XFX RX 480 8 GB Unstable Over Standard Clock

casiophreak

Reputable
Aug 3, 2015
9
0
4,510
Hey guys, I have my XFX RX 480 8 GB Custom Blackplate Edition for over a year now. This one has factory OC to 1328 Mhz.

The issues I found: since a few months ago, the card has been behaving oddly, crashes when playing games and doing benchmark. Here are some instances:
- 3D Mark Time Spy and Firestrike always crash after a minute
- Unigine Heaven does some black screen with music plays in background, or sometimes it just freezes
- Cinebench OpenGL always crash in seconds
- Interestingly, Furmark doesn't crash, no problem at all
- For games, Dota 2, The Surge and Nier Automata always shows screen freeze but games are actually still playing in the background. I can move my characters, pausing games and such, indicated by the sounds and music. But screen is still freezed. This usually happens right after I take control of the characters ingame (not main menu, intro movies, etc)
- Mass Effect Andromeda simply crashes after a few seconds in main menu. Error msg: GPU removed bla bla bla
- Deus Ex Mankind DIvided also crashes with error msg: not enough resource (already tried the lowest graphical settings)
- Planetside 2 playable with occasional flashes of black screens

Then, I tried lowering the clock to its standard of 1120 Mhz. All the problems gone, the card works flawlessly. I tried bumping up the clock and it seems it can only works flawlessly at 1190 Mhz. Above that, all the issues persist.

My specs;
Mobo: Gigabyte GA-Z77x-D3H
CPU: i7 2600k
CPU Cooler: Deepcool Lucifer v2
RAM: Generic SK-Hynix 8 GB (single, I took it from a built-up PC)
PSU: Seasonic X750 W Gold
Case: Corsair Carbide 400c
OS: Win 10

Now I have migrated to Ryzen with specs:
CPU: R5 1600
CPU Cooler: Stock Wraith Spire
Mobo: ASrock X370 Taichi
RAM: Corsair LPX 16 GB (dual channel) @2933
everything else is the same, and I can still reproduce the same error on the GPU.

GPU on latest BIOS

So my question;
1. What's the cause of this? Because from the first time I bought the card prior to this issue, I can play games fine at its factory OC settings in 1328.
2. Any insight as to make it work again on 1328 factory OC?

Thank you good people!
 

c4s2k3

Reputable
Sep 17, 2015
347
0
4,960
Have you checked the temperature of the GPU during those situations? I agree with parani that if you were sold a card that can't run reliably at the advertised clock rate, the manufacturer needs to make good on it, but if the air flow in the case is causing the GPU to overheat, it may not be a warranty issue.
 

casiophreak

Reputable
Aug 3, 2015
9
0
4,510


Prior the issue, unfortunately I never tracked the temps. But then I remember that before I used my Corsair 400c, I used a generic case with bad airflow. The issues showed right before I swithced to my Corsair case.

For now though, with 1190 Mhz my temp never get passed 80 degrees C in 3d Mark and Unigine.

With 1328 Mhz, it's hard to measure the temp because the GPU already crashes even before it gets to 50 degrees C. It was once when I tested in Unigine that it got to 90 degrees and then crashed.
 

c4s2k3

Reputable
Sep 17, 2015
347
0
4,960
Do you have a GPU utility such as EVGA Precision X or MSI Afterburner? They let you set up custom fan curves for the GPU (a temperature to fan RPM map, essentially). Try setting up a curve in the utility, and make sure the "Auto" fan mode is actually engaged in the utility as well. Try some tests and see if the issue improves. I would re-run your 3dMark and Unigine tests @ 1190 first to confirm the curve is cooling the card better first. You saw around 80c previosuly. A more aggressive fan curve should lower that by a few degrees. If that looks good, try it at stock (1328) and see what kind of result you get.
 

casiophreak

Reputable
Aug 3, 2015
9
0
4,510


Thanks for the advice! All the time I only used Radeon Wattman to set fan speed but now I noticed its fan curve setting take quite a long time to shift to higher rpm when needed. I followed your advice and installed MSI Afterburner, the fan speed ramp up in no time after a few seconds of 100% GPU activity. I set the fan curve for 50, 60, 70, and 80 degrees C.

This has led to the fact that the GPU behave strangely; when it gets to 100% load, it reaches 80 degrees C temps in under 10 seconds! I mean, is that normal? I thought temps should built up progressively in a space of few minutes or even hours, but this... in under 10 seconds, the temp already reached ~80 degrees. Signs of thermal problem inside the card?