System freezes sometimes when loading 3DMark when using RX Vega 64

[Stamper_]

Hi everyone!

Currently have a "suspicious" RX Vega 64 AC in my system which sometimes causes freezes with 3DMark. Unigine Superposition does not seem to cause any issues. Very curious if others have similar issues with their RX Vega.

Computer Type: Custom build GPU: Gigabyte RX Vega 64 Aircooled 8GB HBM2
CPU: Intel i5 4690
Motherboard: Gigabyte GA-H97-HD3
RAM: 2x Crucial 4GB DDR3-1600Mhz
PSU: Fractal Design Edison M 650W Gold Certified
Operating System & Version: Windows 10 64-bit
GPU Drivers: Radeon 17.9.1 & 17.8.2 Crimson
Description of Original Problem: System freezes sometimes when loading 3DMark when using RX Vega 64

Steps to reproduce:

• Install 3DMark Advanced Edition & RX Vega on Windows 10
• Repeat the following 10 times:
• Open 3DMark Advanced (Home tab with the Timespy background should be loaded)
• While it is loading, move the mouse frantically over the 3DMark window
• System should freeze

Atleast once every 10 times this issue occurs.

Troubleshooting:

• Undervolt & lower core clocks. Does not seem to help as this issue happens when idle.
• Take out RX Vega and use integrated graphics seems to resolve the issue (even with AMD Crimson installed)
• Turn off iGPU
  
• Reinstall Windows 10.

I would appreciate any reports of reproducibility immensely!

 

[Shiari]

As engaging an activity it is to try and reproduce that, the sheer joy of the process meant that I just couldn't keep count. But no, I've not had luck reproducing, I did see a couple of times after exiting 3D Mark that the system froze for a second or two. It also sounds like something that would only show on particular configurations and not reliably (if it was reliable you wouldn't need to do it 10 times). Does freezing mean "total lockup" or does it recover after some time? I'd suspect a driver issue here rather than a bad card, though 17.8.2 should be far more likely to have had such an issue as it had quite a bit freezing going on for everyone. 17.9.1 might still have some freezing issues, it's only been out shortly though.

On another note: though it's not a bad PSU, your 650W PSU isn't really big enough for a Vega 64 and it could be causing you issues under load. When benching I regularly see my system pulling around 600W at the wall (sure, that is overclocked and with 4 HDD's and lots of fans), and you wouldn't want to stress your power supply basically to the max over and over again, it's just not good for it and it may have problems keeping up where it might not be able to provide stable power as it ages.

 

[Stamper_]

As engaging an activity it is to try and reproduce that, the sheer joy of the process meant that I just couldn't keep count. But no, I've not had luck reproducing, I did see a couple of times after exiting 3D Mark that the system froze for a second or two. It also sounds like something that would only show on particular configurations and not reliably (if it was reliable you wouldn't need to do it 10 times). Does freezing mean "total lockup" or does it recover after some time? I'd suspect a driver issue here rather than a bad card, though 17.8.2 should be far more likely to have had such an issue as it had quite a bit freezing going on for everyone. 17.9.1 might still have some freezing issues, it's only been out shortly though.

On another note: though it's not a bad PSU, your 650W PSU isn't really big enough for a Vega 64 and it could be causing you issues under load. When benching I regularly see my system pulling around 600W at the wall (sure, that is overclocked and with 4 HDD's and lots of fans), and you wouldn't want to stress your power supply basically to the max over and over again, it's just not good for it and it may have problems keeping up where it might not be able to provide stable power as it ages.

Thank you so much for the report! I know that must not have been an interesting 10 minutes xD But I really appreciate it!

With freezing I really mean complete lock-up. As in screen freezes, no movement or responses on any USB devices and a hard restart with the power button is required to get everything running again.

I noted the same driver suspicion to OCUK but, as the hero's that we do not deserve but get, they want to RMA the card to see if they can reproduce it. Your report shows atleast some evidence that hardware failure is an option.

Regarding the PSU, thank you for the headsup! Already saving for a new 850W PSU

 

[Webhito]

Hi there, first time poster, wanted to know if @Stamper_ was able to figure out what was happening with his cards. I currently have 2 that are doing exactly what he mentions, gaming, browsing and watching videos do not have this behavior but 3d mark does.

 

[chaosophy]

Welcome to the forums.

What driver are you on? A couple of us are having issues with the latest 17.12.x drivers.
If you are on those try 17.11.4 or 17.11.1.

 

[Webhito]

Thanks for the speedy reply!

Its actually 17.9.3, 17.11.4 was giving me the same issues. All this was prior to updating to the latest amd driver, which was causing some stuttering on my system, as well as freesync not working properly. Did a fresh windows install but nope, same thing.

 

[melmac]

Thanks for the speedy reply!

Its actually 17.9.3, 17.11.4 was giving me the same issues. All this was prior to updating to the latest amd driver, which was causing some stuttering on my system, as well as freesync not working properly. Did a fresh windows install but nope, same thing.

Run DDU in safe mode and uninstall all AMD drivers.
Then install 17.11.1

I have the same issue with 17.12.1 and can only solve it by installing doing running DDU and installing 17.11.1.

