@12348766754 Look on page 68, I posted a list of the trouble shooting AMD sent me. If that doesn't fix the issue you might need to replace a part or something, it would depend on your setup.
If AMD's troubleshooting doesn't work for you please contact their tech support.
If I recall correctly I think I contacted them from this page https://www.amd.com/en/support
Adding to the list of people who have been running into these issues as well.
Mobo: MEG X570 Unify
RAM: Crucial Ballistix DDR4-3600 (4x16GB, Micron rev.B, 16-18-18-36), currently at stock (I've tested other kits too, to be safe - doesn't make a difference)
PSU: Seasonic Focus PX-750
GPU: EVGA XC3 Ultra RTX 3070, currently at stock
I've been consistently running into issues with WHEA-Logger (ID 18) and Kernel-Power (ID 41) events pretty much since I first built my system back in December. I've already RMA'd my 5950X once, and even RMA'd my mobo with MSI to see if they could find anything. Both 5950X had the errors paired with a hard crash, both with different physical cores (12 for the first one, 14 for the second one IIRC).
I've done a lot of pretty extensive troubleshooting as well - switching out RAM, playing with BIOS settings, testing in and out of the case, etc. Nothing's worked with my current CPU. The only way I can get it to run stable is by disabling CPB, which gimps the CPU. I submitted a new warranty request with AMD.
At this point, I genuinely can't tell if it's just really bad luck or if the board I have still doesn't play well with Ryzen 5000 yet. Anyone have any thoughts? Anyone run into similar issues with this particular pairing of CPU and motherboard?
I am one of the many that though was also affected with the 5900x.
Ran very cool at 35-40C.
From searching online I found my temporary fix: DISABLE CPB & P¨BO (make sure to disable in all locations) and anything that would bring the frequency above 3700mhz / voltage around 1.008, +/1 0.012v. This means my CPU seemed to always go between 0.996, 1.008, and 0.120 on Idle.
It works, it's stable, I ran every other test on my PC and every component seems to work very well. Including the CPU until you boost it to stock settings.
I was then writing this post as I had questions about RMA and I was conflicted, then while describing it occured to me that my mobo has 2x 12V connector for the CPU: a 4 pin and an 8 pin. I only connected the 4 pin as I did not realize you could plug a 4 pin in half the 8 pin connector
Long story short, issue is now fixed. I have all the latest drivers, bios, and after days of troubleshooting I realized that it was underpowered which means I had to run it in an undervolted setting.
Wow what an adventure!
Hope this may help someone else! 🙂
That did not last long.
I was so sure this would fix the issue.
Unfortunately while running additional benchmarks it crashed during Borderland 3.
Shame.... I am in the beginning of my RMA but I don't trust my seller.... they want to run tests and then charge me if they deem it works but I also contacted AMD, they acknowledged that I was suffering from whea 18. and ya it's exactly that, including some kernel p, and bluescreen view seems to confirm that as well. So they suggested a few other tests which I ran or had already ran, and then advised to send a warranty claim if I was still having problems.
Should I wait until my 30 days with my seller so I can go with AMD instead? or can I still go with AMD regardless of my 30 days with the seller?
Or should I sell it and downgrade to a 3900x?
I am conflicted, spent quite some money on all this and I kinda regret going with the latest Ryzen after all these years with Intel. I was excited to change, but I don't think it is too late to drop the sponge yet. I still have some faith after seeing many people receiving replacements CPU that solved their issue.
Anyway cheers mates!
Please confirm that you yoi did set in BIOS the PSU to 'typical' power instead of AUTO.
Just doing that, with everything default (but RAM forced down at 3900 in plave of 4400), fixed all of my issues!
No more crash nor error 18. Just enjoying my new config!
TECH TIP: Look at the timings your manufacturer gave you for your RAM chips. For good RAM those numbers are something like 3600-15-15-15-15-32-56. Okay, now go into BIOS, and look for the tRCDRD and set it to +1 or +2. For example, I have some of the worlds best RAM, and it was running at 3733-14-14-13-12-28-42. ONLY under very heavy special loads, I had errors and crashes. So I changed tRCDRD +2. The errors went away. The crashes stopped. AIDA64 and other benchmarks reported much more consistent performance metrics. Keep in mind... The RAM modules, the motherboard and the CPU ALL SAID the timings should have been fine. But in fact, the read timing had to be bumped up a little bit, OR I had to boost my DRAM voltage more, OR I had to reduce my system clock rate. I think something with the combination of their LCLK PCIE and their new chiplet design... they needed an extra memory cycle and it exceeded the original spec, but they don't want to tell us WHY ARE they working with major memory manufacturers RIGHT NOW to change the chip specs ONLY for AMD? hmmm... it's almost like they know exactly what I'm talking about, huh.
