Cooling Cooling a Ryzen 5900X and Crappy Motherboard (Was: Cooling a Ryzen 9 9950X... on air?)

I'm going to start with a swap to the Wraith Thingy cooler. The 2700X was always happy under that. In fact, I'll do that now...

 

Yeah, that was the plan - it lets me adjust the speed in either direction while retaining the XMP timings.

 

Okay - old Wraith Thingy cooler sat there RGBvomiting all over the place, but more importantly also blowing air over the VRMs. I've also put the third 120mm Thermaltake fan I got in the side panel as an intake, to feed it cold air - connected to the CPU fan header, so that when the CPU fan is working hard so is the intake (and also 'cos I've already got three 120mms and a 140mm hanging off the singular chassis fan header, and Asus says that's only good to 1A - didn't want to push it with a fifth!)

Now to test it under load, see what happens. Cross 'em if you got 'em!

EDIT:
Much, much noisier than the twin-stack with two slow-spinning 120mms, obviously, but not appreciably worse at cooling: it's keeping the CPU around 70°C full-load, which is fine. The extra cooling on the VRMs has made no real-world difference, though: scan-to-PDF run still takes 3m8s, same as before.

Now to keep an eye on stability, and if I get any more crashes dial back the RAM from 3,600MHz...

EDIT EDIT:
Oh, and the roof fan's an outtake again.

EDIT EDIT EDIT:
Went back to my old-fashioned cliff-shaped fan curve: now the CPU (and side panel) fan(s) won't ramp up from minimum until at least 70°C, and won't get properly angry until above 80°C. No performance impact, big improvement in noise under load. I mean, it's still louder than the Thermaltake, but... I'll take what I can get at this point.

 

Very kind of you! Sadly, my case is a Define Mini - mATX or smaller only.

 

No crashes yet - two of the three happened shortly after resuming from standby, so I left it on standby over lunch. Other than the intermittent USB delay, no problems. Rebooted manually a few times to fiddle with the fan curve - I was idling at 54°C with the CPU fan duty cycle at 20% so I bumped it up to 40%, then down to 35% which has given me a good balance of not too noisy but idling in the mid-30°Cs.

Going to reboot again to go turn off IOMMU, as that seems to be to blame for a few unclaimed IRQs I'm seeing at startup. Don't think that has any impact on stability, but it'll be nice to silence the errors on boot!

EDIT:
It did not silence the errors on boot. Bah!

 

No crash this morning. Well, yet, anyway. It is pretty cold in the office, tho'.

 

Still no crashes. Threw another heavy workload at it for a few minutes, perfectly happy. Suspended over lunch again, woke up fine - no USB delay this time. Plenty of browsing, no issues. Even threw it into the powersave governor, to lock the cores at 2.2GHz and test really low load... perfectly happy.

Maybe it really is just that this motherboard isn't happy with anything other than a top-down cooler... but also is incompatible with pretty much all top-down coolers. Nice one, Asus.

 

10:59:24 up 1 day, 19:54, 1 user, load average: 0.57, 0.60, 1.01

Still no crashes, though it did take a while for the USB devices to wake up this morning. Still don't know what that's about.

 

That's PCIe, this is only affecting USB. Looking at the logs, something's throwing it off and I have to wait for it to time out. But only sometimes, and it wasn't a problem on the old CPU. It's not a big deal, it only takes 30-40 seconds or so before I can log in.

In other news...

16:52:41 up 2 days, 1:48, 1 user, load average: 0.10, 0.41, 0.67

Still hasn't crashed! That's with Eco Mode off, CPU running full-fat full-boost, memory still at 3,600MHz.

 

Spoke too soon: it just hard-reset. Bum!

Dropped the RAM speed down to 3,533MHz. See if that does anything.

 

Another crash, same symptoms as before: I'm doing very little - in this case I'd just brought up a save dialogue for an image in Firefox - and the screen goes black and the desktop instantly hard-resets.

Dropped the memory to 3,466MHz, but I really don't think that's the problem.

I'm going to have to take the CPU out and put the 2700X back in, aren't I? <sigh>

EDIT:
This is the longest of long shots, but I've also pulled the USB cable for the card reader in the multifunction DVD drive. That's the only other change I made when installing the CPU, and I am having that weird USB delay issue - so maybe that's to blame?

'course, if this does fix it I won't know whether it was that or the RAM speed change that was the solution...

 

No crashes today. I'll give it the rest of the week, then risk bumping the RAM speed back up to 3,600MHz. If it doesn't crash after that, then it was the USB card reader all along!

 

Exciting new twist this morning: hit the keyboard to wake the system up from sleep, and nothing. Clicked the mouse, nothing. Hit the physical power button, wakes up fine... but all my USB peripherals are dead. Unplug and replug the hub... still nothing. Plug a keyboard into the front panel... nothing. SSH in, everything's fine. Restart via SSH, boots up, all USB devices spring back into life.

And that's with the suspect card reader disconnected.

Le sigh.

EDIT:
And a hard crash. Right, RAM at 3,200MHz...

EDIT EDIT:
Also moved the hub out of the CPU USB 3.1 Gen. 1 port into the chipset USB 3.1 Gen. 2 port. At which point the keyboard became unreliable, skipping keys. So, back into the CPU port it is! Shuffled things around to leave the chipset ports empty.