 

[Webhito]

Run DDU in safe mode and uninstall all AMD drivers.
Then install 17.11.1

I have the same issue with 17.12.1 and can only solve it by installing doing running DDU and installing 17.11.1.

Will give that a try, will reply in a bit after testing =).

 

[Webhito]

Nope, did the ddu uninstall plus the driver you suggested. This is what happens.

https://youtu.be/ZXsNlx0Zo-c

 

[Stamper_]

@Webhito Ever since, I have written it off to a software issue somewhere (probably 3DMark). I RMA'ed the card and they found it faulty. However, since then I have switched to a RX 580 with which I have the exact same issue. I do not have the issue on my (Intel) iGPU. I think something with the combination of 3DMark + AMD is causing it.

 

[Webhito]

@Webhito Ever since, I have written it off to a software issue somewhere (probably 3DMark). I RMA'ed the card and they found it faulty. However, since then I have switched to a RX 580 with which I have the exact same issue. I do not have the issue on my (Intel) iGPU. I think something with the combination of 3DMark + AMD is causing it.

Funny, I actually picked up an rx 580 as well just to cover my needs while I rma'd both of these since they both have the same behavior.
They used to work fine, just all of a sudden started locking up from the blue. Flashing the bios did nothing, neither does undervolting, underclocking, reinstalling windows. What a pity.

 

[melmac]

Nope, did the ddu uninstall plus the driver you suggested. This is what happens.

https://youtu.be/ZXsNlx0Zo-c

Sorry it didn't work for you. Maybe it's just a matter of trying each driver until you get one that works?

17.11.3 and 17.11.4 both crash in 3d mark like that if I don't install 17.11.1 first.

What CPU do you have?

 

[melmac]

@Webhito Ever since, I have written it off to a software issue somewhere (probably 3DMark). I RMA'ed the card and they found it faulty. However, since then I have switched to a RX 580 with which I have the exact same issue. I do not have the issue on my (Intel) iGPU. I think something with the combination of 3DMark + AMD is causing it.

Funny, I actually picked up an rx 580 as well just to cover my needs while I rma'd both of these since they both have the same behavior.
They used to work fine, just all of a sudden started locking up from the blue. Flashing the bios did nothing, neither does undervolting, underclocking, reinstalling windows. What a pity.

Now I am beginning to think my card might be faulty. I wonder is the 3D mark crashing thing a sign of things beginning to go wrong and soon I will be having the same crashes no matter what driver I use.

 

[Webhito]

Sorry it didn't work for you. Maybe it's just a matter of trying each driver until you get one that works?

17.11.3 and 17.11.4 both crash in 3d mark like that if I don't install 17.11.1 first.

What CPU do you have?

That's quite fine, silly things happen, sadly it seems to always occur with the amd stuff I purchase.

Its an 8700k with an asrock extreme4 board, 16 gigs of ddr4 3200 working as intended. I was also getting crashes playing homefront the revolution and needed to sign out of windows just to get it working again, but 3dmark is the only program that seems to completely freeze/hard lock it.

Not sure or not if moving anything in wattman causes the game crashes though, It seems that if I do ramp up the fan profile it tends to crash, including a radeon application has crashed error, I did a clean driver install again and have not moved anything yet just to see if this could be part of the issue. Will report if anything new happens.

 

[Shiari]

No, I'm pretty sure it's 3DMark, that initial screen is iffy. Always jerks a bit when it starts.

 

[Webhito]

From what I have read, if I mod the card via bios, it will no longer load drivers? Because my solution could actually be just modding it instead of using wattman and be done with it.

 

[henson0115]

i had the same in timespy on the 11.4's, havnt checked since updating to the 12's.

 

[Shiari]

This also happened to me in VRMark yesterday btw. I'd definitely put it down to what the software is doing in combination with our drivers.

 

[Webhito]

Well, I am glad to see I am not the only one with this issue. The weird thing though is that these cards were just fine until I updated to 17.12.1, either that or some really odd coincidence.

 

[melmac]

Well, I am glad to see I am not the only one with this issue. The weird thing though is that these cards were just fine until I updated to 17.12.1, either that or some really odd coincidence.

I had problems with every driver apart from 17.11.1. Either crashes in certain games or not been able to run games at all. I thought maybe the issue was that the 17.11.1 were WHQL while the other 17.11 drivers were not. So when 17.12.1 drivers came out I was hoping they would solve my problems. They solved a few of the crashes, but, 3D mark kept hard locking my PC(even without running any benchmark) and Superposition score was down 800 points. And Performance in several games had also dropped, most noticeably in VR games. I had to change Robo Recall to low settings and turn off all super sampling to get it running. So I went back to 17.11.1s and everything worked perfectly again.

But, today, I was bored and said I would give the 17.12.2 drivers a go. First I booted into safe mode and ran the AMD clean utility. Then I ran DDU and rebooted.
Then I downloaded and installed the latest chipset drivers. Rebooted again and then installed the 17.12.2 drivers and another reboot.

Ran superposition and, damn, same performance drop. Was going to just go back to 17.11.1 but I said I would run 3d mark, since I would be rebooting anyway. And, it worked perfectly, no hard locks. Then I tried some VR games and the performance was fine.

So, not sure what to say now!! Have you RMA'd your cards already? Maybe it might be worthwhile trying 17.12.2 on a clean install? ​