"""According to Gigabyte, the AMD Ryzen 9 3950X was stable at 4.3 GHz with a voltage supply of 1.416 Volts. AMD advises an overclocking voltage of 1.3-1.45V for their AMD Ryzen 3000 series CPUs. The temperatures at maximum load under Prime 95 reached 101C (hottest junction temperature). The X570 chipset temperature also hit 68C which shows why the extra fan cooling is needed. When it comes to performance, at stock, the Ryzen 9 3950X posted a score of 3932 points in Cinebench. Do note that the older R15 version was used for evaluating the performance.
Gigabyte also managed to make the CPU run at 4.4 GHz, successfully running Cinebench R15 with no issues at all. With an additional 100 MHz bump across all cores, the chip managed a score of 4475 points which is simply amazing considering it's being done without the use of LN2. Voltage supply of 1.452V was used which is just past the safe boundary but Gigabyte also mentions that not all chips would be able to hit 4.4 GHz putting the 4.3 GHz overclock as the sweet-spot for the Ryzen 9 3950X 16 core processor.
Source: wccftech_com"""
Your processor appeals to professional users. Give your processor back and get a lower processor instead. The vcore value of a processor drops with the Prime95 test. For this reason, the BIOS manual vcore value should be higher than it. The problem with your system is that your motherboard cannot turbo your processor properly. If you cannot do this with a BIOS change, the best thing is to switch to a different processor. Because the processor does not appeal to you. Because I don't recommend OC to you actually. Actually I'm trying to explain to you how to find the correct settings.
Your English is fine for whatever my opinion is worth, with my horse blinkers on and all 
I don't claim to be enough of a creator or enthusiast for the chip I have, but I do enough video encoding and CPU rendering to still find this chip appealing. If it came down to upgrading the motherboard I'd be fine with that, but I have about the highest end b550 board you can get so there should be no reason for it not to turbo properly. We seemed to agree before that the mobo was unlikely to be the culprit unless it was the type of issue which could be fixed with a BIOS update.
I do actually appreciate your attempts to help me/everyone else. As snark-less-ly as possible here... don't you think that if you are trying to find these most accurate settings for someone's processor and cooler then it might be a good idea to ask which cooler they are using first? The section of the article you keep quoting only tells me they didn't use LN2 and they hit 101C peak junction temps. It doesn't mention relevant things such as the cooling solution that was used, or which settings were used/instruction sets enabled in Prime95. It's also pretty dated, written almost a year ago, and the concerns about degradation at those voltages are more recent as far as I knew.
There are steps I was still hoping to try before resorting to an RMA. But if I went through the RMA process and it didn't fix the issue then I would rule out the CPU (or at least worry significantly less about it) and continue troubleshooting the issue. I'm sure AMD would prefer amateurs not try to manually fix stuff by applying more volts before starting an RMA, don't you think? As nicely as possible here, your last reply leaves me wondering if you even understand the issue described, as there is no instability apparently induced by Prime95. Your last two replies seem to contradict one another with respect to what you actually recommend. By trying to guess at the correct settings with insufficient information, you reveal that you are not an engineer.
Let me try, one more time, to summarize the issue for you. Some Ryzen systems end up crashing with "Event 18: Machine Check Exception; Cache Hierarchy Error" in event viewer. They seem to do this only when dealing with some formerly Intel-specific instruction sets (or power viruses, to us amateurs) such as in this recently released game I'm linking to an article about to show you how citations work: Marvel's Avengers Is the First AAA Game to Feature Intel Specific Optimizations on PC. If you check the date, you'll see it is much later than the date of your linked article 
What is odd is that the issue doesn't seem to be provoked by Prime95 Small FFTs with AVX/AVX2 enabled, even though AVX2 seems to be the relevant instruction set.
It's not just that one program either; it's a host of applications, which seem to only have in common that they use AVX2 instructions. It's almost as though AMD has claimed to support the feature before understanding how to properly or stably implement it. At least it might appear that way, to an ignorant amateur such as myself. Are Intel's engineers just more knowledgeable about their own technologies, or are AMD's engineers simply not as good as Intel's in your opinion? It's difficult to say without your expert guidance, for an ignorant amateur like me. Is that why laypeople tend to buy Intel products in your experience, because of reliability? If you ask about this stuff on the AMD forums, it turns out the response you get is a condescending lecture in broken English which doesn't actually resolve your issue. Is that by design?
So to answer your question... if a BIOS update doesn't fix it, and a CPU RMA doesn't fix it, and a motherboard RMA doesn't fix it, and manually tuning down all the clocks and voltages across 4 CCXs myself doesn't fix it... then I might start to think that AMD shares your philosophy on engineering. The issue is pretty infrequent as it is though and a BIOS update seems like it might have helped the OP already, so I'm betting AMD is just working out some kinks when it comes to supporting these instruction sets. You certainly didn't seem to be very knowledgeable about the distinction between the X570 and B550 chipsets, so I have to assume AMD's engineers know more than you do. Maybe you would find it more appealing to troubleshoot older computers, they are probably more your speed.
I am always happy to help a hopeful English speaker aspire to articulate himself, but then let's not have any illusions about who is assisting who here.
