5950X WHEA Cache Hierarchy Error

[Daylight_Invader]

Recently received my brand new 5950X which I swapped out my 3800X for in my Asus I-Gaming X570 motherboard.

Running the board in full defaults mode causes no issues, however if I attempt to use an XMP/DOCP profile, all hell breaks loose (cache hierarchy errors + corrupt files + constant spontaneous reboots), despite the RAM being fine. Strangely running memtest86 for a few hours whilst in XMP/DOCP mode shows no errors, but I am unable to reliable run Windows 10 like this. The reboots only take place when the machine is in relative idle.

Currently going back to defaults mode in order to keep a stable system. I've also stopped PBO, but PBO did not seem to be the triggering factor. For safety however I have turned it back off, although processor seems to happy to spin up a single core to over 5Ghz in some workloads even without PBO enabled.

Currently on the 2816 bios and have a feeling this might be BIOS rather than CPU related.

Any suggestions out there? Putting machine in full default mode and running any stress tests yields no issues whatsoever, but default memory is around 2129 Mhz which does not really do this CPU justice.

Thoughts most welcome, otherwise I guess I will need to do an RMA. Technically my RAM will run to 3600 Mhz, and even putting it back to 3200 does not help. The system just keep rebooting.

 

[Daylight_Invader]

Quick ? Did you reinstall windows after putting in the new CPU? I had very similar issues (crash/reboot when idle) and had to do a full reinstall to get a stable system. Has been ok for over a week now.

TomDD

Absolutely. Changing processor architectures in my mind does mean a full rebuild.

it is indeed bios related - I had the same experience onmy 5950x on X570-E.

You can solve it in two ways:

1: Revert to latest non BETA bios
2: Manual edit voltages in bios as follows:

Cpu Soc Voltage 1.1v
*Vddg iod 950mv
*Vddg ccd 900mv

Rock solid for me since adjusting voltages.

Cheers

C

Not sure I want to go down the non-BETA path as posts on other forums including ASUS, MSI and Reddit all seem to suggest the older BIOS versions also have issues. Basically from what I can gather it is a bit of a mess. I'd prefer to tinker with voltages in the BIOS rather than go backwards, especially as I know we are waiting for a new AGESA code to be included shortly anyhow. My guess is that AMD and the motherboard partners have some work to do. Just surprised the reviewers out there are not finding these issues and it's just people like me complaining on forums.

 

[Daylight_Invader]

it is indeed bios related - I had the same experience onmy 5950x on X570-E.

You can solve it in two ways:

1: Revert to latest non BETA bios
2: Manual edit voltages in bios as follows:

Cpu Soc Voltage 1.1v
*Vddg iod 950mv
*Vddg ccd 900mv

Rock solid for me since adjusting voltages.

Cheers

C

Thanks for the tip. Not seen those exact voltages listed anywhere before. I've put them into the BIOS and the good news is that I have been able to run DOCP/XMP and the machine seems stable. Nothing strange in the logs at all. In fact I am quite tempted to rebuild again just in case as I have not been able to keep the machine anywhere near this level back at auto.

Cheers!

 

[Daylight_Invader]

it is indeed bios related - I had the same experience onmy 5950x on X570-E.

You can solve it in two ways:

1: Revert to latest non BETA bios
2: Manual edit voltages in bios as follows:

Cpu Soc Voltage 1.1v
*Vddg iod 950mv
*Vddg ccd 900mv

Rock solid for me since adjusting voltages.

Cheers

C

FWIW, those settings did not work when setting up windows. They work fine once everything is set up, but need to go back to full defaults for initial build.

If anything, it shows these new 5000 series processors certainly seem to be a little touchy on the voltages.

 

[Daylight_Invader]

Another thing I have noticed is that I get random bad block entries in my System Logs. Strangely if I then do a chkdsk, sfc or look at SMART data, no issues are to be found. Again I suspect this is all down to voltage related issues (ID=7). Keen to see if others out there are seeing these entries too.

I think after this morning's latest crash which was back on full defaults, I am going back to 2802 as a BIOS as this instability is really driving me up the wall. I don't mind limping along for a few weeks on lower settings, but I really hope they sort this out.

 

[Daylight_Invader]

I was getting WHEA errors on my new 5900x also, finally latest beta BIOS sorted that out and all working great now

Great news, but I suspect we have different boards from different manufacturers. Which AGESA version are you on out of interest?

