Question 9800X3D reporting 200 Celsius on HWMonitor?

[Murkxx2]

Hello, I just built a new rig and I have an issue that has been driving me crazy and can’t seem to figure it out and am looking for any input/solution. All hardware/software drivers are up-to-date with latest chipsets/firmware updates.

System Specs:
OS- Windows 11 Home
CPU- AMD Ryzen 7 9800X3D
GPU-Gigabyte Windforce 5080 SSF OC 16GB
MB- Gigabyte X870E Aorus Elite Wifi7
CPU Liquid Cooler- Artic Freezer III 360mm
RAM- G Skillz Trident Z5 Neo RGB Series (AMD Expo) DDR5 RAM 32GB (2x16) 6000 MT/s
CL30-38-38-96 1.35V
OS SSD- Crucial T700 Gen5 NVME M.2 2TB
Storage- Samsung 990 Pro Gen4 NVME M.2 4TB
PSU- MSI MEG Ai 1300W PSU ATX 3.1 80+ Plat
Case- Fractal North XL


So, let’s start at the beginning-

At all stock settings: CPU temps during loads are reporting relatively high even on small tasks such as opening a browser, or just general gaming. The high load temps seem to come in spikes- spiking from 45 to 90 Celsius in a few seconds, then back down to 45-50. Additionally, these load temperature spikes are causing my fan curves to run like a roller-coaster and I am not a fan of it. Nor am I fan of this CPU spiking temps that high just randomly while not at an extensive, heavy load. It is concerning.

My idle temps are fine, reporting usually anywhere from 38-45 Celsius. The thermal throttle limit is 95 Celsius for this chip out of the box, so I understand it is designed to run at higher temps. But the cooler, the better, right? Especially for chip longevity.

Monitoring Software:
HWMonitor
AMD Ryzen Master

So what have I done for possible resolutions? After extensive google research and finding some similar cases, I decided to try undervolting the CPU.

I have tried 2 different profile tweaks with undervolting/Overlocking.
*Both profiles have been tested for 100% stability using AIDA64.

1st Tweak
XMP Enabled- Adjusted RAM Speed 6000 MT/s
Precision Boost Overdrive – Advanced
PBO Limits- Motherboard Curve Optimizer – All Cores/Negative/20 Platform Thermal Throttle Ctrl- Manual – 95.
PBO Scalar Ctrl- Auto FCLK Frequency- 2000MHz UCLK DIV1 Mode- UCLK=MEMCLK

Results: Slightly lower temps, but still getting those high temp spikes during utilization from 45 to 80 in a few seconds, then back down to around 55-65. Idle temps -35-45. Load Temps 70-85. Better than stock, but not solving the issue it seems.

2nd Tweak
XMP Enabled – Adjust RAM Speed 6000 MT/s
Precision Boost Overdrive- Advanced
PBO Limits- Motherboard.
Curve Optimizer- All cores/Negative/20 Thermal Throttle Ctrl- Manual- 95
PBO Scalar Ctrl- Auto FCLK Frequency- 2000MHz UCLK DIV1 Mode- UCLK=MEMCLK
*CPU Ratio- 50.00 (5000MHz)
*CPU Core Voltage- 1.100V

*= solution to my issue

Results: Wow! My temps reduced drastically. 30-40 Idle Temps. Load Temps 50-60. No constant spiking utilization/temps. Tested stability for 10 mins- rock solid, perfect cool temps all around!!! Quiet as a mouse even under load, I thought my troubles were resolved.

However… HM Monitor reported a max value of 200 Celsius on my CPU Temp while gaming! It was reported on 2 separate testing occasions, on the “package” reading and also on the “CCD” reading of the CPU. Noticing this, I heavily monitored temperatures while idling and while gaming under load. Everything seemed fine, idle and current temps under load were solid. Ryzen Master had no reported high temp such as 200, but HWMonitor did.
I immediately did a stability test for 10 mins in AIDA64 with 100% CPU utilization. CPU temps holding steady at 50-65 underload. No report of failure or extremely high temp.
That seemed very odd to me. At 200 Celsius, I am pretty sure my PC would have turned into a nuclear reactor. PC would have shut down on its own. It would have literally melted. Weirdly, I noticed HMMonitor only reports the 200 reading when following the 2nd Tweaked profile, which the only change was setting the CPU Ratio to 5000MHz and the CPU Core Voltage to 1.100. It does not report a 200 Celsius reading at stock or the 1st Tweaked profile.


Now if the 200 reading was accurate, I would definitely NOT have a working CPU. So this can’t be accurate, right? So I am wondering if this is a software issue with HMMonitor or possibly even a sensor reading issue on my MB? It’s just really odd it only seems to happen while changing the CPU Ratio and the CPU Core Voltage. I would really like to leave it on this 2nd Tweaked Profile because the temps are running perfect and utilization is massively efficient while under load. It literally solved the issue I was having. However, the random 200 reading every time I open HMMonitor has me worried.

I am so close to having this solved.. and this happened 🙁

Should I be worried at all? Are my temps being reported accurately after the CPU ratio/voltage change? What is the likely cause of this so called faulty reading? Hardware? Software? CPU?

 

[JayGau]

You could try Ryzen Master and run something like Cinebench. If the temp stays reasonable it's just HWMonitor that is wrong. You could try to reinstall it to make sure it resets it's sensors.

