Crazy-high Ryzen 3900X temperatures [solved]

[Centauri]

Ryzen 3900X with a Corsair H100i rad and MX-4 in between. ASRock X370 Taichi.

I just noticed that temps were idling at 70ish and immediately hitting the 95c thermal limit on load. Which got me looking at voltages and the BIOS had the CPU at 1.45ish for god knows what reason. So with a bit of research I dial it back all the way down to 1.275v. And while that lowered the idles into the lower 60s, it still hits 95c immediately on load.

And I see no problems with the contact between the heatsink and the CPU.

Ideas?

 

[VirtualLarry]

I had approx this problem, with my 3600, and my B450 Aorus Pro Wifi board, with AGESA 1.0.0.2 and the original 7/7 chipset drivers release.

 

[DrMrLordX]

@Centauri

We have at least one user here on the forum with the same CPU + mobo combo as yours. He hasn't reported such trouble.

Seems like you're getting a lot more voltage than normal if those numbers are to be believed. Load temps SHOULD be accurate even if idle temps (and voltages) are suspect on all Matisse setups.

Are you operating at default settings? Do you have the latest UEFI revision for your board (there are some betas out there that have helped @Markfw )? And do you have the latest chipset drivers from AMD?

You can also try applying a negative voltage offset inside the UEFI to try to bring down those voltages and temps. Might be worth a shot - just be careful not to go too far, or you'll run into clock stretching which is actually pretty annoying for desktop users.

 

[Centauri]

Trying the latest chipset drivers as we speak, but the flaming temps are reported at BIOS level so I'm not sure it'll help.

Default BIOS settings, latest BIOS. Even did a Load Defaults just to be sure.

Also tried the neg offset but it would only let me drop -.1v. And that's what still left with me with low 60s on idle and 95c immediately on load.

My Corsair is also set to Extreme for both pump and fan.

 

[Centauri]

No change with the chipset drivers.

 

[Centauri]

Also just re-applied the heatsink. No way it's not on solidly.

 

[IEC]

Also just re-applied the heatsink. No way it's not on solidly.

How are you applying the TIM?

 

[IEC]

The other way to determine if it's the cooler-IHS interface that has poor contact would be to try the stock heatsink and see if your max temperature makes sense or not.

 

[Centauri]

How are you applying the TIM?

The same way I've done on every every build I've ever done around a heatspreader'd CPU; a drop of TIM the size of a pea right in the middle, flattened out by the heatsink.

The other way to determine if it's the cooler-IHS interface that has poor contact would be to try the stock heatsink and see if your max temperature makes sense or not.

I'll give that a shot tomorrow. But I really feel like there's another problem here. I've been running this build for 2 years and the only thing that has changed was the CPU two days ago. Never had anything even resembling high temps with my 1700 installed.

 

[Markfw]

I have a 3900x coming next week, and its going in an x370 Taichi. I will reply when I have it up, as to how its going.

 

[scannall]

The overclocking part of bios is still pretty raw on some boards. Set it to defaults and enjoy. When it gets cleaned up, then tinker. Tinkering when your tools are broken usually ends badly.

 

[DrMrLordX]

The overclocking part of bios is still pretty raw on some boards. Set it to defaults and enjoy.

He is using defaults, near as I can tell.

@Centauri

Okay, voltage offset is going to be a no-go. You're going to have to force static voltage to get to the bottom of what's happening here. There are two ways you can go about it: use Ryzen Master to tweak things in Windows, or just set things in the UEFI.

I would highly recommend that you set a static clockspeed that is very conservative, like 3800 3000 MHz, and a static voltage of 1.2 1.0v. Then run a benchmark like CBR15 or CBR20 and see if it crashes, what temps you get, etc. Also! Be sure to use CPU-z to measure your CPU voltage during the benchmark. Check this value carefully. We need to see if the board will stick to your specified voltage setting. You will probably see some vdroop, but honestly with your board feeding so much voltage right now, I would hesitate to recommend any LLC settings until you've done a little testing with a static voltage first.

 

[Topweasel]

The same way I've done on every every build I've ever done around a heatspreader'd CPU; a drop of TIM the size of a pea right in the middle, flattened out by the heatsink.



I'll give that a shot tomorrow. But I really feel like there's another problem here. I've been running this build for 2 years and the only thing that has changed was the CPU two days ago. Never had anything even resembling high temps with my 1700 installed.

