Showing results for 
Search instead for 
Did you mean: 


Journeyman III

WHEA: Cache Hierarchy Error Processor Core

I haven't had any issues until today when 2 of these errors happened. One APIC ID 9, the other APIC ID 1. mI was playing the new Marvels Avengers game, which is why I think it is tied to it. But anyways, the game randomly crashed and my entire PC went down. Attached is the files in event viewer. The only thing I did recently is upgrade from 16gb of 2400Mhz ram to 32Gb of 3200Mhz ram. I stress tested the machine and it was fine. I also played all day today and it was fin until I played this game. I am using A-XMP profile 2 to get the ram to run at 3200Mhz.

As of writing this I am testing it again with prime95, passing every test so far, its been going for 45 minutes now. HwInfo during the stress testing does not seem to indicate to me that something is wrong.

Sorry I been busy with class and have not go around to replying. I am unable to reproduce the issue at all. Like I said I ran prime95 over night and nothing. I went right to a 3 hour gaming session after that of Marvels Avengers, nothing. It was a one time thing. So I am at a loss on what to do and a tad worried.
I did update the bios to the latest for my motherboard which is MSI B450 Gaming Plus Max. Maybe the bios update fixed it?
69 Replies

I didn't think anybody thought that having a max boost of 4.7ghz meant it could do 4.7ghz all core lol!


With 4.7Ghz, you can run all cores with a good cooler. But you need to set AVX-AVX2 multiplier. If you try to operate all cores with AVX feature with 4.7Ghz, 300watt power consumption may occur. It is quite difficult for you to cool such a power. 


Do you know where to find this AVX/AVX2 multiplier in AsRock's UEFI?


No. This may not be an option for Ryzen systems.


I didn't think it was, so I had to ask and make sure lol!


If you run an AVX application with manual BIOS without giving a negative multiplier, your system may be


I don't intend on running at anything but stock, as I've said a few times now.  It seems to downclock itself when under heavy enough AVX/AVX2 load, such as in Prime95.  It will just downclock itself to 3.3-3.5ghz in Prime95, but I think in a real world scenario like a game where there is no detection or downclocking of AVX instructions some CPUs fail to meet rated speeds at stock voltages under those circumstances.  If anyone has a processor that seems to not be able to perform up to specification in this way that seems like a perfectly valid reason for an RMA.


If Prime95 decreases the CPU multiplier, then it's fine. No need to give a negative multiplier. Otherwise, your system may run into corruption. This is important. Because if your system tries to run an AVX application at turbo speed, your system may break. Manually editing the BIOS can cause this and I just warned you.


It decreases it automatically, below the base clock, and this process seems to be hidden from the end-user (like you said, there is no user-definable AVX multiplier or offset for Ryzen).  It's almost like AMD designed them not to fail certain stress tests and synthetic benchmarks which use AVX instructions, but that of course doesn't mean that they've implemented this in such a way that every application which uses AVX instructions knows to downclock/apply this hidden offset.  I'm not really worried because everything in my BIOS is on default values, and my philosophy is such that if my game(s) crash(es) enough on stock settings then I complain lol.

As far as warnings go, some of the recommendations about settings you've given I believe to be unsafe.  For example, my RAM needs 1.35v to run at rated speeds, but you said 1.2v.  That's at least wrong if not potentially unsafe, why it's good to check up on the specs of listed components before making recommendations about settings, in case I were the kind of person who didn't understand anything about their RAM lol.  Also, I've seen anywhere from 1.325-1.35v listed as a maximum safe vcore, well below your recommendation of 1.45v.  I'm not really into overclocking so tell me if I'm wrong, but by many peoples accounts you would definitely be risking degradation at that voltage.  If I were trying to manually tune and stability test everything myself, I wouldn't go above 1.325v without an exotic cooling solution.  If the end-user NEEDS to manually tune and stability test everything themselves though, that's a problem... one which can hopefully be fixed with an RMA.


I might have mistyped it because I don't know the part number of your RAMs. You must make the necessary correction for the system part you did not properly specify. What you do also concerns you.


Maybe you shouldn't give recommendations until you get all the specifics you need.


I think you buy a different processor. The 3600 is the right one for you.


I think you should read more recent articles, and maybe ask more probing questions before making recommendations.  It's not really responsible to recommend manually configuring settings in the BIOS to potentially unsafe values before even knowing what hardware someone is using.  This is especially true if you think you are speaking to a novice user, who's system may already be unstable.  It isn't sufficient to say "you warned them" after assuring them that the unsafe values which you recommended that they manually enter are totally fine.

