I specifically wanted to make a separate post on the problem of the black screen of AMD video cards.
For Ryzen processors, there are 2 parameters that generally determine the operation of the processor itself.
The first parameter is cpu voltage, the second parameter is cpu soc voltage. On the one hand, the processor speed depends on the cpu voltage value (the higher the voltage, the greater the speed), but on the other hand, the degradation of the CPU depends on the same voltage (the more voltage is applied, the faster the degradation).
The second parameter is even more important than the first. The cpu soc voltage determines the stable operation of the processor with the motherboard and other computer components, including work with the video card, memory, etc. And it is very important that the motherboard adjust this voltage automatically, because. the stability of the PC depends on the exact value of this voltage.
You ask, what is the relationship between the black screens of AMD video cards and cpu soc voltage? By slightly increasing the cpu soc voltage, we get a more stable operation of the memory controller of the Ryzen processor, BUT at the same time we worsen the performance of the pci-e bus, especially if the pci-e version is 4.0. And vice versa, slightly lowering the cpu soc voltage, we get a slightly more stable operation of the pci-e bus devices, but at the same time worsen the performance of the processor memory controller. With the cpu soc voltage value, we need to find a middle ground, the voltage value at which both the memory controller and pci-e bus devices will work stably. And it is for this reason that the motherboard must adjust this voltage automatically.
By default, cpu soc voltage for all Ryzen processors is 1.1 V. It seems that we get that with this value of this voltage, both the processor memory controller and the pci-e bus should work stably. But we use different RAM each (different memory frequencies, voltages, memory timings, etc.), so the motherboard tries to change the value of this voltage in each case so as to get the most stable PC performance when using our memory sticks. The motherboard tries, but it does not always succeed. And it happens that the cpu soc voltage, issued by the motherboard in automatic mode, is underestimated and we get constant errors in the memory controller, and therefore the RAM itself, and as a result, we get the notorious black screens (but without crashing the video driver). Or vice versa, the cpu soc voltage output by the motherboard in automatic mode is too high and we get pci-e bus errors and, as a result, the same notorious black screens, but with a video driver crash.
As you can see, two different categories of errors lead to the same result - a black screen.
Therefore, it is important that when receiving black screen errors, it is always necessary to analyze the cpu soc voltage value, this is how we will understand in which direction we should move to eliminate errors.
Ask how hardware acceleration affects the stability of work here? With the help of hardware acceleration, part of the work being done is shifted from the processor to the video card. And if the pci-e bus is already unstable due to the cpu soc voltage value, then disabling hardware acceleration will return this work being done back to the processor (hence there will be fewer errors on the video card) and we will get a slightly more stable PC operation (just a little more stable than completely stable). And how to get a completely stable PC? To do this, you need to get the exact value of cpu soc voltage, at which both the memory controller (and hence our memory sticks) and the pci-e bus (and hence our video card) will work stably. And how to get a more accurate cpu soc voltage value in automatic mode? This is determined precisely by the BIOS, as motherboard manufacturers taught the motherboard using AGESA to work with a specific processor. Therefore, in these black screens, the fault is not only AMD, but also motherboard manufacturers, because. they didn't fully teach the motherboard to work stably with a certain processor when using different RAM.
PS. It is really possible to manually select the cpu soc voltage value at which the computer will work stably, but this is very long and dreary.
This seems to be solid info. I don't tinker too much in the BIOS, other than enabling resizable bar and XMP, but this could maybe be something to look at if I ever have the black screen issues people are reporting.
It sure is solid info.
Coincidentally I messed with SOC voltage a while ago when I had some stresses with my memory. I had some freezes while at 3200Mhz. Tried everything including SOC voltage to keep the memory stable.
But I never thought that it would impact PCI-E bus.
I currently have a tad over on the SOC voltage to keep the memory happy, had to loose the timings a bit and lower memory voltage too.
Curious thing is, I had 3 black screens in the last 3 months, not even critical at all but it could be the SOC voltage, so I'll give it a try.
I do tinker a lot in the BIOS to squeeze out the max performance and to lower the temps (and fan noise) as much as I can, but this often leads me to some rare crashes once in a while. Oh well... ¯\_(ツ)_/¯
I also forgot to point out that it happens that during the transition from the loaded state of the processor to the state of low load or even idle, sharp jumps in cpu voltage and cpu soc voltage occur, which go beyond the standard short-term voltage deviations equal to 350 mV and accepted according to by default in BIOS (in practice, this is more applicable only to cpu voltage, there are no such abrupt changes for cpu soc voltage). In this case, this protection will work out and the PC will simply turn off (restart). There are many reasons for such a sudden change in voltage, ranging from the settings of the BIOS itself to the VRM circuits of the motherboard and computer power supply. To remedy this situation a little, we can change this limit to 400 mV, which is intended for use in overclocking conditions.
I had this annoying problem for a month and i thought my card died then i looked to reddit, amd forums, some tech forums and realized so many amd users having this problem so i tried solutions from these topics.
Nothing worked tbh. My card is rx570 and here are my tricks: Uninstall drivers with DDU and install latest WHQL driver(common knowledge) Then undervolt your card(i undervolted from 1150 to 1050) Here is the "key" solution: Before you launch your game use amd chill and set your max fps as your monitor refresh rate. Because i realized the "fake fps" at game launch (about 500fps) caused crashes.
After game launched im disabling chill because i prefer in-game limiter. And i think you guys should also fps limiter atleast untill AMD fix these problems.
You guys also try windowed fullscreen or lowering your resolution. But dont forget to end explorer.exe task while using windowed fullscreen.