Can't speak for the Tiachi. But I just built a system with the 3900x, Asrock 570x itx board, and H100i. The H100i I swapped out the fans for some Noctua quiet fans (think it was the PWM 120x25) so no the same but somewhat close.

For Tim I did 3 peas, one in the center and 1 over/very near the 2 core dies. It's been mentioned but its probably not smart to use the grain in the center approach on these CPU's the parts that get hot are the 2 core dies and a grain in the center might not spread all the way out to cover the dies.

But the result is high 30's/low 40's at idle, low 70's when med load, mid 80's with all cores at 100%. I don't hit 90 unless I am running PBO with AOC at 100% all core usage.

 

[IEC]

The same way I've done on every every build I've ever done around a heatspreader'd CPU; a drop of TIM the size of a pea right in the middle, flattened out by the heatsink.

I'll give that a shot tomorrow. But I really feel like there's another problem here. I've been running this build for 2 years and the only thing that has changed was the CPU two days ago. Never had anything even resembling high temps with my 1700 installed.

Word of warning: I've heard the stock TIM they use is like glue and sticky, so you may need to use dental floss or fishing line to help loosen the heatsink before removing it. I never had issues, but I never use stock TIM.

 

[Snarf Snarf]

Can confirm that small pea TIM method doesn't work too well with 3900x. I had to re-apply mine with an X and then I spread it out evenly using the spatula that Thermal Grizzly includes with their TIM. Mine sits at mid-low 30's and max load 80-85 with Prime95 mixed test at 4.35 all core on a 360mm radiator.

 

[Centauri]

Everything about the AIO Corsair told me it was working from the iCue app, to feeling the vibration on the pump/heatsink housing to feeling movement through the hoses.

Swapped in the 3900X's heatsink and everything is immediately 20 degrees lower across the board and no more throttling.

*sigh*

Glad I figured it out though.

 

[DrMrLordX]

Glad I figured it out though.

Ah good. So it's a faulty cooler. Glad it wasn't the x370 throwing insane voltage at your chip or something.

 

[Markfw]

Ah good. So it's a faulty cooler. Glad it wasn't the x370 throwing insane voltage at your chip or something.

I still have 4 AIO's, but am being more cautious now, as I have had my share of failures, (like 5)

 

[DrMrLordX]

I still have 4 AIO's, but am being more cautious now, as I have had my share of failures, (like 5)

Corsair usually has a good rep for these sorts of things, though. Not like a certain Threadripper AiO with which you are intimately familiar.

 

[gk1951]

Any AIO can fail just like the fan on a cpu cooler can fail.
From what I read from 3900x owners, the cpu is near max OC at stock and runs hot.
I have a 9900k and it has the same "qualities". Good cooling is a must.
The Wraith cooler (I have one that came with my 2700x) is a good cooler that serves the purpose.

As the 3900x gets further into production and AMD refines the BIOS, there might be some slight improvement in cooling. In addition as more of these chips get into the mainstream, cooling producers will improve their products for them.

Imagine the cooling needs for the 3950x

 

[DrMrLordX]

3950x may require custom water. I'm quite happy to have put custom water on my 3900x since it (apparently) lets me get away with some clockspeeds that others can't. I haven't hit 95c yet, and I have no intention of going there if I can avoid it.

 

[VirtualLarry]

I've hit... 117C. When I thought that the temps were somehow incorrect.

 

[DrMrLordX]

I've hit... 117C. When I thought that the temps were somehow incorrect.

Yeah don't do that again, if you can avoid it.

 

[StefanR5R]

3950x may require custom water.

It's a 105 W chip. A 120 mm slim tower air cooler should be sufficient.

What kind of cooler you put on the top of the lid of the processor influences the heat flux density within the 7 nm CCDs only to a small extent. Doesn't it?

 

[DrMrLordX]

It's a 105 W chip. A 120 mm slim tower air cooler should be sufficient.

These things get hot even with very low power flux. My 3900x has never dumped more than 190W of heat, yet the temps are high. Typically package power is 160W or less. The real concern is core power though.

What kind of cooler you put on the top of the lid of the processor influences the heat flux density within the 7 nm CCDs only to a small extent. Doesn't it?

It's not going to change heat flux density at all. The question is: how much heat can your cooling solution handle per mm2 with an acceptable temperature delta?