PC Drivers & Software

[RPX100]

I am creating this topic after being active within this community since december 2020.
The topic is meant to gather crash types and reasons in one place - with possible solutions.
My own RX 5000 series card had a lot of problems and I was able to resolve most of them.
(check my profile for most recent posts to see all the info in detail)

instability on the Radeon 5000 series can be caused by:
- GPU core clock too high (Radeon Software defaults to very high boost clocks!)
- GPU mem clock fluctuations (switching between DPM power saving states: driver bug)
- GPU mem clock too high (overlocking)
- GPU core voltage too low (or clock too high for that voltage)
- GPU hot spot temp >90°C (bad cooling, fan speed too low)
- temp delta between GPU core and hot spot >= 20°C (thermal paste dried out)
- wrong display cable standard (you need: HDMI 2.0b/display port 1.4)
- Radeon Software custom game profiles (random system crash after closing the game)
- general usage and auto overclock info for RX 5700 XT (link)

crash types and reasons in detail:

crash type #1: APPCRASH to desktop (often with screen flickering)
- app freezes, sometimes followed by a short black screen or screen flicker
- display driver resets itself and you are back on desktop
- Unreal Engine Games will crash with "d3d device lost"
- the screen flicker can sometimes be seen a few moments before the crash happens
-> reason: core clock too high (or core voltage too low for that core clock)
-> solution: increase voltage by +10 mV (repeat until crash is gone)
-> if that does not help, set max boost clock (P3) to manufacturer specs
-> note: stay below 1200 mV while using air cooling!

---

crash type #2: black screen and system freeze with system reboot
- app freezes, system freezes and screen stays black
- system will either restart itself or you have to force reboot (power off)
-> reason #1: core clock too high (overclocking)
-> solution: decrease core frequency by 25 MHz (repeat until crash is gone)
-> reason #2: GPU hot spot temperature above 90°C
-> solution: increase FAN RPM, try to stay below 90°C
-> reason #3: temp delta between GPU core and hot spot >= 20°C
-> solution: dismantle GPU cooler and reapply thermal paste to GPU chip

---

crash type #3: green screen and system freeze with system reboot
- screen goes green instantly, system freezes and restarts itself
-> reason #1: memory clock fluctuating (driver bug, still present as of Jan 2022)
-> check memory clock while gaming and look for clock fluctuations
-> memory clock might be fluctuating between the different power saving states (=bad)
-> workaround: 2 monitors with different resolution / refresh rate (will lock mem clock to max)
-> reason #2: memory clock too high (overclocking)
-> solution: reduce memory OC by 25 MHz (repeat until crash is gone)

---

crash type #4: (generic) screen flickering / driver crashes / system freezes
- possible reason: wrong display cable standard (HDMI/display port)
-> solution: use correct standard for RX 5000 series cards:
-> HDMI 2.0b / display port 1.4

---

crash type #5: random system freeze and crash when using custom game profiles
- system freeze after closing games (crashes when back on desktop, but random)
-> reason: some weird driver behavior when using Radeon Software custom profiles
-> solution: only use the default performance tuning profile
-> or install "driver only" and use MSI Afterburner instead

---

This topic is meant to gather crash types and reasons in one place - with possible solutions.
The list above is by no means complete, but might already be enough to help a lot of people.

[RPX100]

pushing this topic in order to add yet another possible reason for GPU crashes.
... and while I am at it, I will talk about PSU specs and 12V rails.

crash reason: power supply unit (PSU) 12V rail(s) configured in a wrong way.
-> this is most likely happening for PSUs with more than 1 PCIe power connector 
-> (or ATX 2.0/2.1/2.2 standard where each rail has 20A limit)
-> do not worry if your PSU has only 1 PCIe power connector (8pin and 6+2 pin "daisy chained")
-> for PSUs with only 1 PCIe power connector the Amps (A) for that rail is the important spec
-> If you are using ATX 2.0/2.1/2.2 PSU then you should use both PCIe connector cables

crash behavior: black screen and system freeze with system reboot
- system will either restart itself or you have to force reboot (power off)

general PSU 12V rail info:
- A rail is a wire/path inside the PSU that carries electricity of a certain voltage
- The 12V rail provides power for GPU and CPU
- a 12V rail can be split into multiple 12V rails (each providing 50% of the rated wattage)
- source: https://en.wikipedia.org/wiki/Power_supply_unit_(computer)