 

[Daylight_Invader]

Still keen to see if others out there are seeing ID 7 errors in their Windows system logs (bad blocks) which seen totally spurious and (thankfully) false. I am running a WD SN850, so not sure if that is a big factor in this however.

 

[Daylight_Invader]

I've also made a few other changes to ensure stability.

I've moved back to the 2802 bios and applied the following with PBO & XMP/DOCP:

SOC = 1.1V
IOD = 0.950V
CCD = 0.950V
DRAM = 1.37V

Seems I have finally achieved some much needed stability (for the moment). Obviously keen to move back to auto defaults with XMP/DOCP + PBO2 when everything settles down. Still watching out for those strange ID7 bad block issues, but will need to wait as these seem to be generated at unexpected moments. I have had one crash, but it was before I enabled XMP and PBO and was with default voltages on the non-beta BIOS, so even the non-beta is not 100% stable.

 

[Daylight_Invader]

AGESA ComboV2 1.1.0.0 D

I was on same version with patch C, but I’ve dropped back to patch B which for the moment appears to be more stable.

 

[Daylight_Invader]

nice cant wait to see this in action

To be fair, I suspect patch D is really just another beta iteration. I think we still are waiting for the next iteration of the AGESA code, and some suggest that we won't see a proper new BIOS until early January 2021. That means all of us need to get a little bit better with going into the BIOS and playing around with some settings ourselves.

 

[Daylight_Invader]

New 3001 NON BETA bios just released for Strix X570-E - interesting to test if it has the WHEA´s solved as well

C

I-gaming too which is almost the same board (but cut down).

 

[Daylight_Invader]

So....

I have installed the new 3001 Asus bios on my ITX X570 Motherboard.

Initial tests are really promising, although please note that this is the strangest update I have ever had on an Asus board. The bios update itself restarted the machine 3 times to complete various update steps (inc the LED lights firmware).

Please note: Once updated, you will need to reapply PBO via Ryzen Master as the settings won't stick even if they were successfully applied before, so please do re-test and apply PBO if that is something you want.

My own testing shows no WHEA Windows Cache Hierarchy faults, and the machine does not do the strange idle reboots like it did before. XMP/DOCP also seems to work. Scores in Cinebench seem very good once PBO is re-applied (scores over 27K consistently assuming you kill most background programs). Boosts have gone up to 5Ghz, but not quite as good as when I set some voltages manually myself where I was closer to 5.1. CPU temps seem to range between low 50s and high 80s with my current cooler setup, but I am upgrading from a 240mm to a 280mm cooling setup in the next few days so hopefully that will help lower these. I believe AMD when they recommend an AIO for these new chips, although I suspect some of the high end air cooling solutions might also be good enough.

Those of you thinking AMD had messed up big time and that a bad batch of processors were sent out to a lot of customers really need to try these new BIOS that are starting to come out from the board partners, as they honestly seem to work. It was pretty obvious reading between the lines reading all the forums out there that a lot of this seemed to be down to the way voltages and other settings were being handled in the background, rather than fundamental flaws with a chip production batch. Obviously an hour or two of testing is not the end result, but I would say this is a massive step forward as I know the older bios would have had me blue screening in minutes on basic web tasks, let alone anything more creative.

The only thing that is still buggy appears to be the LED Aura settings which don't always survive a reboot. That's an issue that seems to be infected the last 2-3 BIOSes from Asus, and I don't see this as a deal breaker myself, although it would be good if Asus fixed this. it's clear Asus testing team are not always up to the task, but at least the more fundamental issues seem to be resolving themselves.

As always YMMV, so keen to hear what others think about these new BIOS versions and how positive (or negative) your own experiences are. Personally I am now pretty happy with this 5950X based on these updates.

 

[Daylight_Invader]

I've had one blue screen although strangely it does not appear in logs. I also suspect it was driver caused rather than to do with this bios, and no WHEAs.

I doubt we are totally out of the woods yet, but I still see this as several steps forward.

 

[Daylight_Invader]

Since the new bios, the bad block errors have gone but I’m now seeing sporadic page fault errors. SFC and Chkdsk comes back clean. Any one else seeing these? Only seem to come up during boot or shutdown, and never during normal use.

The only other issue I am seeing is no sleep mode, despite seeing there being no requests that should stop sleep mode being activated. Monitor also does not switch off despite being set to do so. Any one else seeing this?