5800X constantly crashing machine

[shogglet]

I've just recently installed my shiny new 5800X processor, and immediately noticed instability in my system following the upgrade. (I upgraded from a 3700X).

The CPU seems to run pretty hot, and I'm not sure if this is what is causing it to crash, but the temps are within the stated maximum of 90 degrees celsius.

Here is my setup:

MSI MAG B550 TOMAHAWK Motherboard (latest BIOS update from November 2020)
AMD Ryzen 5800X
Corsair H100i RGB platinum 240mm AIO
nVidia RTX 3090
Corsair Vengeance RGB PRO 32 GB (4 x 8 GB) DDR4 3200 MHz C16 XMP 2.0
Corsair HX 850i Platinum PSU

500GB NVME SSD for OS

Pretty fresh, clean Windows 10 Pro install (after motherboard installed).

I've re-mounted the AIO pump/heatsink block twice to be sure it was contacting the processor correctly. Both times lifting the block up I could see I had applied the thermal paste well, as it had a good, even spread.

Idle temps seem to hover around 50 degrees celsius, and when loaded, the chip will go all the way up to 80+ degrees.

All it takes to crash is one or two runs of 3DMark timespy defaults, or a bit of gaming - for example the Anno 1800 benchmark loop for 10 minutes.

I've tried setting precision boost overdrive to "eco modes" (65w and 45w) and those don't seem to help.

The RAM XMP profile is set so it is running at 3200MHz CL16.

Lastly, boost speeds on the chip seem to hit their stated maximums at just shy of 4.8GHz.

Has anyone seen anything like this? Are there some settings I'm missing on the motherboard bios that need to be changed?

Thanks!

 

[snips86x]

Given the temps you've stated, that would absolutely looks like thermal protection kicking in.
Have you changed the thermal paste to something more well known/reliable and checked the pump is working correctly on the AIO? (give it a shake and listen for excessive noise)
What about the state of the radiator?

 

[shogglet]

Given the temps you've stated, that would absolutely looks like thermal protection kicking in.
Have you changed the thermal paste to something more well known/reliable and checked the pump is working correctly on the AIO? (give it a shake and listen for excessive noise)
What about the state of the radiator?

I'm using a new tube of Arctic Silver 5 thermal paste. The radiator is in good condition. I haven't tried giving the pump a shake, but it does seem to operate silently. The monitoring in the Corsair iCUE software reports the pump and fans operating OK.



Even the internal pump temperature seems indicate the fluid is not too high in temp either.

I have wondered about getting another cooler to test with - maybe a beefy air cooler to throw on and rule the AIO out.

 

[snips86x]

That would definitely be worth a try.
If you check your BIOS, is the pump header set to full speed?

 

[Doug2507]

Tried changing fan profile to extreme and seeing if that drops temps at all?

 

[shogglet]

Tried changing fan profile to extreme and seeing if that drops temps at all?

I did have a go last night at raising the fan profiles up to max and with a lot of noise, it didn't seem to help the temperatures much at all.

I ended up swapping my 3700X back into the system and everything is stable again. I can game and run benchmarks on loop with no crashing or instability. However the temps still seem fairly high to me. Under sustained load, is 70-72 degrees celsius normal for a Ryzen 3x000 series chip (3700X in my case)? I've never really paid attention to the temperatures before as its always been rock solid for stability.

Here is a run I setup this morning with timespy CPU on loop, and various monitoring tools running.

You can see the pump RPM and pump fan RPM is good, and the CPU-Z temps reported at around 70 degrees under load. This is of course the 3700X now which is running stable like this.

With the 5800X the temps reported were around 5-10 degrees higher for the CPU package and it would crash constantly.

So to clarify, below screenshot is 3700X stable. 5800X was in the same system, but reported 10 degrees higher temps.

 

[pc-guy]

what have you done with the Zen3 PBO settings etc?

whatever you have done to that 3700x looks like a 1.35V all core OC at 4GHz. zen3 does work like that.

also not sure why your AIO is not cooling the CPU down. mine is a 7yr old H100i and it can handle 100% core load on my 3900x outputting 150w with less CPU temp than yours. Mine peaks under Aida and OCCT 72C.

