Drivers & Software

[alexx855]

Hello, im having this weird problem since i brought my new GPU a few months ago, my whole build is new, the GPU was the last  purchase, previously i had a R9 290x in this same build, all works perfectly!

The crash is sometimes a black screen on all monitors and fans turns 100% speed, nothing to do here just force shutdown, sometime i got a blue screen with the message, THREAD STUCK IN DEVICE DRIVER.

All this happens randomly, between 1 hours to 4 maximum,  while gaming and casually browse on my second monitor.

i have tried all the possibilities, clean windows install, a lot of drivers versions 18.x.x / 19.x.x, diferente pci-e socket, diferente PSU, two diferents power rails, undervolt, power limit +50%, GPU and motherboard BIOS update, basically all known tricks. 

My systyem:

• AMD Graphics Card
  • Vega 56 Gigabyte OC version gigabyte.com/Graphics-Card/GV-RXVEGA56GAMING-OC-8GD
• Operating System
  • Desktop Windows 10 64bit
• Driver version installed
  • 19.11
•  Display Devices
  • LG IPS FULLHD (1920x1080@59Hz Active DP to HDMI adapter
  • 23MP55 (1080x1920@59Hz) HDMI
  • HISENSE (1920x1080@59Hz) HDMI
• Motherboard + Bios Revision
  • Z370 AORUS Gaming 5 + Bios F10
• CPU/APU
  • Intel Core i7 8700k
• Power Supply Unit  Make, Model & Wattage
  • Seasonic 750W Gold 
• RAM
  • 16GB Gskyll 4133hz

EDIT UPDATE:

The only solutions that works is the driver 17.11.1

https://www.amd.com/en/support/kb/release-notes/rn-rad-win-17-11-1  

whql-win10-64bit-radeon-software-crimson-relive-17.11.1-nov10.exe - Google Drive 

EDIT UPDATE 2020 

I have been testing the new bios, for a week i think, no crashes.

UPDATE THE BIOS USING THIS VGA Bios Collection: Gigabyte RX Vega 56 8 GB | TechPowerUp  (or search for your manufacturer if its not Gigabyte )

Im using the default settings, with the lastest driver version,  turned off zero rpm,  changed the fan speed and power limit -50%

EDIT UPDATE MID 2020

Radeon pro drivers seems to fix all problems

https://www.amd.com/en/support/professional-graphics/radeon-pro/radeon-pro-wx-x200-series/radeon-pro... 

[mojoe_90]

Windows drivers don't allow any oc performance gains.  If it was a software issue it would effect every single card on the market not just the odd card hear and there.

To me this has to be a hardware issue to be so random.  

If your happy with running a basic driver then good luck to you. 

[vendett3]

Eh, no. Faulty drivers have always affected either some or all gpus of the same familiy, there is no rule about that. An example of this happend to me not so long ago with an Nvidia gpu (980Ti), custum build ALSO by gigabyte, which had a weird crash very few people reported having. That is also how I came to know DDU uninstaller. The only solution was to rollback to older drivers that had came like 3 months prior. And no, we are not "happy" to run old drivers, we have to do it becuase how RMA works in different places around the world. 

[wolfiet]

The way RMA works in my country prevents me from replacing the card, so no. I am not "happy". It could very easily be a faulty driver as well, or even a faulty bios. The fact that many of these issues come from gigabyte cards is starting to point at the later.

[alexx855]

You're wrong, i can oc the card with the aorus engine sofware, or the old msi afterburner with the windows drivers....

[wolfiet]

Exactly this. You just loose access to wattman, not things like afterburner or aorus. just because you can't use AMD's official program doesn't mean you are unable to poke at the card. It does however mean we miss out on a LOT of bug fixes and performance improvements.

[vendett3]

Actually, after you downlowad (or windows does) the 17.7.2 patch (thats the precise number, look it up on AMD's webpage), you get the Radeon Adrenaline thing alongside wattman and the whole package. It's the older version mindyou, one of the first, but it's still something.

[wolfiet]

Yeah... I noticed, because windows just did that for me and now my card is unusable again. I guess I HAVE to pull the card now and go back to integrated. Great.

[wolfiet]

I am having the same problem listed here in every way, except I am on the reference vega 56, not the OC version. I am about to test the "fix" and use the standard windows drivers, but hopefully this draws some attention to this issue.

[wolfiet]

To add to this, using windows drivers works, but of course you loose access to wattman and the like. AMD needs to take a close look at this issue ASAP, something is wrong with the newer driver packages and the vega 56

[wolfiet]

As of today, my particular windows install patched and has started installing newer version of the driver that includes wattman. As a result, I can no-longer use the card at all. Until AMD decides to take action on this matter, I suppose we need to consider the vega 56 a no-go entirely.

[vendett3]

If you still want to fight it out, you could always uninstall the newer drivers with DDU in safe mode (or use the AMD utility), download the driver here Radeon Software Crimson ReLive Edition 17.7.2 Release Notes | AMD, and see if you get any results. Hope it helps! 

[wolfiet]

unfortunately, in my case any driver that adds wattman also brings with it the unpredictable black screen crash. I don't know what gigabyte did, give this is even happening on reference cards, but it messed them up entirely.

[wolfiet]

To update my problem and my horrible confirmation (because the stupid integrated gpu wouldn't run my monitor) I re-flashed the bios of the gigabyte card, first to an older version to see if that fixed, then I gave up and slapped the MSI equivalent on it. Magically all problems are "resolved" when I did this. I can't recommend this solution for those of you that still have warranties, but it did seem to make the card suddenly stable almost the instant the card was no-longer detected as gigabyte.

