I would look in BIOS and see what version you have. If it's ending in a zero. it's a pre-release and can cause issues. It's meant for Ryzen 5000. The latest Agesa for 30000 is 126.96.36.199 Bios's 188.8.131.52 or 184.108.40.206 are problematic. Agesa applies to all motherboards, it's the Ryzen micro code and power profiles. One other sore problem would be if you installed an AIO or removed the AM4 backplate and scratched the back of the board. I had that happen to an AM3+ and it was ruined. Had shutdown problems left and right.
Ahhh, check the back of the board, like in my last post. If that got scratched around the AM4 socket or mount area, it's a toasted board. It might work a week or 10 seconds, eventually it will show itself. The tiniest of scratches will do it.
That sounds as likely as anything else, at this point. God knows it got moved around enough to risk that organizing things.
I'll keep an eye on the black friday and cyber monday sales for a replacement. Worse case, I sell it again and get most of my money back.
Take it apart man and look real close at the back of the board where the plate mates. It's so easy to accidently micro scratch it. I needed a magnifying glass to see it but it was around the holes on the AM3+. MSI board and the fine gold wires got exposed, shorted the entire board. Then I could see the white burn marks going to the RAM from the CPU. The CPU lived, the board no. Why they solder or melt critical wires so close to the surface of the PCB during manufacturing I have no idea. 3 layer boards and they still probably make them the same way.
Slowed to a few a day.
Next step: inspecting the board (not easy, where all the mounts are. And it would probably be around a standoff. It's likely it just wouldn't be visible.) Going to try the old power cable I was using, just in case it's a shielding or short thing. After that, going to swap in my old GPU (a GTX 950 FTW Edition) just to try narrowing it down, except in that case I'd expect it to happen in games more than Firefox, or at least be more repeatable.
If those aren't it, it's the house wiring, surge protector, CPU, or motherboard. No other possible hardware issues that I can find.
Not ruling out software, but I wouldn't expect that to be so sudden or to reboot so hard it doesn't deliver a crash dump. And aside from updating or rolling back, there's nothing I can do about software anyway. And I'm not eager to do that without knowing what it is, because rolling back some drivers will cause issues.
Update: Switched power cord. It's a known good one. Doubtful cause, but -1 variable. Might make no difference, won't hurt.
Checked through BIOS again, Auto had my RAM at 16-16-16-39 at 1.30v (from what I can tell; there's a lot of numbers with no explanation, often a big string of 16s, no 18s) for some reason. Enabled DOCP, which adjusted it to match the 16-18-18-38 at 1.35v as on the label. I'm running 4 sticks, bought as two separate pairs, so hoping Auto was being dumb and this might level things out.
Memtest wasn't reporting any errors before, but I figure actually matching the label is the last thing that will hurt stability, and at least that's not risking underpowering it. Especially since I bought two at a time instead of a full kit. Don't know when that happened or if it's always been that way, or why either the timing or stability would change after months. Burn-in time maybe? Or the BIOS trying too hard.
Whatever. Keeping an eye on things.
Did you check the CPU back plate to the AIO area? That's the area that I meant most. Because when mounting it's easy to let that slide against the board and scratch it. Those RAM timings should be fine either way, especially if it passed Memtest. The box timings are usually tighter for compatibility but not "optimal", which is why tuners, like myself, use DRAM CALC or a similar program to calculate tighter timings for reduced latency and increased read/write speed and for the 10-20 FPS gain. But stock can't hurt.
What mobo do you have? If it's AsRock, like me, that is more of a problem. I need to RMA my board and buy a different brand.
It's an Asus TUF Gaming x570 Plus (wi-fi).
I haven't done any checking except a quick glance yet, until it's crashing again. The case doesn't leave too much visible and the PC's not in a particularly great place for it. So it's on the list, but will be a pain.
I'm also not too upset about 10fps or so, at the framerate of today's programs. Stability is a distant #1 from any other priority. If I really wanted the extra frames, I would be better served spending more on hardware instead of trying to milk it cheap from a 3600 or a GTX 1650 GPU.
I just wish Asus's default wasn't in overclocking everything. Especially in messing with RAM timings, which usually needs higher voltage IME.
Hey, don't be knocki'n my 3600X, it was all I could afford and ranks high in 1080p gaming CPU's, which is what I do. I think 165 average FPS in BF4 using "Ultra" is decent. FarCry 5 running at 131 FPS "ultra" is good too. I have a RX 5600 XT with the 14Gbps BIOS.
Your RAM timings default wasn't an OC, it's what the BIOS read from the JDEC default on the RAM chip. Which I might add, is not too shabby, but wasn't giving you any real gains. That is more in the sub timings and the "1T" command rate, etc.
No, not true in most cases that overclocking RAM means higher volts. Mine with custom timings at it's stock 3733mhz is 1.35v, like on the box. The most juice DDR4 needs is maybe 1.4 if you try to increase frequency from 3600-4000 or something. Anymore and, yes, one delves into the danger zone and should buy faster RAM. That said, the "sweet spot" for Ryzen 3000 series CPU's is 3733mhz. It gives 1:1 Infinity Fabric to RAM clock and is the fastest you can go without decoupling to 1:2 or something. Doing that can hurt single thread performance, which AMD has a hard time competing with as it is. Your 3600mhz is fine, not a huge leap from 3733 performance. Straight CAS 16 like it put in is great and stable as Memtest showed.
Doesn't your case have a cutout to see the back of the CPU also to change the cooler if need be without removing the board? If so, you can remove the cooler carefully to inspect the area under that plate.
Oh, not knocking it. My 3600 is awesome when it's working right. But if I was one to want fps higher than my monitor's refresh rate, I would have shelled out for at least the 3600X, and a 2000-series GPU (I built it just pre-3000 series), and probably a more expensive motherboard, and WD Black or Samsung SSDs, and...
Anyway, my point is I'd much rather have something acceptably fast that never crashes and didn't hurt my bank account as much, versus trying for every ounce of speed and dealing with the headaches and/or costing twice as much.
And I did OC the RAM a couple PCs back, and know it's not a huge problem in itself. It's tightening up the latencies and undervolting it at the same time, across non-kit cards, all at once without testing along the way, that can be an issue.
And the case does have a cutout on the backside of the socket, but only there. Scratches around the standoffs are more likely, if less likely to cause issues, but even in that spot it would be a pain. The PC needs to be unhooked from everything to get it out where I can see that, and the cooler needed it's own great big mounting plate versus the stock cooler mount's... equally huge mounting plate.
Still better than my last PC's Intel mount, with the dumb push-and-twist studs that wanted to either stick or break every time.