I am using P12 fans at 800RPM under stress tests

 

[shogglet]

what have you done with the Zen3 PBO settings etc?

whatever you have done to that 3700x looks like a 1.35V all core OC at 4GHz. zen3 does work like that.

also not sure why your AIO is not cooling the CPU down. mine is a 7yr old H100i and it can handle 100% core load on my 3900x outputting 150w with less CPU temp than yours. Mine peaks under Aida and OCCT 72C.

I am using P12 fans at 800RPM under stress tests

I haven't changed anything there - just left the PBO at it's default settings in the BIOS. The 3700x seems to maintain roughly 4GHz on all cores when running workloads that stress all cores to their maximum. I confirmed that earlier by running Cinebench R23 - about 8 passes of that, and the processor maintained 100% load and a 3.9 to 4.0 GHz speed.

As a test over lunch I took out my Corsair H100i AIO and put the stock AMD Wraith Prism RGB cooler on the 3700x (still had it boxed and unused). I booted up and noticed the Idle temps were already slightly better than the idle with the H100i...

Load temps seem to get close to what the H100i was giving too. I ran a 3DMark timespy test and I got an extra 600 CPU points over what I got off the H100i, so it seems like the chip is boosting a bit better on the wraith prism. Not good for the H100i!

Maybe that is the problem - the AIO is just not cooling the 5800X well enough and it has some sort of internal blockage or something, and with the 3700X I am seeing stability because its TDP is lower than the 5800X and the H100i is only just coping with it...

I think the next step might be to get a beefy air cooler, or perhaps an Arctic Freezer II 240 AIO and try that out...

BTW thanks for the insight on your own H100i. It helps to have a few points of comparison to guage how my own hardware is performing. Cheers!

 

[pc-guy]

If you are struggling with cooling u set PBO dial in manual OC with lower volts and go from there.

 

[BrokenWindows]

I have had a similar problem with a 5800x. I kept having games crash a few mins in. Checking in prime95 gave lots of rounding errors. Memtest86 was passing on everything but test 6 which was only showing search results for people with CPU issues. Only when I underclocked the 5800x to 3ghz did it remain stable. It wasn’t even getting that hot. @3.5gh it was 55 degrees ish. I’m sending it back soon.

looks like I’m waiting till Jan for both a CPU and GPU now...

 

[BrokenWindows]

Snip


Thanks!

have you tried prime95 or Memtest86? Now I’m wondering if there is a similar problem that has a software fix?

 

[shogglet]

If you are struggling with cooling u set PBO dial in manual OC with lower volts and go from there.

Thanks! I'll try dialling in a manual PBO setting with lower volts if I still get bad temperatures on a new cooler.

have you tried prime95 or Memtest86? Now I’m wondering if there is a similar problem that has a software fix?

I haven't tried those tests. As soon as I put the 5800x back into my system I'll give that a go and update this thread. Not likely to get a chance to do that till the weekend though unfortunately. Sorry to hear you're having trouble with yours. I'm lucky I still have this 3700x to tide me over till I get the 5800x stable, or failing that return it.

 

[pc-guy]

Thanks! I'll try dialling in a manual PBO setting with lower volts if I still get bad temperatures on a new cooler.
.

with the PBO you cna select motherboard or manual (i also think there is an option for AMD predetermined value). I would say go with motherboard as a first port with the 5800X. if that is struggle then dial in a Volt 1.3V or something like that and go from 4.2GHz all cores and keep on going up see where you land. dial in volt and all core OC is not ideal as you are not taking advantage of the single core boost that PBO is able to do etc. but it will help you wiht the thermal throttling issues.

also it would be useful to have HWinfo running at the back ground to record the temp to see if your CPU is being thermally throttled.

 

[ukbionic]

Did you check my startup apps? Epic game launcher was causing my idle to be 55 C after startup closing that down helped. Heard icue software also spikes the boost a bit.

