 12/27/2019 at 3:43 PM, Noah0302 said:

I can relate to something just stop working overnight, like my PC just would not boot with the OC Ive had for months. Probably a degraded chip...

So I was wondering if maybe something on your card also degraded, be that the chip itself, memory, VRMs...

 

Sadly I have no idea on how to check those things, but I wish you good luck man.

You were correct muh dude, actually was degraded, managed to bring it back to life

[SavageNuke]

sounds like a driver issue

 

Edit:

you think it was degraded... it would boost under certain circumstances though, correct?

[Noah0302]

20 minutes ago, RebelKhaotic said:

You were correct muh dude, actually was degraded, managed to bring it back to life

Really? Congrats

 

How did you do it?

[RebelKhaotic]

2 hours ago, SavageNuke said:

sounds like a driver issue

 

Edit:

you think it was degraded... it would boost under certain circumstances though, correct?

If you read the recent updates I posted in an edit at the top, I got it fixed. I tried rolling back drivers through the last year. It was a hardware issue, power delivery solder joints appeared to have been the culprit. 

I know the post is long winded, but it details nearly everything I tried. including different drivers, flashing firmware, reinstalling windows, trying a different machine just about everything. 

 

[RebelKhaotic]

2 hours ago, Noah0302 said:

Really? Congrats

 

How did you do it?

A little solder, and a TON of patience.

[janak2345]

I have the exact same issue with my Zotac GTX 1060 since a couple of weeks now..

 

could you PLEASE tell me exactly which connector did you re-solder apart from the power connections (6 pin or 8 pin I presume) when you say "a few of the components that looked a little "less than brand new" all on the power delivery side".

 

also, everything on my gpu looks brand new to me... nothing looks degraded... but I am still willing to try this out since I have tried everything you have too.

 

Hell, I even went and got a new PSU :( which is now useless to me.