Vega 56 constant crashing

[broksaj]

Hey, everyone, I'm kinda desperate.

 

I bought the GPU on november, it's an ASRock Vega 56 Phantom Gaming card.

 

Since I got it I tried some overclocking (undervolting and increasing power limit to increase performance overall), it looked promising, though I quickly realized I didn't exactly win the lottery, since it wasn't overclocking a lot. I increased the power limit to 50%, undervolted, kept the rest the same, playing with memory clocks wasn't too successful, either, I could only get around 880 MHz before it would crash when stress testing and gaming.

So after some more time testing and trying overclocks, pretty much any setting I tried gave me crashes 2-3 times per day in games, sometimes when browsing, watching videos, etc...

I decided to not use any overclocks, and while that worked better, I realized recently after playing F1 2019, Read Dead Redemption 2 and several other games that I'm getting crashes even on stock settings...

So I'm playing a game for 1-3 hours, no problem, and then all of a sudden both of my monitors will go black and after 20-30 seconds it would come back, the game would crash and it would say that Radeon Host Application has stopped responding, or something similar...

I would really like to know if this is a driver problem or a faulty card, because I would like it sorted out as it's really annoying and not something I expected when I spend 220€.

 

Now some more info about my system...

I'm using the latest OPTIONAL GPU driver, freshly installed a couple times, tried the factory reset, tried the DDU uninstaller from safe mode, but nothing seemed to help.

I've got a 2 monitor setup going on, one 60Hz, other 144Hz + an LG TV plugged in but disabled in Windows display settings, I only use it when I want to watch a movie or something.

Here are the rest of my components.

Ryzen 5 1600x 3.60GHz w/ Cryorig H7 cooler
MSI B350 Tomahawk
Ballistix Sport LT 16GB DDR4-2400MHz
Gigabyte SSD 240GB & 600GB HDD
Seasonic FOCUS PLUS GOLD 650W
AOC 24G2 144hz (main monitor) & Lenovo Thinkvision X24

 

 

 

https://streamable.com/3z6bbc here's a video I took today, was playing Anno 1800 for around 3 hours and this happened. The sound of the game kept going, also the mouse was responsive and I "fixed" the problem by opening task manager and killing the game and moving my mouse around the screen to remove the bugging...

 

I also contacted ASRock and AMD and got an answer from ASRock as to what to try, but I don't really have other components to troubleshoot properly. They also told me that 1.5hrs of furmark stress testing should be enough to guarantee that my GPU is fine, so I did that without issues.

 

What should I do? I'm still quite confident it's either the card or the drivers, I've been testing for 2 days now and both days I've had crashes on stock settings. I also don't have any overclocks on CPU and RAM. So again, it would crash after minimum 1 hour, sometimes it'd take it a bit longer, but overall pretty consistent crashes in every modern game I tried so far.

[TetraSky]

13 minutes ago, broksaj said:

and not something I expected when I spend 220€.

And what, pray tell, did you expect after overclocking? OC'ing isn't a harmless practice. You're pushing the hardware past its limitations. It is quite possible that you broke it by doing so. You can't blame the card for your own actions.

 

As whether or not the issue is physical or software related... We can't tell for sure. Usually for physical damages, you'd see things like artifacts. It might even blue screen with an error. In your case, the screen goes to black and comes back after the display driver failed...

This could actually be a number of things, even unrelated to the GPU. In fact, sometimes it can be caused by faulty RAM.

 

What I would suggest, is using DDU to get rid of your current drivers, reinstall the latest ones and see from there.

 

If it still happens, I see you have a first gen Ryzen CPU. Have you updated the bios of your motherboard lately? New fixes for better memory compatibility and system stability have come out since the release of that CPU and it might be a good idea to update it.

After that, do run Prime95 on Blend for a while to test the RAM. (Faster than memtest86, which can take hours upon hours... while prime95 blend, usually finds issues within 30 minutes or so, but it might still be a good idea to run memtest overnight for 12+ hours after that.

