[SOLVED] How can I disable L3 and L4 CPU Cache?

[GameCrucnh]

So my friends got a faulty CPU and we narrowed it down to being the cache, it causes random game crashes and BSODs.

I'm wondering if there is a way we can disable some of the CPU cache's to be more stable. I tried looking online but all of it was pretty advanced and only seemed to be on a per program basis.

He has the 2600x and a B450 if that helps.

 

[rcald2000]

GameCrucnh I don't know the answer to your question. However, you should simply submit your CPU to AMD for a warranty replacement. I've done so myself in the past, therefore I know it can be done. Good luck.

Addendum: The odds of a brand new CPU having a design defect is extremely low. I recommend that you create another message and title, seeking help with troubleshooting your problem. For example, have you run DDU and reinstalled the graphics drivers? If you create a new message thread, I would also recommend listing your friend's complete specs.

 

[hotaru.hino]

The 2600X has no L4 cache and there's no way to disable L3 because it's integral to the processor. Besides, if you disable it, you'll just lose performance and it won't even be worth having the processor.

 

[Karadjgne]

You cannot disable cache. Doesn't work that way.

My question is just how did you come to the conclusion that L3 cache was the issue in the first place and not know it cannot be disabled?

What was the issue in the first place?

 

[DSzymborski]

By what process did you make the determination that the CPU was the problem in the first place? Defective CPUs without some kind of mistreatment is rarer than hen's teeth. A faulty CPU simply causing BSODs would be an order of magnitude rarer. It's far, far more likely the culprit is something else, more typically memory, software, or another hardware issue.

 

[GameCrucnh]

By what process did you make the determination that the CPU was the problem in the first place? Defective CPUs without some kind of mistreatment is rarer than hen's teeth. A faulty CPU simply causing BSODs would be an order of magnitude rarer. It's far, far more likely the culprit is something else, more typically memory, software, or another hardware issue.

He replaced his ram, GPU and PSU and all the issues still occurred.

I came to the conclusion that its a faulty CPU Cache since I got him to download and install Aida64 and run all the tests one by one, within 9 - 40 seconds the CPU Cache one gave a critical error and none of the other ones did.

It causes a lot more than generic BSoDs, He will get BSoDs sometimes but usually whatever program he is running will just end up crashing and not giving any error code or indication of it crashing. I've got him to look through event viewer and there wasn't really anything there

 

[hotaru.hino]

If the CPU is really faulty, then:

• I don't think you should trust the CPU anyway.
• Again, disabling cache if you could will cripple performance of the CPU since by design, the CPU uses that cache.

If it's still under warranty or the merchant return period, I highly recommend just getting it replaced.

 

[Karadjgne]

^ exactly. You have no idea exactly where in the cache there is a defect, if any. If it's in the first 64k, you could disable almost the entire cache and still not catch it, in affect forcing any software to use that error more frequently.

Cache is a deposit. The cpu demands data from the ram, which is data from the storage. The cpu holds that data in cache, queued up to be next through the cores. If the original data is corrupted, it's going to go all the way through everything before ending up in the cache and pulling an error when it hits the cores.

The error could originate with the controller in your ssd, the downloaded files, anywhere. It's only when in use by the cpu it's being tagged.

Throw something out of a plane. Nothing cares what it is, it has no bearing on the surroundings, the air doesn't care, birds, bees, ppl walking by, nothing. Until it lands. Then it makes a massive difference whether it was a bucket of water, a feather, a beachball or a brick.

Aida says there's an error in the cache. It can't tell the difference between a code error that originated from your downloads or a physical malfunction in the cpu.
https://www.windowscentral.com/how-use-dism-command-line-utility-repair-windows-10-image?amp
I'd start here.

 

[USAFRet]

So my friends got a faulty CPU and we narrowed it down to being the crash cache

The problem lies elsewhere.

