Drivers & Software

[cristi_trohin]

So I've bought this MSI RX 5600 XT Mech OC since November 2020. Since then, there hasn't been a single day without a crash. It's **bleep** near impossible to play anything on this thing. I've send it to warranty, but the service tested the card with Heaven, which is not crashing since it doesn't have a typical gaming load. So I got the card back with the status "Reported defect not manifesting". I've tried every driver of 2020, and the last two of 2021, but the crashes still happen. It has days when it crashes 1-2 times, and days when it crashes constantly. I've reinstalled Windows 10 almost every week.

My configuration is this:

- CPU Intel Core i9 9900K cooled by a Coolermaster Hyper 212 EVO

- MOBO Gigabyte Z390 UD

- Storage: Micron 240GB M.2 SATA SSD, WD 12TB HDD

- PSU Silentium PC 700W (had a Seasonic 500W PSU before, but changed it thinking it could be a power issue)

- 48 GB RAM (1x32GB Samsung 2666 CL19, 1x16 GB Samsung 2666 CL19)

Nothing is overclocked in my system. I disabled all "multicore enhancements", set the PL limits for the i9 to 95W/119W.

The only thing I cannot fix is the incompetence of the AMD driver team. That is the only logical conclusion. If the driver team is not at fault here, then the hardware is bad and AMD is opened to class action lawsuits for intentionally selling bad hardware.

[JirayD]

This sounds an awful lot as if the PC is getting insufficient voltage from the grid, especially the differences between days. Are you plugged into a power strip? Try connecting your PC directly into an outlet, or if possible into a different circuit in your house, closer to the circuit breaker.

[RPX100]

Some more information would be useful:

What are your settings within Radeon Software?
- which profile did you select? (standard / gaming / e-sports / energy-saver)
- auto performance tuning or manual? (post your actual clocks and voltage)
- GPU temps while under load? (core / mem / VRM / hot spot)

---

Have a look at this topic (link)
- info is regarding RX 5700 XT, but general tips should apply to 5600 XT, too.
- just make sure to use actual manufacturer-clocks and voltages

[cristi_trohin]

I didn't select any profile (and I assume that would default to Standard). Tuning control is on Automatic, and Auto Tuning is on default. No overclock, no undervolt, nothing.

The Mobo bios is updated to the latest version, so is the GPU bios (via MSI live update). I have 4 intake 120mm fans in the case, and 2 120mm exhaust fans, so airflow is plenty.

While running Furmark (after 20min), the stats are:

GPU Clock around 1540

VRAM clock 1746

GPU voltage tops out at 918mV and hovers around 878mV

Power Consumption tops out at 136W

GPU Temp 76C

GPU Temp Junction 86C

GPU memory junction 96C

GPU VRM 78C

GPU core current 106A

GPU memory current 16A

[cristi_trohin]

After reading your post in the other topic, I noticed the P3 clock was set to 1780, so I set it to 1620 manually (the value in the card spec) and lowered the voltage a bit to 960mV. Now the clock during Furmark hovers around 1440.

However, the temps are the same, but I haven't change the fan profile yet. Once I set an aggressive fan curve (with 100% fan at 75C) the GPU temp dropped to 73C, GPU Junction to 84C, GPU mem junction to 90C, GPU VRM to 67C, while the other temps remained the same as in my previous post.

[delevero]

I did not read all of the above, but here is a tip that you can try to make your pc stable.

Go into bios and turn on advanced mode ( maybe press F2 to turn it on )
find the location where it say  "CPU Vcore"
Change the setting AUTO to 1,4V

Personaly my computer with a 5900x and a 6900xt have been crashing in every game for 2½ month now and i tired 3 different bios and all avaliable drivers but still it crash.
But yesterday I manual enter the voltage for cpu Vcore to 1,4v and after that I have had 0 crashes.
Example i play hitman 3 and it crash in the "japan" mission where we run on a bridge in the start after 5 minutes every time.
But now i can play the mision without problems as well as other games.
Give it a try.

ps.. 1,4v will not damage your cpu  Ryzen goes up and down the volt range all the time when on auto from xx to example 1,5v
Try with 1,4v and then maybe try and lower it if your pc get stable.  ( maybe to 1,35-1,4v )
pps..: its my impression that the automatic voltage software in bios that control the power to the cpu maybe have a bug.
Allot of people on the forum have issues with stability and windows error  Power-kernel  ID 41 ( 63 ) and my personal guess is the bios provide to little power in some situations = shut down/restart the pc due to little power at x critical time.
The funny thing is in stabilty tests my own pc was 100% stable, but in games it crash after example 5 minutes.
Maybe it have something to do with when you play a game that is not so demanding that the single core clocks get to little power or something like that and that trigger instabilty.. But that is just my guess.
Good luck and if it help you then help spread the word so other can get a stable pc

[cristi_trohin]

@delevero My CPU is an i9 9900K, and I had no trouble with it so far. I run virtual machines with software builds every day, and I've stressed the living crap out of it without any issues. I doubt that is my problem. But what I can tell you is that if in your case raising vcore worked, it means either AMD's QA is slacking off (allowing a CPU to be in the 5900X batch without being stable at the spec vcore) or the AMD AGESA is messed up. It seems that software is AMD's weakest point, and it reaches every department from CPU microcode to GPU drivers.

