Ryzen 5800x WHEA errors

[Jamin280672]

As the title suggests, I'm getting WHEA errors left right and centre on my Ryzen 5800x when running OCCT stress test, I've set everything to stock settings and still getting them, the only thing that is running overclocked is my ram which I have dialed into xmp settings but dialed in manually with a bit of extra voltage, it's giving me an ID on the error of 0 which I relate to core 0, would this be right ? I've removed PBO, curve optimizer, the lot and still getting them, any ideas anyone ?

Running Windows 11
Ryzen 5800x
32gb (4x8gb crucial ballistic 3600mhz cl16 1.4v)
Gigabyte x570S master
4 X nvme drives
Asrock RX6800 phantom gaming X GPU

 

[Jamin280672]

I had these whea errors on my 6700k ocuk told mee to up the cpu voltage a wee bit .

It's on auto, because when it boosts it pulls upto 1.5v, so it has all the voltage it wants.

 

[Jamin280672]

WD nvme drives can cause whea errors.

3 X Samsung 980 pros and 1 X crucial nvme.

These are the settings my system required to fix WHEA errors on core 0 -

RAM 1.38v
SOC 1.124v
VDDG IOD 1075
VDDG CCD 950

Anything over 950 for CCD caused more problems than it solved.

Thanks, I'll give this a try when I get home, seems to be a common problem on the 5800x, I've been reading through a load of threads today, apparently they fixed it with agesa updates and then it came back, you can't back flash agesa unless your board has some sort of bios flashback, thankfully my X579S master does, so I may try some other bioses out.

 

[Jamin280672]

Out of curiosity, what cooler are you running with your 5800x ?

I'm custom water cooling.

I didn't even know I was getting these errors, I'm not a big gamer, I just have a tinker with call of duty, battlefield and the new Microsoft flight sim every now and then when I'm bored, I was passing all benchmarks I threw at it like IBT, cinebench (all of them both multi and single core), memtest and all sorts, I decided to give it a run with OCCT last night and that's when I first discovered the WHEA errors, as OCCT basically shows them right to you without opening event viewer.

I was using the gigabyte active OC turner in the bios where I was using PBO for under 45amps and a manual overclock of 4.7ghz when the CPU requested more than 45amps. I immediately removed everything to do with overclocking the CPU and disabled the active OC turner, disabled curve optimizer, even backed off on the fabric and ram speed and still getting the WHEA logger errors.

 

[Jamin280672]

I’ll try running Occt for a while then and see if crashes and gives me an error. That being said my whea errors BSOD and restart the machine so i may not be able to see the error details.

Yes do, it didn't take long at all, I just used the standard test with medium data set, only took 2 mins of testing before the WHEA errors started to appear by the tonnes.

 

[Jamin280672]

Check the event viewer and see which error it is.

Event 18 means CPU is borked, RMA will most likely be needed.

Event 19 means it's fclk related, lower fclk or play with voltages.

Thanks, I'll double check when I'm home, I don't finish till 9 tonight, I'll post some Picts.

 

[Jamin280672]

Thanks for your help guys, looks like ive cracked it, I basically got home last night, reset the whole bios and then went through each setting 1 at a time with a quick test inbetween, looks like either the ram or the fabric was the problem, so far ive got PBO on with curve optimizer of -20 across all cores, I need to set this to per core and find out which ones are better than others, this is with +125mhz boost, so its boosting to 4975mhz and holds there for quite a while

Ive enabled XMP and set the fabric to 1800mhz with ram at 3600mhz but some really rubbish timings, need to sort this out yet, but CPU first hey, ram voltage is at 1.38v and the FCLK voltage is at 1.15v.

So far so good, im getting some pretty good results in the quick bench on CPU-z, right now as I type this OCCT is running and has been for 30mins without any errors, the CPU is bouncing around between 4.8 and 4.7ghz, a bit more testing to do and then i'll start working on the ram, Ive had this kit of ram running happy all day long at 3800mhz with 16-19-16-38-1T-GDM enabled, 1.45v.

 

[Jamin280672]

Sounds promising. I spent more time with mine last night as well. I’m beginning to lean towards it being a power delivery quirk when the processor is down stepping. Apparently the 5000 series can momentarily draw a high spike of current when they step down which can trip the vrm protection, leading to a hardware error being flagged. Older boards like mine were not aware the spike, so more likely to trip.

The solution possibly being to one of the increase the Pbo limits so that it doesn’t trip. I’m still messing about to find a happy medium.

this might explain why I see the crashes at lighter loads and not when running stress tests.

I have my PBO settings set to 280, 200, 200. That's just to start, better more than not enough, I will fine tune them a bit further down the line as hwinfo shows you what it's using out of that lot, I'll run an extreme benchmark like IBT and then just add a tiny bit more to what hwinfo is reporting being used on the PPT, EDC, etc

 

[Jamin280672]

In the interests of time ( which I don't have much of at the moment ), I've given up a wee bit on it. Turned off CPB and PBO and just set the multiplier to x44. For my setup, that seems to hold all cores around 4400 / 82°C under cinebench load with score around 14700. Apex seems to run fine without crashing, and other programs fly along, and the computer is quiet during normal working where it sleeps the cores and idles around 35°C.

To me, its good enough, more than fast enough, so I'll stick with that for now!

To be honest, I was getting to that stage, and that's fair enough, I know the feeling, I work 16 hour days so only get an hour in the morning and in the evening, 2 days off a week so I'm very limited for time too, if it didn't behave itself last night, I'd of taken the same path as you and you're only losing out on about 300mhz single core speed, you won't even notice that in real world usage.

 

[Jamin280672]

I finally cracked it in the end, 2 full days of testing, curve optimizer is a nightmare to do per core, but got there in the end, obviously core 0 is my best core, so it wont optimizer much more than it is already, RAM is on XMP with tweaked timings, and then I have an all core overclock of 4.7ghz above 50amps and +150mhz on the PBO side of things, so 5ghz max CPU on light loads.

 

[Jamin280672]

wasnt the fault of the cpu in the end, it was either the fabric or my RAM, anyway seems as all my ram sticks are the same, right down to the ICs used on each stick, I took 1, downloaded the SPD data and XMP data off it and flashed the other 3 with the same data, so now they are basically a quad matched kit.