 

[rcald2000]

@GameCrucnh As I've already mentioned, I can't be 100% certain but I still believe that it's highly unlikely to be a CPU defect, although not absolutely impossible. Were I troubleshooting your friend's problem, I would start off by googling "Aida64" + <the exact error given>. also post error in AIDA64 forum in addition to here (https://forums.aida64.com/). I suspect that the error is the result, not cause, of another issue. Again, I would have made a message board thread listing all specs including cpu fan and power supply's mfg & model #, I would notate the current BIOS version and date and then update the BIOS, reset BIOS to default settings, remove ram overclocking while diagnosing problem, install firmware directly from AMD, monitor and notate temps while running AIDA64. Personally I would go as far as swapping in another SSD or HDD for testing, and install a clean copy of Windows 10 without a license key, since this is only for testing purposes. I know that what I propose sounds like a lot, but it can all be done relatively quickly. Again, if you're 100% certain that it's a manufacturing defect in the CPU, then submit a warranty claim, and then you could always continue with troubleshooting while waiting for AMD to make a determination of whether the claim is approved or not: https://www.amd.com/en/support/kb/warranty-information/rma-form

Let us know what you decide and how you plan to proceed. Whatever the ultimate solution is, I'd appreciate you sharing it here. Thank you.

 

[drea.drechsler]

Just like @rcald2000 recommends: make sure BIOS is up-to-date for the processor, reset CMOS and operate at default settings for both CMOS and memory and test again. Only then will you have a valid claim on warranty with AMD if you get a repeat of the problems. And if you do continue to have the problems, then that's definitely what you should do to get it fixed.

I suspect your friend may be overclocking and you're experiencing this problem, so your thinking you might just disable cache to continue overclocking. As previously noted, even if you could do that any benefit from the overclock is utterly decimated by the loss of the cache anyway.

 

[GameCrucnh]

@GameCrucnh As I've already mentioned, I can't be 100% certain but I still believe that it's highly unlikely to be a CPU defect, although not absolutely impossible. Were I troubleshooting your friend's problem, I would start off by googling "Aida64" + <the exact error given>. also post error in AIDA64 forum in addition to here (https://forums.aida64.com/). I suspect that the error is the result, not cause, of another issue. Again, I would have made a message board thread listing all specs including cpu fan and power supply's mfg & model #, I would notate the current BIOS version and date and then update the BIOS, reset BIOS to default settings, remove ram overclocking while diagnosing problem, install firmware directly from AMD, monitor and notate temps while running AIDA64. Personally I would go as far as swapping in another SSD or HDD for testing, and install a clean copy of Windows 10 without a license key, since this is only for testing purposes. I know that what I propose sounds like a lot, but it can all be done relatively quickly. Again, if you're 100% certain that it's a manufacturing defect in the CPU, then submit a warranty claim, and then you could always continue with troubleshooting while waiting for AMD to make a determination of whether the claim is approved or not: https://www.amd.com/en/support/kb/warranty-information/rma-form

Let us know what you decide and how you plan to proceed. Whatever the ultimate solution is, I'd appreciate you sharing it here. Thank you.

He is 100% not overclocking. When we thought it was a ram issue he underclocked his ram but now he got new ram sticks and the same problem occurs.

He has reset his Cmos before, I doubt its a Power supply issue as he fully replaced the power supply and the error still happened.
Aida64 did not give an error other than "Hardware Failure detected"

He has another backup hard-drive, i'll tell him to install windows on that and run a test as-wel as fully reset windows.

 

[drea.drechsler]

He is 100% not overclocking. When we thought it was a ram issue he underclocked his ram but now he got new ram sticks and the same problem occurs.

He has reset his Cmos before, I doubt its a Power supply issue as he fully replaced the power supply and the error still happened.
Aida64 did not give an error other than "Hardware Failure detected"

He has another backup hard-drive, i'll tell him to install windows on that and run a test as-wel as fully reset windows.

https://www.amd.com/en/support/kb/warranty-information/rma-form

I really do not think a hard drive could bring on CPU cache failures. You've ruled out overclocking with CMOS reset and default setup and ruled out memory and PSU with swap-outs. The only thing is the entire motherboard if you have one available. Then start a warranty claim with AMD...link above.

 

[gamerk316]

He replaced his ram, GPU and PSU and all the issues still occurred.

...

It causes a lot more than generic BSoDs, He will get BSoDs sometimes but usually whatever program he is running will just end up crashing and not giving any error code or indication of it crashing. I've got him to look through event viewer and there wasn't really anything there

I note the Motherboard was not replaced.

The random crashing of application programs is typical of some sort of memory fault, and I suspect that's what AIDA64 caught.

Given RAM was already replaced, I'd suspect an error in one of the physical RAM slots on the motherboard. This type of fault does crop up from time to time, although it is fairly rare.

As a test, try running with only one stick in one physical slot and see if any errors pop up. If so, try a different slot and see if they go away.