hardware issue Board power sensor reporting wrong values causing RTX 2060 Mobile to throttle core clock

[muzikhan]

Hello everyone, 

 

As the title suggests, my now 20 month old RTX 2060 Mobile decided that it was time to slow itself down. My GPU core clock has been stuck at 300MHz for about a month now and the issue started when I did a near full software upgrade for my HP Omen 15-dc1030nr laptop. (S/N: 5CD9193WS3) I updated the BIOS and the VBIOS using the utilities provided by HP's driver website and also downloaded and installed the newly released 461.09 GPU driver from NVIDIA GeForce Experience. Now, I do not think that any of these factors actually contributed to the problem but I think it would be a good idea to share what happened before the issue started.

 

Anyway, after updating the system, I rebooted, as usual, and launched Rocket League to have a nice gaming session with my friends. The thing is, the framerate was damn near locked to 25 at the lowest settings and the input lag rendered the game completely unplayable. So I decided to reinstall Windows multiple times, each time using a different BIOS and graphics driver version and I even experimented with supported VBIOSes from the VGA BIOS Collection site. None of them made a noticeable difference to gaming performance. After that, I decided that the task at hand was bigger than I could handle, so I took the laptop to UBreakIFix, which HP Support said worked with them. After a week and a half of probing the motherboard and trying different driver revisions and BIOS versions, they too weren't able to find a single thing wrong with the board. They just told me to wait until NVIDIA or HP released a patch to fix it. NVIDIA released 2 hotfixes and a major revision but the problem still persists.

 

So I decided to do some digging and found out that, even at idle with absolutely nothing running, the board power sensor for the GPU was reporting upwards of 280W. (image below, disregard the temps, I unplugged the fans to prove that nothing except that sensor reading matters to the core clock.) Which is quite odd since the RTX 2060 in here is only a 90W card, and even weirder is the fact that the power sensor in the GPU die is reporting 5W at idle and 18W at full load. After seeing that, I decided to disassemble my laptop and see the possible failure points. I found this: 6 VRM stages and right next to those, the MPS1910/MP2886A PMIC (image below, link to chip: MP2886A). Am I right to have the suspicion that this chip is the one reporting the wrong values? If not, what do I need to do to repair the fault for as cheap as possible? HP and UBreakIFix told me that the only way to fix it would be to replace the whole motherboard which costs $534 (The number they gave me when I asked HP for a quote. UBreakIFix told me it was a reasonable price and that it wouldn't get much cheaper than that even with used parts.) One thing that I think is important to mention is that I used the laptop without a battery for a few weeks before the issue started and HP confirmed that the battery assists the PSU while under load (which, apparently, is why they swell so much, this was the reason I took it out in the first place.) That might have done some damage to the PMIC beforehand.

 

In conclusion, there is absolutely nothing left on the software side that I can test or change, but I am very willing to shell out a few tens of bucks for replacement parts to get this problem sorted, since this is unfortunately the only rig I have. Any help would be very much appreciated. (Sorry for the really long read, I don't know how else to describe the things I experienced.)

 

P.S: some extra weird stuff that I think is worth mentioning: the sensor reported board power draw randomly dropped to 110W after I flashed this VBIOS: HP RTX 2060 Mobile 6 GB I thought I could now go back to the stock VBIOS provided by HP on their driver website since most of the problem seemed to be gone, but nope, the reading shot back up to ~280W and I have not been able to get that number down ever since. This problem persists both on battery (the reading drops to 230W on battery which is still way too much but the die power draw remains at 5W at idle and 18W at full load.) and on AC power but I do not think this is a power supply problem since the output voltage on the DC jack has always been the same: 19.5-20.0V (results from internal power sensor, confirmed with a multimeter test.)