common misconception and useful information:
- "you always need 1 cable for each 8pin PCIe connector on your GPU"
-> this is only true for PSUs with ATX 2.0, 2.1 and 2.2 standard (12V rails are split into max 20A rails)
-> the important part is to know how many Amps (A) your specific PSU model can handle on a single 12V rail
-> ATX 2.3 (or 2.4) can handle more than 20A on a single 12V cable (allowing daisy chained 8pin & 6+2pin cables)
-> if your PSU has the daisy chained 8pin & 6+2pin cable you can check how many Amps (A) it delivers
-> RX 5700 XT maxes out at 22A/264W (OC variants) - default is 19A/228W (on reference design)

- "you need at least 750W PSU for RX 5000 series cards"
-> yet again this is not true and certainly overkill
-> it is far more important to check the capability of the PSUs 12V rail
-> and also if that PSU is using 1 or 2 PCIe connectors
-> If you are using ATX 2.0/2.1/2.2 PSU then you should use both PCIe connector cables
-> If you have ATX 2.3/2.4 PSU with a single 8pin & 6+2pin cable, then this is perfect too
-> this really depends on how your PSU was designed (check your manual)

conclusion for PSU Amps (A) and Wattage (W):
- RX 5700 XT (max 22A/264W) + powerful CPU (13A/156W) = 35A/420W
- there are a few other 12V devices like fans (around 1.5 to 2.0W per fan) = +10W (5 fans)
- then you also need to consider that you do not want your PSU to run at 100% load 
-> "80 PLUS" standard PSUs are most efficient at between half and three-quarters load
-> much less efficient at low load, and somewhat less efficient at maximum load

So we take the 430W (power consumption limit) and devide that by 0.75 and 0.6:
-> optimal PSU for RX 5700 XT with powerful CPU is in the range of 570-720W

another example for Ryzen 5 5600X (65W): 264W (GPU) + 65W (CPU) + 10W = 339W
-> optimal PSU for RX 5700 XT with Ryzen 5 5600X: 450-565W

[RPX100]

pushing this topic after yet another issue that I had (with driver 22.1.2) yesterday:

crash type #6: generic/random game/app crashes without system freeze
- games/apps might instantly close with or without error messages
- I have encountered this event with no "Windows Event Log"-Messages being logged at all
- so this might actually just close a game/app without Windows reporting it as an error
-> reason: this might be related to Windows Update messing up your graphics driver
-> solution: fresh driver install with DDU followed by "sfc /scannow"
-> a clean/fresh driver install fixed this issue for me (4+ hours of testing and this crash is gone)

(before the clean driver install, the game Outriders would randomly close without any event log)

[RPX100]

found another multi monitor issue, that does not affect single monitor setups:

crash type #7: (multi monitor) green screen and system freeze with system reboot
- (fullscreen/3D-gaming screen) goes green instantly (for a short time)
- HDMI/display port audio devices might get a reset/restart
- system might continue to run with audio crackling or audio device delays

-> reason #1: memory clock fluctuating (driver bug, still present as of April 2022) [22.4.1]
-> check memory clock while gaming and look for clock fluctuations
-> memory clock might be fluctuating between the different power saving states (=bad)

-> reason #2: memory clock too high (overclocking memory above 1750 MHz)
-> solution: reduce memory OC by 25 MHz (repeat until crash is gone)
- it is worth noting that 1800 MHz memory clock can work with 1 screen connected
- but using a dual (or triple) monitor setup can lead to this crash (or general instability)
- crash can be triggered even more easily with different screen resolutions or refresh rates

So if you are overclocking your RX 5000 series memory modules - be warned! ^^

[kazeookami]

I have another crash I am getting fairly often that doesn't seem to be here.

Computer blackscreens, but never recovers. Have to do a hard reboot. Then when I come back up/online only one monitor works but resolution control is still in place which suggests there are drivers. AMD software no longer launches successfully as it says the drivers are not compatible. I then have to manually remove the drivers that are in place by using a tool named Display Driver Uninstaller because there seem to be some busted drivers in place still that blocks any AMD installer from installing new drivers. After wiping out the drivers I can then install new drivers and get back to work as usual. 

I don't meant to dirty up your library here but I couldn't find a means to send you a private message.

[Dsmity00]

This is the current issue I am having.  Have you found a solution? I first suspect my power supply because it’s only 550w