Help troubleshooting Sapphire R9 390 instability and crashes

[stormylagrange]

I have been having a lot of issues with my Sapphire R9 390 graphics card. I have gone through a total of 4 as a result of RMAs, and I am currently using card #4.

Here is the list of symptoms I have had with each iteration of the card.

 

• Card 1: Developed fault after 5 months. Display would randomly cut out to a black screen, but just before it happened, vertical lines appeared. Very difficult to reproduce, crashes are random throughout the day, may not even occur for a week. Only one crash during a game. I have tried stress testing the GPU with unigine and furmake, and this card did not fail. It only seems to fail when GPU is idle. If there is audio in the buffer, I would hear a small part of it repeated over and over. Crash is like this: http://imgur.com/a/i21nc
   
• Card 2: In addition to card 1 issues, some horizontal flickering can be seen. Crashes and flickering occurs in Windows and in game. Easily reproducable. I also did get this spectacular crash: https://vimeo.com/178229176
   
• Card 3: No 3D games would work. Every time anything 3D rendered would come up, the game would crash and AMD Crimson threw an error: "Application has been blocked from accessing graphics hardware". Event log shows "Display driver amdkmdap stopped responding and has successfully recovered". I went out and got a GTX 970, was able to play games without error, though I did not thoroughly test the 970 for problems similar to prior cards. Easily reproducible.
   
• Card 4: Same as card 1. Some horizontal flickering as card 2, but hasn't crashed during games yet. Very difficult to reproduce. Just as I was typing this entire post in Notepad, the vertical line crash occured, but the display recovered and I did not have to do a hard reset (so far this has only happened once).

I have some videos detailing similar crashes, these are mainly card #2: https://vimeo.com/user47724713

Some specs:

 

• CPU: Intel Core i5 6600K OC to 3.5 GHz (BIOS preset)
• Cooler: CM Hyper212 EVO with Scythe Gentle Typhoon 120mm x2
• Fans: 5x Fractal GP-14. 4 intake, 1 exhaust
• GPU: Sapphire Radeon R9 390 8GB no overclock
• MB: Gigabyte GA-Z170-HD3P rev 1.0 latest BIOS
• RAM: Crucial 16GB (2x8GB) DDR4 2400
• PSU: EVGA SuperNOVA 650W GS
• SSD: Mushkin Chronos 120GB
• HDD: Toshiba 3TB, WD 3TB, WD 2TB
• Wifi: Asus PCE-AC56
• OS: Windows 10 Anniversary
• Case: Fractal Define R5

 

Here are the troubleshooting steps I have gone through (no particular order), but yielded same problems.

• Clean install
• Various drivers: Windows Update WHQL, Sapphire Drivers, AMD Catalyst Stable and Beta, latest AMD Crimson, latest AMD Crimson hotfix
• Removed CPU overclock
• Reset BIOS to factory default
• Moved R9 390 to different PCIe slots
• Used Windows 10 Anniversary, Windows 10 RTM, Windows 7 Professional (all 64-bit), Linux Mint, Ubuntu (surprisingly using Linux had more frequent crashes on card #2)
• Ran Memtest
• Upgraded to latest BIOS

Other things I have tried:

• When using card #3, I got a GTX 970 and gamed worked perfectly. Did not test for other symptoms.
• Switched to integrated graphics. Problem does NOT occur.

 

I do NOT believe the R9 390 is defective. I highly doubt I can go through 4 defective cards. Something else in my system must be causing this.

 

Motherboard isn't exactly the best, I have read reviews that the VRMs and power phases are not the best quality. I am thinking of upgrading to a better board (MSI Z170 or Asus Z170-A) if it is indeed the issue.

 

PSU is of high quality. I was told that maybe I should upgrade the PSU to something like 750W, but I thought 650W was sufficient. I didn't want to go higher because it would be a waste.

[Mindovermaster Moderator]

I'd blame the board, not the PSU/GPU...

 

Can you put the R9 390 in another computer, and verify that it is not the card?

[stormylagrange]

1 hour ago, Mindovermaster said:

I'd blame the board, not the PSU/GPU...

 

Can you put the R9 390 in another computer, and verify that it is not the card?

I actually don't have any spare computers lying around. I really doubt it is the card, I have already gone through 4 of them.

 

Can I ask why you think the motherboard is at fault?

[Hum]

I vote for the board too.

 

You might try looking at the motherboard voltages going to the slots -- see if they go too high or vary a lot.

 

Look in the BIOS.

 

Edited October 9, 2016 by Hum

[stormylagrange]

Just now, Hum said:

I vote for the board too.

 

Can you at least try a different pci-e slots ?

I have tried different PCIe slots, same problems occur.

[Mindovermaster Moderator]

33 minutes ago, stormylagrange said:

I actually don't have any spare computers lying around. I really doubt it is the card, I have already gone through 4 of them.

 

Can I ask why you think the motherboard is at fault?

I would ask a friend if you could try the card on their system.

 

I think it's the board, as in, the PSU is top notch. It couldn't be anything else than the board. Also, as you said above "VRMs and power phases are not the best quality" that puts a big pointer at the issue.

[TheReaperMan]

I would agree with the above, try the card in a other system(friends). As it could be a faulty motherboard.

 

You will also know 100% it is not the gpu causing your problems. 

[Mindovermaster Moderator]

3 hours ago, stormylagrange said:

PSU is of high quality.

btw, care to tell us what that means? High quality to you can be crap to us. Model? I'm guessing it's a 650W.

[Andre S. Veteran]

Do you still have the GTX 970? Just use that... Yeah seems unlikely to go through 4 defective cards but I've had 2 defective ones in a row from XFX, they come in batches sometimes. I bet you're losing money on shipping for all those RMAs, I'd just switch to a different brand of video card at this point.

 

PSU or mobo could be bad, but it's really hard to diagnose this. Or voltage spikes?

[T3X4S]

Just because the PSU is a good one, doesnt mean it isnt @ fault.  Even the good ones can have issues.

I would look to see if you have a bad PCIe cable(s).
If its modular, order some new ones, if its not, might have it looked @ or just buy another one, run it for a ~month - see what happens.  If the problems stop - there is your answer.  If the problems continue, return the PSU withing the timeframe of whatever retailer you bought it from.

Its a good PSU, but I would personally have gone for something a little heftier.  Maybe its a little flaky and not delivering quality current when pushed.  Only way to know is swap it.  PSU testers are pretty pointless, as they just tell what you have @ each connection @ idle, the normal ones dont push the PSU seeing if it falters under stress.


BTW - Seasonic just released (3) Titanium rated PSUs in 650,750, and 850 wattage.  Not to say that is what you need, or what you should do, but when I went to see what a Seasonic 650W would cost, I saw they had Titanium PRIME Flagship PSUs now.

[Mindovermaster Moderator]

8 hours ago, T3X4S said:

BTW - Seasonic just released (3) Titanium rated PSUs in 650,750, and 850 wattage.  Not to say that is what you need, or what you should do, but when I went to see what a Seasonic 650W would cost, I saw they had Titanium PRIME Flagship PSUs now.

And the 750 is on sale