[broksaj]

Well, I didn't really overlock per se, it was moslty undervolting and increasing power limit to get some more performance out, though I wasn't gaming enough back then, I only had a crash every few days. Though, even then I was able to pass these stress tests with ease. I'll do what you said and I'll test the RAM.

 

and yes, I updated the BIOS before I got the GPU, no new updates after that for my mobo, though.

 

edit : https://prnt.sc/smfr0t prime95 blend & temps in HWMonitor

[IAmAndre]

Maybe you should monitor your temperatures and power consumption. The crashes could be due to that. I have one personally and the spikes in power consumption make the PC crash at times, so I'll just get a better PSU.

[broksaj]

I did, when stress testing on stock I wasn't getting past 170W of power usage and the temp was stable around 80-81 with memory around 83-85 degrees at full GPU usage. When gaming it's even cooler than that usually. Though, I'm certain my PSU is more than capable enough to run my setup + I bought it with my GPU after doing a research on what PSU to get.

[broksaj]

Ran Memtest overnight (8-9hrs) and prime95 for 1 hour, both without issues. I restored the high performance in the power options, as I remembered playing with that when I had my old GPU and some other issues that made my PC freeze and have a weird buzzing sound when that would happen. As of then no crashes today (played RDR2 for a few hours). I'll do some more testing in the coming days to confirm that fixed it.

[Nena Trinity]

If latest drivers still have problem I can give you my current driver version it seems stable. U_U

[broksaj]

Just to update on this, I have confirmed with the AMD staff that it's a driver issue, my GPU is fine, as well as other components.

[Nena Trinity]

Tried driver 19.12.1?

[broksaj]

So... I managed to get my friend to test my components in his system and vice versa... Recently I found out a way to reproduce the crash in my system (running memtest in OCCT that tests GPU memory)... I don't know his specs exactly, but he has a 700w 80+bronze PSU, fx 8350

 

Vega56 + my PSU in my system on either PCI-e lanes = crash

Vega56 + his PSU in my system on either PCI-e lanes = crash

Vega56 + my PSU in his system = no crash

Vega56 + his PSU in his system = no crash

 

I already know it's not the ram and this mixed components test likely excludes the PSU, too, so it's probably the motherboard? I guess the only way is to order a new mobo and  test it with my components and if it works I can refund it and claim warranty on my own mobo(I bought it 1.5yrs ago so probably still have the warranty on it).

 

Please do let me know if there's something else I may have missed, but I think I pretty much did all I could to identify the problem.

 

Edit : could it even be CPU? I see ryzen on b350 chipset communicates directly to PCI-e slot?

[LOST TALE]

On 5/23/2020 at 1:28 PM, broksaj said:

I did, when stress testing on stock I wasn't getting past 170W of power usage and the temp was stable around 80-81 with memory around 83-85 degrees at full GPU usage. When gaming it's even cooler than that usually. Though, I'm certain my PSU is more than capable enough to run my setup + I bought it with my GPU after doing a research on what PSU to get.

80-81 is above thermal design. As a v56 owner, I know I get a warning anytime it exceeds 80C.

 

When you OC the v56, you want to lower all the steps to the lowest and undervolt that to the lowest that is stabe, and move up each step that way. You OC when you get to the final step, adding extra 1-3mv voltage to transition steps.

 

Trickiest OC ever. Then again I have the airboost. You really want to turn up the fan speed. I can't imagine what that's like on the stock cooler! I initially minimized the fan speed until I got used to it. Basically once it's really bad, it doesn't matter how much worse it is so you might aswell yolo it at 4000 rpm

[broksaj]

Well... I tried increasing the power limit to 50% and decreasing to -50% while also trying undervolting the card, it crashed either way, I don't think thermals are the issue.

 

When doing the OCCT memtest on my friend's system with either my own or his PSU, the GPU maxed out at around 65 degrees, so that can't be the issue, we tried stress testing it on stock on both systems and had identical thermals outside the case, so again, probably not the issue here. I'd really like to know if we did a good job at isolating the issue (motherboard) or if we missed something obvious.