GTX970 power question

[Carosel43]

I have a ASUS gtx 970 strix, Is anyone running one of these? ii am trying to repair the one i have and it seems to be sucking down the juice big time. 

 

In GPUZ it lists the perfcap reason as mainly power, the the graph also lists VRel and VOp jumping around all together as the frequency bounces like crazy and frametimes suck. There is also loads of coil whine pulsing with the hitches in frametime. It eventually hard crashes needing a reboot. every other nvidia card i have (760, 960, 1070) all list VRel as their perfcap reason and they all sit solidly showing VRel as the perfcap reason and clocks etc are stable. If i crank the power limit in afterburner pwr disappears as the limit and VRel and VOp fight it out...until the card crashes. The 960 i have was broken (listing power as the limiting factor) and i was able to repair it by removing all of the inductors from the VRM and soldering them back on. It now lists VRel as the only limiting factor so im looking at power delivery. 

 

The card seems happiest when VDDC hits 1.2v, but gets quite upset when this drops. It holds 1.2v when i crank the power, but it then draws 195w! this seems excessive, if anyone has one of these cards, or any 970 i suppose, can you post some shots of your gpuz monitoring while the card is under load? I just want to see where i am at and where to look for problems. 

 

thanks

[SimplyGone]

When i get my 6600k/strix 970 rig back i'll run some numbers unless someone beats me to it

[Carosel43]

brilliant thanks. 

 

This card has been a rollercoaster. I got it knowing it to be faulty, crashing after a few minutes in games. Crispy thermal paste and overheating i thought to myself. 

 

Sure enough, the thermal paste was crispy, so i cleaned it and thought it was all good. Nope, crashes as described. 

 

Fiddling around i wondered to myself if the bios had been messed about with. Found the right bios, flashed it...oops, nvflash took a dump and soft bricked the card. nothing would fix it, no dos -F, no autoexec, nothing. It was dead and after 5 hours trying i gave up. 

 

Ordered one of those USB programmer clip things to program the chip directly, managed to get the original bios back on it and its alive again but still with the original issues. 

 

The saga continues...

 

 

[SimplyGone]

My strix is a bit of a legend like that.  I was helping my cousin build his first gaming PC and he didn't really listen to me much and ran out of cash before getting his GPU.  So while he was building we'd swapped his PS4 for a retail copy of windows 10 i'd bought.  So i took the PS4 down to CEX and part exed it for the Strix 970 (think i cost me 100 quid on top of PS4) and i said to him i'd sell it to him for the 100 quid provided i got it back when he upgraded for no money.  4 years he ran it for before upgrading to a 2070 super.  True to his word i got it back for no money so all it stood me for was the 100 quid for the retail windows 10 key.  Then about 8 months ago the 2060 super in my Nephews PC died and he used the 970 until he could get a replacement.  I use it in my VR sim rig now but that whole computer is away helping someone else at the minute with their things going on lol

[Kilrah]

This sounds like the card has issues in the power supply circuitry, and is potentially measuring power wrong. As a result it thinks it's drawing too much and limits itself even when it probably isn't. 

[Carosel43]

I was wondering if the power was reporting incorrectly. The card does get very hot though so that power is going somewhere. Core temps report in the mid 60's -70'c, but the backplate is good for cooking bacon. 

 

One thing i forgot to mention is the physical damage to the card. Its been abused a bit, probably dropped? the fins on the heatsink were very bent and the whole pcb is both bent and twisted. I wonder if this physical twisting has cracked a solder joint somewhere. I would give it the heatgun reflow but that is a bit of a last resort bodge job. 

[Carosel43]

small update. 

 

I removed all the inductors from the vrm, cleaned, and replaced them. As with my GTX960 repair this got rid of the power problem (VRel VOp now only perfcap reason) but the card still crashes under heavy load. Downclocking the card about 75mhz on the core (it runs at 1270mhz now vs 1253 official boost clock) seems to keep it running. Anything near 1300mhz causes a crash. 

 

Investigations continue..