Drivers & Software

[thecyclegeek]

I've been having random black screens and system lockup with my MSI Radeon R9 390 8GB for about a year or so. It happens when running desktop applications and never with full screen games. Turning off graphics / hardware acceleration in applications (Firefox, Chrome, Spotify, etc) seemed to make it less frequent but it still occurs randomly and without warning.

Now, within the past several weeks, the graphics driver has started crashing in games, and making many of them unusable. Within 5-10 minutes of playing Running with Rifles, Fortnite, Deep Rock Galactic, and other games, I get screen flickering and then a crash reporter popup telling me the drivers crashed. I never receive any feedback from those reports, nor can I determine the source (I have the latest drivers, always). Temperatures seem stable and don't reach crazy levels. I don't know what to do now since I can't buy any replacements (all cards on the market are sold out or ridiculously overpriced in the second-hand market).

What are the steps to rollback the drivers to see if that helps? Can cards of this age be serviced / fixed? Are there any logs that I can reference for the card or any troubleshooting steps I can apply?

Specs:

Motherboard: MSI Z170A GAMING M5 

CPU: Intel Core i5-6600K 6M Skylake Quad-Core 3.5 GHz 

Video card: MSI Radeon R9 390 8GB

Memory: 32GB DDR4

[benman2785]

hi,

as the R9 390 is getting old you should "REPASTE" them now... (hotspots can be critical on old paste - even if sensor says "normal temps")
its usually not that difficult - just use a good thermalpaste like ThermalGrizzly Kryonaut or similar

if error than persists:

downclock max VRAM speed about 10% and add 20% to powerlimit...

[thecyclegeek]

---

@benman2785 wrote:

hi,

as the R9 390 is getting old you should "REPASTE" them now... (hotspots can be critical on old paste - even if sensor says "normal temps")
its usually not that difficult - just use a good thermalpaste like ThermalGrizzly Kryonaut or similar

if error than persists:

downclock max VRAM speed about 10% and add 20% to powerlimit...

---

I'm not sure how the thermal paste I have stacks up with the one you mentioned (Arctic Silver 5 High-Density Polysynthetic Silver Thermal Compound AS5-3.5G) but I can attempt that. I noticed in a newer, high-end game, my temperatures reached about 82C, which is higher than what I saw before, and then it crashed.

Questions: What's a safe solution to use to remove the old thermal paste without damaging other components? Should I apply the new compound as a blob, an X, or smooth it out prior to application?

I've reduced the max clock speed already, from 1040Mhz max to about 1000Mhz and it seemed to help, but I haven't touched the VRAM at all.

[kingfish]

@goodplay  has some experience with the 390 cards..maybe he can add something

[benman2785]

@thecyclegeek 

arctic silver is okish - but better buy something as good as ThermalGrizzly Kryonaut (its not that expensive - and you only need the small one)

do a FULL COVERING of the complete DIE! no X or no dot - the whole chipsurface has to be covered (at GPUs reasonable more is better)

to clean buy some iso-alcohol - its cheap and gets the job done...
or you buy "Arctic Clean Thermal Remover Kit" - 1x is citric acid that "melts" hard thermalpaste and the other is iso-alc (it is numbered what to use first)

[thecyclegeek]

Thanks!

Due to the pandemic, I'm having a hard time getting iso-propyl alcohol at a high-enough concentration (+91% - all local stores are sold out so I had to order something online) to clean off the surface and the 1-2 cleaning kit seems overpriced or would take too long to reach me. Regardless, I'll try replacing the thermal paste this week, using some Youtube videos I found as guidance.

I understand that I need to cover the entire die. I've seen some people place enough compound in a dot that they know it will cover the entire thing when the heatsink is replaced, and I've seen others apply the compound and then smooth it out with the provided plastic tool before reassembling the card with the heatsink. I have a set a baseline temperatures for certain games now that I've been monitoring so I'll use that as a gauge for whether the new compound makes any difference.

I've been able to run games for extended periods of time as long as the temps don't start reaching around 78-82C. That should be well within the parameters of the current card. I'm using AMD Radeon / Adrenalin software to monitor the temperatures and performance. I've wanted to adjust the fan speed via the graph that others have appear when they enable the Advanced options so that I can keep the card at a cooler temp, but I'm not seeing anything like that available in my version (which appears to be up-to-date). Do you happen to know if they removed that option, or is it limited to certain graphics cards?

[goodplay]

Fan control through wattman/tuning settings never worked on the R9 390.

Use afterburner, disable wattman/tuning (to avoid settings conflicts).

I used to just let the card run from its vbios control (only used msi-ab on certain demanding games).

[thecyclegeek]

So, I removed my card and replaced the thermal paste. The thermal paste spread from before seemed alright but I removed it with Arctic Clean and replaced it with thermal grizzly anyway.

Idle temps are about the same, but some 3D Mark benchmarks and gaming temps are on average 3-4C lower than before. 30 mins of gaming hasn't caused any crashes, whereas 2-3 mins in the same game before would've. I've re-enabled Hardware Acceleration in a number of applications and I'm going to push the card over the next several days to see if I get the black screens again. I'll post here with the results.

I also picked up a 5600 XT when one became available, so I'll be trying that after I gauge if this card is still problematic or not.

Is DDU necessary when installing another AMD card?

[thecyclegeek]

No black screens and no crashes after using the card for 2-3 days. I didn't touch any of the default settings since I was worried it might cause some instability. Idle temps were 60-61C, whereas graphic intensive games brought it up to around 75-79C. Maybe someone can comment if they think that temp is normal for this card. I compared with the side panel of my case off and with it on, but it didn't make much of a difference (maybe +/-1C), so case airflow doesn't seem to be an issue.

[goodplay]

Load temp looks good, but idle is about 20C over.

Check your memory clock load (on idle) with gpu-z (sensors).