[vendett3]

Wow, so this potentially could mean that there was some SERIOUS mess up from Gigabyte, not to mention foul play if they knew they were sending non ready gpus.

[wolfiet]

I have no idea. I ran an 8 hour long test and never managed to get it to crash under the MSI bios. Whats worse, looking in the bios, I can't even spot a real difference. I have no idea what is going on here.

[wolfiet]

well, scratch all that. Even under the MSI bios it gets unstable, just much less often. Whatever the problem is, I am starting to suspect physical issues with gigabyte cards. Since they have made it clear they are not interested in doing real repairs for customers, we might be getting into lawsuit territory. Great.

[jdrobinson314]

There may be more to your monitor not working with your igpu than is letting on. This might be a holistic view however.

[wolfiet]

nah, the monitor is 4k and the igpu just plain didn't have the guts to drive it.

[jdrobinson314]

Anyone tried spying on their vrvdcc/vrm temps in hwinfo64? These gigabyte cards might be deficient in cooling or have poor contact with heatsink.

[vendett3]

So I tried to do a little bit of research (like, just a couple of videos, there wasn't that much info about it, if you have a post or something, please link it if you can) and there seems to be somewhat of a consensus that vrms run hot, like way over what other components set for normal. So I checked with hwinfo 64 and, on the sensors section, the three areas (thermal diode, HBM, and hot spot) showed an average of 70-69 degrees (with the exception of hot spot that, at one time, it showed a 94 degrees mark) while the card itself run at 67-74 degrees (CPUID HW monitor).These are not by any means outworldy temperatures in my opinion (they could be cooler, for sure). What's your take on this? Honestly, I dont think this is a temps issue, but it could fit into the Gigabyte messing up theory, whether if this is a drivers or hardware issue.

[vendett3]

PD: Keep in mind the measures I toke were while I was gaming, not in idel. On idel, temps on all boards (that is, HWinfo 64 and HW monitor) sit on 34-35 degrees.

[jdrobinson314]

Those are about right, but they aren't your vrm temperatures. This is called vddc in hwinfo64, make sure you have the latest version and not ~5.7.

There was a nasty incompatibility bug with that version and vega, ref; The RX Vega 64 Owners Thread | Page 197 | Overclockers UK Forums 

Make sure hwinfo64 isn't auto starting at login or boot and/or update hwinfo64.

[vendett3]

Hey, kind of an update on your previous comment. I finnally got to download a version of Hwmonitor that did read VCCD temps and, according to a test I did early this morning, they got as high as 83° celcius. While the overall gpu temp got as high as 74°. What's your take on this?

[vendett3]

*VDDC, sry.

[system2000]

i'm suspecting the VRMs from the beginning, if it's a software issue why is it happening with this model and not the reference model by gigabyte and other brands, but alex says he is being using drivers 17.7 without any problems,these drivers don't support vega which makes me think that, the card's full potential is not being used, i added a video showing how cooling the backplate using exhaust fans on top helps in reducing VRM temps, but it's under moderation

[wolfiet]

Yeah, at this point it is looking like a complete design flaw. Unfortunately, I will just have to use mine as a paperweight until money comes for the next financial mistake. I really don't think I will be using gigabyte again for much of anything.

[vendett3]

Someone said earlier in this post that it was using a reference gigabyte and had the same issue. 

[system2000]

but that's rare unlike the gaming oc the majority are bad reviews, i looked for gigabyte vega 56 reference on amazon to see the reviews, it has 8 good reviews and rated 5 stars, search this model GV-RXVEGA56-8GD-B on amazon

[wolfiet]

from what measurements I got, the temps were in line with any other version of the same card. I am not sure at all.

[jdrobinson314]

My prior comment is awaiting moderation because it contains a link. The hwinfo64 version 5.7 had a bug related to probing for temperatures that was causing crashes. It was solved by not starting hwinfo64 at boot.

I doubt that's the problem, but the VDDC temperature is what I was talking about, not the other temperature - those seem normal to me.

[vendett3]

Oh, sorry, my bad. Well, I tried to look after the VDDC temps, but hwinfo64 wont show any. Just the VDDC voltage, which is sitting at 0.7-0.8 volts. Perhaps it has to do with the driver (the old 17.7 we are forced to use)? Or maybe I should try older versions of Hwinfo and see if they pick them up. 

[jdrobinson314]

The older versions probably won't be much use - probably good idea to try the 6.03 beta version.

[vendett3]

Tried with the beta 6.03, same thing. Will show the voltage, but not the temp.

[wolfiet]

Is there any way we can get this marked from "assumed answered"? This issue needs to be addressed by an actual AMD rep.

[mojoe_90]

I have messaged a rep amdmatt and asked him for his input on the problem.   

[jdrobinson314]

Any differences in or ability to run with 19.3.1?

[vendett3]

Haven't try for fear to put the GPU on risk, and since it is from the same family that the previous one and the release notes do not adress this specific issue, I have not dare to try them. So far, the 17.7 are still holding. 

[jdrobinson314]

AMD will likely only ever knowingly resolve known issues as fixed issues in patch notes.

At one point, audio over displayport broke in one driver - and then got fixed in another. All this went on without so much as a peep out of AMD.

[wolfiet]

None at all. Still completely worthless under new drivers. And now that I cannot seem to get the old drivers with no AMD settings app, the entire gpu has become a paperweight I have sitting on my desk to glare at from time to time.

[vendett3]

Hey, can confirm what wolfiet said. The 19.3 drivers did nothing to resolve the issue. I'm waiting on a response from Gigabyte Tech Support on monday; they will probably agree on the RMA... At least I'm praying on that...