Processors

[Distaste]

System:
Ryzen 5900x
Gigabyte Aorus Pro AC
Gskill Trident CL14 3200 4x8gb
Seasonic Focus PX750
MSI GTX 1070 Armor 8gig
Adata SX8200 Pro 2TB NVME drive
Phantek 500 case
NH-D15 cooler(2 fan)
6x 140mm fans

Long story short is that I've been getting random reboots since building my system in March last year. This PC has been down waiting for RMA parts almost as much as it's been up and "working". I've RMAed everything in the system twice, except the CPU which has only been RMAed once. The first CPU was giving me WHEA 18/19 along with the reboots. After getting the RMA replacement and that not fixing the reboots I figured maybe it was the PSU, or memory, or motherboard. Another round of RMA's later and nope. I tried reinstalling windows, updating bios, installing newest chipset drivers, memory at stock, only 2 sticks of memory, another GPU, a 4th PSU. I've used up an entire 10g tube of noctua paste just on this PC alone. It's 99% stable outside of certain games, I've only had it reboot once on an OCCT test out of maybe a dozen or more tests. In certain games though it's going to reboot it's just a question of when.

So here I am back to CPU again, which of course makes me none to happy given the amount of time, effort, and money I've spent so far.  I fill out the RMA form with all of the things I tried and parts I RMAed and how frustrated I was. Did they read any of that? No. I gave them my dxdiag and MSinfo and now I got the response to try reinstalling windows and reinstalling chipset drivers and updating motherboard bios. Of course that's all stuff I told them I already did in my original RMA request. Nothing irks(keeping it PG) me off more than CSRs ignoring what you already told them. Then taking 24hrs to respond again even though you responded to them within an hour.

I've tried everything short of bumping voltages because I don't want to risk voiding a warranty, especially since I'll apparently need it.

I have 3 requests in my RMA that I don't feel are too egregious and hopefully Niraj didn't just gloss over them.
1. Advanced RMA- I shouldn't need to have even more downtime because you're putting out unstable products I will give you my credit card number you can charge me if the processor doesn't come back.
2. Don't make me do the picture of the heat spreader again, it wastes my time, Thermal paste, and you already know what processor I have because you sent it to me, I have the box and import papers to prove it.
3. Can I get the B2 stepping in the hopes that it is more stable.

In 20+ years I've only ever had 2 bad CPUs and they were both in the last year and one was the replacement for the first one. I've been an AMD fan since the Athlon smacked down the PIII. This experience though has definitely made me rethink that though. If CSRs don't acknowledge my requests and almost certainly if my next replacement is unstable I'm probably done with AMD. Seasonic on the other hand I'll be a repeat customer of because they sent me a new PSU and cables before I even finished the testing they asked me to do. More importantly their CSRs acknowledge the information I gave them, gave responses with that information in mind and their requests weren't a hassle.

Just had to vent while I wait another 24 hours for a response that will probably ignore what I already said.

[urielejh]

RAM clock and, consequently, infinity clock too high...? had a kit that operated @3800 with xmp but no cpus could sustain it (1x5900x, 3x5950, 5800x3d; 3 different MBs). i ended up getting the same kit but @3600 and better latencies. you're on 3200 so that is not the issue... but you may have defective kit.

[Distaste]

I've taken the kit down to 1 stick and at 2066(XMP off) after a CMOS clear just to be sure, did the test twice with different sticks with the same results. So I'm fairly sure it's not memory being unstable. Even taking the the ram up to 3600 cl14 the reboot doesn't occur more often, it can even pass TM5 Anta777 extreme profile back to back. If I push it further I get start getting blue screens on testing. That obviously doesn't insure the memory/infinity clock is stable, but I'd expect if that was the root cause the reboots would happen more often or during TM5 when increasing the clock speed.

[Zolex]

Can you try with 1 stick of ram? First without and then with XMP . If its fine go with 2 stick and back to 4. 

Can you also post you ram module info? 

F4-3200C14Q-32GFX & F4-3200C14Q2-64GVK  have your specs and should be compatible with the Motherboard.
But even then it can go wrong.

[Distaste]

I've done 1 stick no XMP(2066 irrc) after cmos clear. Also tried a different stick and then the remaining two stick as a dual rank set in the proper slots. None of which stopped the reboots. OCing the memory doesn't make the reboots worse and when pushing the memory timings too tight/far I get BSOD not a reboot.

My kits are F4-3200C14D-16GTZKW and are on the QVL list from Gskill for my motherboard, Gigabyte has similar kits on their QVL list but not that specific one. The kits are also essentially matching/quad channel kit even though they are dual rank kits, as in their numbers are all sequential. GSkill tech said the memory should work fine just using 4 sticks makes it harder to achieve a higher stable OC.

Keep in mind this is also my 3rd RMA round of memory. First set had a bad stick and all 4 were RMAed for two matching kits. 2nd set got RMAed due to the reboots, and now this set. Given that the reboots don't get worse with memory OCing or better when underclocked and different stick configurations don't solve the issue I have to assume the memory is probably not the cause.

[Distaste]

CSR just got back to me, again after a full day. Now they want my Event log, which they could have asked for in yesterdays email or better yet the first email could have asked for all the information they wanted and saved me 2 days. Even better the warranty form could have asked for all of it and saved me 3 days. I'm so absolutely fed up with this being strung out when my first RMA took all of 2 days from request to shipped. They know there are random reboot and WHEA 18 issues, these forums have plenty of threads about them. I can't reliably do the things I bought the CPU to do, but apparently my time doesn't matter to AMD. Again sorry for the rant, I'm just well passed my patience threshold.

[DimkaTsv]

WHEA 18 is either Memory config issue, or straight up FCLK instability. 

If it is memory config, you needed to test every stick and every slot. Maybe it works bad because it is 4x8 and not 2x16 config, as 4 sticks can force lower frequencies if these 8 GB sticks ar 2R each (Ryzen is kind of terrible in working with more than 4 rank setups without HEAVY downclock, sadly). And i cannot lookup if this sticks 1R or 2R each. You can do it in ZenTimings for example. At very bottom you can select individual stick to show parameters and in parentheses there is either 1R or 2R written, which means ranks)

Second possibility is that you rolled bad roulette and got CPU that cannot run 1600 FCLK (3200 memory 1:1). Which should be under RMA conditions as CPU doesn't run with memory at officially supported speed up to 3200 MT/s). 

[Thomasdk1405]

Hi

try to change the PBO - under adv. section to motherboard - follow what I write here and see if that works !

https://community.amd.com/t5/processors/5800x-crashes-now-solved-most-likely-for-all-ryzen-5000/m-p/...

[FuegoDeVolcan]

Hi Journeyman III, I have the same problem with random reboots whose only clue is whea-logger 18, I work with Matlab and Microsoft and MathWorks couldn't solve that problem. My CPU is AMD 5950 128GB of RAM with 1.8TB NvMe when my memory is not enough.