I've also created a question on the Windows forums, and so far have received the usual "roll-back drivers", "refresh windows', etc. XFX has been saying the same things. Even if it is a hardware issue created by faulty drivers, I want to post it to as many places so that we can elevate the issue, and maybe get a fix for those whose cards have not yet been compromised. I don't expect to see XFX send me a new card, or AMD or Microsoft for that matter. It is just frustrating that there is essentially radio silence, or at least negligible support. If my card was new I would have done an RMA already, but I hope that these companies still would like to support those end-users who are outside of that warranty period.
I just discovered something new to help us along. I've put my HD7870 in the secondary pcie slot, and am running my HD5750 as my primary card. On installing the 7870 drivers, the black screen we are getting is actually a windows BSOD with the classic "Thread stuck in driver" error message. Going back to 2016, the main culprit then was the audio drivers. I've tried installing the driver with no audio drivers present, and with my audio drivers fully updated from the supplier (it is an intel mobo, so realtek drivers from win8.1), but to no avail.
While it is still not solved, it at least narrows the search, although AMD drivers have had their fair share of stuck threads. Maybe we'll solve it yet. I refuse to believe this is hardware failure. I have tried flashing some modded vbioses, but they are no help either. My XFX bios does not display its vrm in VBE7, and so I don't have much control over voltages, and trying to set the memory clocks for anything but state 0 doesn't work. I'm wondering if that is due to it not being a reference card. I poked into the bios in a hex editor, but even with the ATOMBIOSREADER, I can find some of the memory straps, but I'm not really sure what to do or where to go from there. It's very different from all the tutorials which are geared to polaris devices for crypto-mining.
Hopefully this helps point some of you in the right direction.
It is just frustrating that there is essentially radio silence, or at least negligible support.
I totally get you, man. This is what gets me on my nerves: no one is able to tell us anything, and no one seems to really care about it (aka AMD/Microsoft). Sorry for not replying these days, I've been quite busy at work. Thank you for your updates regarding XFX/Microsoft and the testing you've been doing!
Does that mean that this may be a solvable issue after all? That apparently our real problem is a stuck thread?
This reminds me that I was thinking about updating my BIOS, and if I'm not mistaken this can usually help solving the "Thread stuck in driver" error. The thing is, I've never updated it and I'm afraid of bricking it. Do you think this could be a possible solution?
If you have a desktop, Go to the Motherboard's manufacturer and they always have a software specifically made to Update the Motherboard's BIOS. Makes it easy to update the BIOS now a days. Just follow the instructions of the program. The ONLY thing that can brick your Motherboard during the update process is if you lose power or the computer shuts down unexpectedly during the flashing process before the BIOS is fully flashed.
I have an Asus Motherboard which I flashed my BIOS several times since I purchased it without any problems using Asus program for flashing my BIOS.
Laptops also have their own programs for flashing their BIOS same as Motherboard manufacturers.
This may help you, although unfortunately it does nothing for me. I am already running the latest bios for my Intel DP67BG board, rev 2209. It has been running that bios since I bought it used, and I reflashed it just to be sure there was no corruption or anything.
Definitely update your bios and see if it fixes anything, but I wouldn't hold my breath.
XFX seems to have lost interest in my case, because after their stock answers (update bios, check windows integrity, reinstall windows) they have not even stated they will look into it, despite the fact that I have insisted there are multiple users with Pitcairn devices with the same issue and that it is quite coincidental that they all failed in the same way at around the same time.
Windows forums have offered the same advice, and unfortunately the "experts" there often suggest bloatware driver updaters and other 'diagnostic' tools that I wouldn't wish on an enemy's computer. I haven't heard a response from anyone who seems even remotely tied to the Windows development team, or even their PR team.
As far as the device goes, I've exhausted all by one last option. I've tried flashing complementary vbioses from similar devices to see if their memory timing or voltage profiles would help (they didn't), even going so far as to flash an R9 270x bios. This did boot, but was still being recognized as an HD7870, and so subsequently failed a driver install.
I have one RAM chip that is not exactly matching my pair, and while they are running in single channel, and I have had no issues with stability during memtest runs, I tried removing the chip and reinstalling the gpu. Again, this didn't do anything expect cause problems with my Windows 10 activation (apparently it doesn't like downgrading Ram amounts, only upgrading).
I created a bootable usb of Video Memory Stress Test (vmtce: https://mikelab.kiev.ua/index_en.php?page=PROGRAMS/vmt_en ) and tested it there. My HD5750 was totally fine, but my HD7870 through a significant amount of errors. On the second test, it consistently output verified its write data as being out by 20 digits. I'm not sure if this is an actual error, or just an issue caused by the fact that this program cannot really handle RAM amounts of 2gb and over. If it is in fact reporting properly, it means some of the gpu memory is munted, which could be consistent with a voltage issue or improper cooling due to a faulty Wattman profile for our cards that could have been causing the failure.
The only thing left for me to try is to reflow the solder, something I am hesitant to do if there is no reason to. I will like pass on the baking method, and instead opt for a heat gun, trying to reflow the main chip and the ram chips. I'm going to hold off on this for a while, and see if anyone finds a solution that is more than just DDUing the drivers and reinstalling.
I was just trying to set ezrawolvenheart mind at ease about updating his BIOS since he has never done it before. Someone else posted that he can risk bricking his motherboard by updating his BIOS. Which is true. But as I mentioned in my reply that all Motherboard and Laptop Manufacturers have specific programs to flash BIOS relatively safe now a days. As long as ezrawolvenheart doesn't lose power to the computer or does something to interrupt the Flashing process of the BIOS it should be a safe and easy task to do.
Upgrading BIOS has solved some Users computer problems concerning drivers in the past. I feel it is a good idea to update your Motherboard or Laptop BIOS to the latest due to compatibility or security issues that arise as technology advances and your motherboard or laptop becomes more obsolete.
Good call, sorry if it sounded like I was negating your advice. I totally agree the updating to the most recent bios is the first part of call in trying to diagnose this issue, particularly for users running embedded gpus. I remember the first time I flashed a Mobo bios, and how worried I was that I would brick my device. You never forget your first time
My caveat to your statement was probably just my old, jaded reality that this solution won't fix the problem. I hope that for most people the solution is as simple as a bios flash. I guess I'm just sad that tech companies think that 5 years is too long to support a device, and that a device lasting that long is an exception not a rule. I think as consumers though we have brought some of this on ourselves, always lining up to trade in our 6 month old device for the newest shiny thing.
I guess the real issue is that unlike the gpus of old, we are really pushing the new silicone to it's limit. With that comes very small margins between optimum performance and critical Hardware failure. The problem is that we as end consumers can sometimes have or devices bricked by miscalculations on the part of the driver engineers, and if we are out of that support cycle range, tough luck. It's also not worth the resources to prove negligence, it to file suit to demand compensation. /Rant
Thanks for walking those who have trepidation about the bios flash process through it, your guide was informative and reassuring.