If I were to follow your advice, then I would be better off with a 3600... but only because your advice would end up degrading a much less expensive chip that way.  Asking a few probing questions about what programs I use might better inform your suggestion of processor.  Personally, I'd suggest just avoiding a career in sales if I were you, not to mention tech support.


Considering that amateurs like you will interfere with the BIOSs, the engineers include all kinds of control measures in motherboards. So even if you want, you cannot damage your system with the BIOS changes.

If you have a system that is not stable at the factory settings and you do not apply what I wrote or similar, you will not be able to stabilize your processor.

In addition, the chipset is of great importance for processor stabilization. It is absurd to claim that the b550 is better than the x570 chipset. Even though you write very ignorant things, I make the mistake of trying to help you.



Considering how you act like you think that I'm running my CPU in a unventilated cardboard box, it is pretty remarkable that you think BIOS settings could fix my problem at all...

Sure, I'm an amateur... and that is one of the reasons why I repeatedly stated that I am opting not to go manually "interfere" with settings in the BIOS, which you would know by now were you literate. Being an amateur as I am, I don't know what kinds of control measures are put in place by these engineers.  Nor do I know exactly which voltages they would say are safe with enough heat dissipation, but I do know now that I should not necessarily trust you to accurately relay information from them or anyone else.  That is the extent of what I know right now.

As an amateur, I was wondering if perhaps you could elaborate with your great technical expertise on the vital role of the chipset in CPU stabilization?  You said before "I guess the chipset adjusts the CPU voltage" and this amateur was hoping you'd have some citation for that beyond simply "I guess."  I wouldn't even doubt the claim about the chipset being important the respect you say so much as your ability to explain it at this point. What role would the chipset have in a specifically CPU/RAM stability test like Prime95 for example?  I'm genuinely curious to know, and I'd be happy to admit to being an "ignorant amateur" as I already do if it helps lol!

The claim I actually made (to someone else, which you would know were you literate) related specifically to the power delivery/VRM design of B550 vs X570 boards across comparable models.  What I said was that one would be "better off" with the B550 variant of a given model if all they cared about were those aspects (rather than other features, like PCIe Gen4 through the chipset).  This is slightly, yet crucially, distinct from the absurd claim which you were first to codify; "that the b550 is better than the x570 chipset."  By trying to put those words in my mouth you let them come out of yours, even if only to indignantly negate them.  That was either the weakest strawman I've ever seen or you could desperately use some basic reading comprehension skills.

Either way, I do appreciate you trying to give me (a humble, ignorant amateur) the benefit of your years of formative experience guessing at which voltages to input and bricking hardware.  I can't tell you how helpful it has been to argue with someone for whom English is clearly a second language while trying to troubleshoot my issue.  Goodbye again, I'll await your next snarky reply!  I wonder what downgrade you'll recommend as a solution next... plz not the gpu X'D


In fact, you are helping me with my English practice with your amateurishness that you constantly refer

Old systems had two bridges. In new systems there is only one bridge and the vast majority of controls are on the processors. Chipset contains controls such as USB and PCI.

But a processor cannot perform voltage stabilization with a control on itself. If a processor talks to itself without a chipset, it will make mistakes like you.

So chipsets were very important for older systems. It doesn't matter much anymore with newer systems. But it is not possible for me to tell you what the problem is before you do the necessary tests and return to me. 

Your processor goes up to 4.3Ghz with all its cores. The voltage value is very close to 1.41. This information is enough for me. However, unless you interfere with the BIOS, it will not properly turbo itself to 4.7Ghz and it will crash. The error message you got states this.

Your processor appeals to professional users. My purpose in writing this quote is to address people who understand the BIOS settings well, and understand server systems, and somebody similar. An upcoming update to the BIOS might fix the problem but it may not. It is unreasonable to wait for a BIOS update rather than trying to specialize on your current processor. By personifying the event, you already reveal your character. You are wearing horse blinkers. What will you do if RMA doesn't solve the problem? 

No one can use such systems with OC without trial (your CPU is OC version). You will try it and find the most accurate settings for your processor and cooler. So you will specialize. If that's not your goal, don't write any more. 


"""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.


Both your motherboard and processor are very good. The only problem is that AMD offers too many cores for your processor with very high turbo speeds, and the CPU's OC potential is pretty low when looking at core voltages. When you want to use this processor at higher speeds than its equivalent, you may have to change the motherboard every 3-4 years without losting performance. This is just my interpretation.

Choosing the RMA way may not solve your problems. Even a small BIOS change without going to RMA can make your system very stable. This is also my interpretation.