Contact the people that sold you the PC and tell them you are willing to process an RMA with AMD directly, versus through the seller/integrator. Be prepared to open the case, pull off the heat sink, and take a picture of the CPU serial number, with enough extra details to convince them that the picture is unique and yours. If you cannot partially disassemble the PC without voiding a warranty, then don't do it yourself. Because you've already told the seller/builder that you're about to process the RMA. Now, you contact AMD with the photo and ask them for a box in which you can put your old CPU for RMA. When they tell you to use toilet paper and hopium to wrap up the chip... You should debate with them ..... The best way to protect the CPU from pin damage in transit for RMA. Tell them you REALLY don't want to take full legal responsibility for any damage that happens.... and you refuse to fabricate your own container, because you're not a professional package designer, and it is your belief that a judge in the court of jurisdiction would not release you from liability unless you use a pre-made shipping container. I mean, there's a good reason that CPU trays are a very specific shape of plastic .... down to the millimeter. IF a customer attempted to make their own packaging, they could refuse to honor your RMA, "due to customer damages". So then eventually, they will send you an empty box with a pre-built shipping container... ORRRRR.... you fill in the blank and enjoy.
Tried both the ram read tweak and the typical setting for the PSU. Still same issues.
Going back to disabling cpb & pbo so at least it works... Like a 3900x...
#1 Reseat RAM modules and CPU to make sure bus is clean. Only do this yourself if you're really careful and good with tiny tiny little pins that can bend and thermal paste that needs re-applied, and all that. Plenty of youtube videos.
#2 reseat the video card, and make sure that if you can, the pwoer lines from PSU to video are on separate cables/rails (just spread out the love)
Verify your power supply is not some crappy 500 watt junker. IF you have a buddy with a spare PSU, maybe put it in there for a day or two.
Ensure memory is set to 'Gear Down Mode' Enabled.
CPU and SOC LLC auto
Consider manually setting VSOC, VDDG IO and VDDP. They all matter. We all underestimated the right numbers, I think.
In the PBO Curve optimizer, you may actually have to set all cores to +15 or +20, to get stability. Not all the CPUs were that great in batch 1 and 2. And your Motherboard might not be using LLC the same way as others.
The following is not for all RAM modules. Need to research what you've got to get your exact numbers. Ryzen DRAM calc and Taiphoon Burner get close, but still may not be perfect.
1.55 RAM voltage
1.15v VDDCR SOC
1.1275 CLDO VDDP
1.15 CLDO VDDG CCD
1.15 CLDO VDDG IOD
14 CAS Latency
15 Read Row Column Delay
8 Write Row Column Delay
13 Row Precharge Delay
12 Cas Write Latency
@ViciousDelicious on page 68 I posted all the first level troubleshooting I got when I contacted AMD, they'll probably ask you to do that if you start the RMA process.
I'd already done most of it except changing the PSU idle current setting to Typical.
This fixed it for me, I still run DOCP and all MOBO settings default.
However this setting shouldn't need to be run, I suspect it's my PSU's fault as it's from 2012 before the new low power specs from Haswell that year. But the PSU manufacturer said they were supported and worked fine for many years. Chips use even less power these days, PSU's often don't advertise the minimum amp spec of the 12v rail if you can get a 0 amp spec you'd be fine. I calculated mine to probably handle as low as 5 or 6 watts and typical idle control forces this to 7w. Although If I recall correctly I think I saw my system often get down to 3w. Auto on the mobo it supposed to detect what modes your PSU can handle, so is it the MOBO's fault?
Anyway for you it could be the MOBO, high speed ram, CPU or PSU at fault. Change some settings swap out any part and the issue might go away. However there's not enough data but seems strange that people that swap out other parts had no success and those that swapped out CPU did, You'd expect to see reports of both.
My plan when I get enough spare time and money is to swap out my PSU to one with a 12v 0a min rail and see if It runs fine on the default setting for PSU idle control. Otherwise I'll swap the CPU out with the store, I only need a receipt and they'll exchange with a new one over the counter. I could of done so already, However I don't want to do that unless I'm 100% sure.
Tommorow replace my PSU, this is old ANTEC 1000w platinum but work fine with 5600x. The new PSU is a Bequiet Dark Power 12 Pro 1200w and have 2 MOBO for test, Asrock Taichi X570 y Asrock Taichi X570 Razer Edition. If the system is not come stable maybe rollback to intel 😞