Had same exact temps as you mention but now followed someone on Reddit. Adjusted vcore offset to negative 0.036 and used pbo with zero extra boost, negative -20 per core. TDP limit of 100 or something. Now testing or stability but making progress.. 4.55ghz all core and 4.85 boosts. Most importantly 65degrees under full load stress tests. Very quiet

 

[shogglet]

Did you check my startup apps? Epic game launcher was causing my idle to be 55 C after startup closing that down helped. Heard icue software also spikes the boost a bit.

Had same exact temps as you mention but now followed someone on Reddit. Adjusted vcore offset to negative 0.036 and used pbo with zero extra boost, negative -20 per core. TDP limit of 100 or something. Now testing or stability but making progress.. 4.55ghz all core and 4.85 boosts. Most importantly 65degrees under full load stress tests. Very quiet

Some good insight here. I'll keep this in mind when I try things out again over the weekend. Great to know what your settings are and what speeds and temps you're reaching with them. I did notice that epic launcher was causing a couple of cores to constantly boost so I did end up closing that down when measuring idle temps. Their software is not good at all, and who knows what they're running in the background in a tight loop that causes high CPU usage like that constantly...

 

[ukbionic]

Some good insight here. I'll keep this in mind when I try things out again over the weekend. Great to know what your settings are and what speeds and temps you're reaching with them. I did notice that epic launcher was causing a couple of cores to constantly boost so I did end up closing that down when measuring idle temps. Their software is not good at all, and who knows what they're running in the background in a tight loop that causes high CPU usage like that constantly...

Yes good look with it. To be honest you may get slightly higher scores if you leave things auto but for me this undervolting approach with a Tdp limit is really helping.. very very quiet gaming now. I play at 1440p so I'd ather have real world noise decrease and heat decrease in GPU bound scenario over some epoints. Having said that, I did get a few crashes in games last night so upped the dynamic vcore offset littl (negative 0.03 maybe now) with a Tdp limit of 110. I'm not running any extra boost limit.

It's now passed realbench, blender benchmark and an hour and counting of small fft prime 95, looking positive.

To assess which cores in curve optimiser to put less of negative magnitude, I ran it all on auto pbo and just monitored cores. Some boosted less than others so I put a higher voltage through them.

Let me know how you get on, I'm really liking the 5800x comingfrom 6700k

 

[ouildmonkay]

Hi guys, I found your topic thanks to Google...

I come from a 6700K too and I now have a 5800X and B550 Tomahawk. My setup crashes like yours. I suspect more the voltage than the temp.

For the moment, I just limit Vcore to 1230mV and the core frequency to 4600 (Settings/Advanced/AMD Overclocking/Manual CPU OC) and it works fine with games (and lowers the temps). The best solution would be to use PBO + Curve Optimizer.

 

[shogglet]

Hi guys, I found your topic thanks to Google...

I come from a 6700K too and I now have a 5800X and B550 Tomahawk. My setup crashes like yours. I suspect more the voltage than the temp.

For the moment, I just limit Vcore to 1230mV and the core frequency to 4600 (Settings/Advanced/AMD Overclocking/Manual CPU OC) and it works fine with games (and lowers the temps). The best solution would be to use PBO + Curve Optimizer.

Thanks for posting your findings!

I had another go over the last couple of evenings and got it more stable with B550 chipset drivers installed, however some games in particular still crash regularly. Anno 1800 is a good example. Doom eternal will also sometimes crash.

As soon as I put my 3700X CPU back in, everything is rock solid stable.

I'm using PBO on all default settings. I personally find it unacceptable to have to go and start tweaking curves and voltage offsets to get a stable chip, so if I can't get it stable over the next few days then I'll be sending the chip back to OCUK and requesting either a refund, or a replacement for a chip that is less likely to operate on the edge like this, maybe a 5600X where there are 6 out of 8 possible cores in the CCX.

 

[ouildmonkay]

My 5800X has never crashed @4600Mhz / 1231mV (PUBG, Hitman 2, AC Valhalla). A friend of mine set the voltage to 1225V and it works well too.

Cinebench R23
Multi : 15879
Single : 1528 (We're losing 5% here)

I recommend you these settings until next AGESA updates.

 

[ouildmonkay]

I set LLC to mode 4.

No more crashes @stock since