[cristi_trohin]

@RPX100 I set the clocks manually, as shown in the post you shared but the problem remains... Still getting crashes every half hour. 

[RPX100]

You have already tried what has worked for my own RX 5700 XT.
So there is not much more that I can offer to help, since this might be another issue.

If you want to try one more step - then I suggest the community favorite:
-> use driver 20.8.3 without Radeon Software (Download from here)

1) uninstall current GPU driver with DisplayDriverUninstaller in Windows Safe Mode.
-> it is important to run DDU as Administrator and in Safe Mode
-> unplug your LAN cable during the entire uninstall process (to block windows updates)
-> restart your system after uninstall is done

2) run the 20.8.3 setup but only do the first part, where it extracts the files to "C:\AMD"
-> when the Radeon Software installer open, just close it (cancel the install process)

3) then open Windows Device Manager
-> click on graphics card/adapters
-> there should be "basic display adapter" (or something like that)
-> right click on it and choose "update drivers"
-> follow the steps to manually select a driver from "C:\AMD" folder

restart your system after install is done
-> you may now reconnect your LAN cable
-> use MSI Afterburner if you want to change fan curve/voltages

[CozaMcCoza]

- 48 GB RAM (1x32GB Samsung 2666 CL19, 1x16 GB Samsung 2666 CL19)

Have you tried removing the 1x16 stick?

I've had so many headaches by not having exact matching RAM on a system.  It's caused me many system stability issues in the past that I will no longer ever consider using another RAM stick that isn't the same model.

[lph]

You could try underclocking like I did in my thread
https://community.amd.com/t5/drivers-software/i-finally-fixed-my-5600-xt-crashes/td-p/294476

[delevero]

try this

1). go into bios
2). select advanced mode
3). find something where it say   CPU Vcore
4). Change Auto to  1,375v
5). save and restart your pc
Try and see if that made your pc stable.

( i have a 6900xt and 5900x cpu and i have had crahes since mid december and could only play 5-15-30 minutes and then computer would turn itself off.. After change the above setting I had 0 crashes.  Im no expert but i have a feeling that either my cpu is defective or the is a bug in bios or amd chipset driver that make the computers crash... allot of people have power related issues ).

[cristi_trohin]

@lph When I was using the latest bios with 14000MHz memory, I tried underclocking, but the crashes kept going. I think it was because of the memory clock. Unfortunately, you can only underclock the core, not the VRAM, hence why I flashed the original 12000MHz bios on the card, and now it seems to work. Until the next crash, probably.

[cristi_trohin]

@CozaMcCoza I've tried with 2x16GB Micron sticks, same timings, same freq (actually the exact same batch out of the factory) and the issues persisted.

Last evening I managed play without crashing for about 2 hours, and here is what I did:

I got the initial bios for the RX 5600 XT MECH OC card, from Techpowerup. This is the bios where the VRAM clock was set at 12000MHz. Then I set the P3 value to 1560MHz in Adrenalin. This seems to be more stable than it was with the newest bios (with VRAM at 14000MHz and core at 1640MHz).

The problem is that when I bought the card, it had the latest bios on it, with 14000MHz VRAM, but I've heard that some MSI officials were saying that not all cards can work at 14000MHz. Why? It seems that MSI is not sourcing the memory chips on their own, they're buying them from AMD, and were put under pressure by AMD to release the cards with the 14000MHz memory clock, even though they were not able to validate the cards at those speeds.

Anyways, I hope this piece of AMD **bleep** won't crash anymore, and I can definitely say that this is the last card I ever buy from AMD. I'd rather pay 20-30$ extra to get a stable NVidia card, than to be AMD's guinea pig for 30$ less spent.

[cristi_trohin]

I have my computer plugged into an UPS, so I doubt there is insufficient voltage. I also tried this card in my son's PC, and it had the exact same behavior, and he's plugged into a different circuit, straight into the wall, with ground pin as well. My previous PSU(Seasonic 500W) as well as my current one (Silentium PC 700W) never cut the power to my system under load. I've even tried an experiment... When a game would crash 2 times within 5 minutes, I'd load up Cinebench R20 and Furmark at the same time, and let it stress the living sh*t out of my system, but it never crashed. Once I stopped the tests and went back into the game, it crashed again. I think no game would ever be able to stress the PSU as much as R20 + Furmark could.

[nathaneddy2004]

I literally just managed to fix this today on my XFX card with 14gbps flashed drivers I moved my voltage slider way back. It went from about 980mv at 1750 mhz on core. With an actual of about 1700mhz. I pulled the clocks back to 1700mhz with an actual of 1650. I then pulled my voltage all the way back to 880mv. Instantly got 20% fps buff in games and hasn't crashed once so far. I believe there is a voltage regulator issue on the flashed cards. Try it out and see if it works.