The best stabilization test for processors is done in 2 ways... First one with AVX, and the second without AVX. When your processor is not using AVX commands, it will try to turbo itself to higher speeds according to processing load. For this reason, it is very important that your motherboard BIOS and chipsets properly adjust the voltage curve. 

You can give the voltage curve with adaptive voltage setting and manual voltage setting and VRM setting and a special multiplier (for adaptive) with the BIOS. And you can make your system stable. This would be obvious if your newly purchased motherboard or processor was faulty.

Thats the problem for you. Your processor will try to instantly turbo itself to 4.7Ghz unless you interfere with the BIOS. And it will need instantaneous high voltage. And the curve of the core voltage has to match this peak. And if your system's BIOS software is not good or the processor cannot speak to the chipset properly, you shouldn't have any problems today. But you may run into problems tomorrow.

I am not suggesting to you. I just wrote what I thought. You should find out what vcore value your processor is stable at what speed and I think you should learn to apply the necessary settings according to this value with the BIOS. If you do not want to mess with the BIOS, it would be better to buy a processor that is not k or x series.

Thanks for the English lesson... Goodbye...


Hey there, here is another poor guy who has the same issue as the post, and I am here to ask what is a cpu rma, cuz I updated the bios and drivers from mobo website and amd website and have ryzen master installed( basically what I am trying to say is that I wanna try everything to make my pc stop crushing while playing games


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.


Please ask this question here... 

I am using Intel. I don't know if an AMD Ryzen processor will tend to execute AVX commands without lowering the processor speed. I don't know your level of computer knowledge, but the problem is that if you guessed it, AMD will need more voltage to make it stable.


mstfbsrn980 wrote:

Please ask this question here... 

I am using Intel. I don't know if an AMD Ryzen processor will tend to execute AVX commands without lowering the processor speed. I don't know your level of computer knowledge, but the problem is that if you guessed it, AMD will need more voltage to make it stable.

AMD Ryzen supports AVX fine, the mix of instructions vary depending on the compiler options for a particular game or app.


I do not think otherwise and I did not share an AMD link for that. 


Getting back to the original problem. I wonder if the OP has been able to stabilize his rig from the BSOD problems.


OP is not getting BSOD. This error was caused by a hardware error that occurred while Windows was running and did not cause BSOD, and one of the CPU cores did not respond for a very short time. It is an error that may be solved with a small change and is not actually a hardware error. The BIOS and the chipset duo fails to raise the voltage for a short time.

In short, I think there is no need for RMA. Because a different part with the same part number may solve the problem today. But the same error may occur in the future due to wearing out the parts. I am trying to explain the exact solution. And I'm trying to explain the reasons. I have been using PCs for very long years and I have been sharing according to my experience.

I have used a PC from the IBM PC 5150 onwards

I'm not attempting to overclock the chip, neither is the OP.  We are both using stock clocks and voltages, at least I am and I don't wish to overclock at all.  It seems like one or more cores is failing to run stock clocks at stock voltages, is the issue.  It could conceivably be an issue with power delivery from the board, but that's unlikely in my case since I have a higher end board.  Like I said, I will try to wait for another BIOS update to see if that resolves the issue, and my BIOS is running with default settings. I really doubt it's an issue with the board, it seems to be turbo-ing normally to me (highest I've seen it boost to in monitoring software is 4616mhz, right where we'd expect with it spec'd @ 4.7ghz).

Also, I think those are pretty high voltages to be running for an everyday overclock.  I certainly don't want to push that much power through my chip anyway.


Yes. I agree the same. It has very high voltage values. I think your system will work very stable by manual adjustment.

If I were you, I would follow these ways and use my system stably.
I'd go into the BIOS and reset is to factory settings

+ I'd go into the VRM settings and "only" set the level by default.
I'd go into the TDP settings and set it to 105.
+ I'd go into the CPU Core Ratio settings and set it to 45.
I'd go into the CPU Core Voltage settings and set it to 1.45.
I'd go into the RAM Speed settings and set it to 3200.
I'd go into the RAM Voltage settings and set it to 1.2.

I'm not sure these settings will work. But you should try to run your system stably with a similar configuration. If the AVX multiplier is adjustable with the BIOS, you can give this multiplier to 4. 

Edit: Not 4 for AVX, 6-7-8 can be tried... Because turbo speeds and core count are very high for you...


Some good news out there. A slew of new BIOS updates appear to be coming out from some of the AMD board partners.

I updated my Asus motherboard last night after having these errors come up left right and centre over the last few days and the good news is that things have settled right down. I'm also able to enable PBO as well as XMP/DOCP. The only way I was able to do this before was downgrading to a B patch BIOS and change a number of my voltages, so this seems like excellent news.