But you should not have to do all this undervolting to get acceptable temperatures. Those spikes to 80c while doing pretty much nothing are not normal. The 9800X3D goes hot when loaded at 100% but when running regular tasks like internet browsing or even gaming it actually runs pretty cold. Mine at stock barely reaches 60c during gaming and the only way to get very high temp like 80-90 is to run heavy benchmarks (or compile shaders). And with curve optimizer at -10 it basically never reaches more than 75c even in heavy tasks.

So your undervolting is basically hiding a problem under the carpet. You also underclocked the CPU to 5000 MHz while it supposed to boost to 5200. You are losing performance.

Are you sure the cooler is properly installed with enough thermal paste? How is your airflow?

Speaking of the cooler, this is a general comment but why so many people go with the artic freezer 3? You can't even monitor the liquid temp with this cooler. One of the main reasons to buy an AIO is to avoid those annoying fan bursts by setting the fan curve with the liquid temperature instead of the CPU. The fans are not cooling down the CPU, they cool down the liquid, and since the liquid temp changes very slowly the fan speed should be set with that. Spending so much money on a pc and going for a cooler like the artic freezer 3 is beyond my comprehension.

 

[Murkxx2]

You could try Ryzen Master and run something like Cinebench. If the temp stays reasonable it's just HWMonitor that is wrong. You could try to reinstall it to make sure it resets it's sensors.

But you should not have to do all this undervolting to get acceptable temperatures. Those spikes to 80c while doing pretty much nothing are not normal. The 9800X3D goes hot when loaded at 100% but when running regular tasks like internet browsing or even gaming it actually runs pretty cold. Mine at stock barely reaches 60c during gaming and the only way to get very high temp like 80-90 is to run heavy benchmarks (or compile shaders). And with curve optimizer at -10 it basically never reaches more than 75c even in heavy tasks.

So your undervolting is basically hiding a problem under the carpet. You also underclocked the CPU to 5000 MHz while it supposed to boost to 5200. You are losing performance.

Are you sure the cooler is properly installed with enough thermal paste? How is your airflow?

Speaking of the cooler, this is a general comment but why so many people go with the artic freezer 3? You can't even monitor the liquid temp with this cooler. One of the main reasons to buy an AIO is to avoid those annoying fan bursts by setting the fan curve with the liquid temperature instead of the CPU. The fans are not cooling down the CPU, they cool down the liquid, and since the liquid temp changes very slowly the fan speed should be set with that. Spending so much money on a pc and going for a cooler like the artic freezer 3 is beyond my comprehension.

What liquid coolers would you recommend on this setup? I’m gonna check proper AIO fit and reapply thermal paste to see what happens.

Edit: Air flow is really good. 3 front 140mm fans that came with the case. I have a 120mm exhaust and 120mm fan on the bottom that is reverse to pull cold air from the bottom. Along with the 3 AIO 120mm fans mounted on top. My GPU barely reaches 55-60 while gaming and idles at 30-35.

 

[JayGau]

What liquid coolers would you recommend on this setup? I’m gonna check proper AIO fit and reapply thermal paste to see what happens.

Corsair H150i, Asus Rog, NZXT Kraken, anything with a usb 2.0 cable that allows you to monitor temperatures and control the fan and pump speed through a software. I personally don't like Asus because I hate their software (Armoury Crate), but Corsair iCUE and NZXT CAM are good.

 

[CountMike]

Setting fan and/or pump rpm according to liquid temps just delays rpm change, possibly even too long, a CPU can overheat during that time and you wouldn't know it and fans wouldn't react. Most modern MBs have BIOS option to delay up or down rpm change. 1-2 seconds is plenty enough. Liquid temps are only useful to troubleshoot cooling.
CPU is not just randomly spiking temps. it's ramping up or down with demand from OS, so OS is "guilty" of it.

 

[gamerk316]

I will say my 7900x acts much the same; it spikes surprisingly fast, but as long it stops at a reasonable limit you should be OK.

 

[Murkxx2]

Corsair H150i, Asus Rog, NZXT Kraken, anything with a usb 2.0 cable that allows you to monitor temperatures and control the fan and pump speed through a software. I personally don't like Asus because I hate their software (Armoury Crate), but Corsair iCUE and NZXT CAM are good.

So I bought a Kraken 360. But the manual states to plug the 3 pin connect to either CPU_FAN or AIO_PUMP. Is there going to be a difference between the two? Which one should I plug it in to? Thanks for the help!

 

[JayGau]

So I bought a Kraken 360. But the manual states to plug the 3 pin connect to either CPU_FAN or AIO_PUMP. Is there going to be a difference between the two? Which one should I plug it in to? Thanks for the help!

If the manual says either works then it doesn't matter. Just put it in the CPU fan header. Usually those are just to tell the board that you have a cooler installed, they don't do anything else. Does it have a USB 2.0 cable going from the block(pump) module? If it does then you should be able to control the pump and the fan speed through CAM.

 

[Murkxx2]

If the manual says either works then it doesn't matter. Just put it in the CPU fan header. Usually those are just to tell the board that you have a cooler installed, they don't do anything else. Does it have a USB 2.0 cable going from the block(pump) module? If it does then you should be able to control the pump and the fan speed through CAM.

Okay appreciate it! And Yes. It has a USB 2.0 from the module going to the MB. About to hook it all up and test it out. I reset my MB back to default before installing the new AIO. So I guess I’ll see how it behaves.

 

[Murkxx2]

https://imgur.com/a/M6DXMRa

View: https://imgur.com/a/M6DXMRa


Success. All fans running and idle temp looks pretty stable.

Gonna uninstall & redownload HWMonitor and a few other monitoring apps. Then I’ll check idle temps in windows for a bit.

Then I’ll start moving on to some stress testing.

From there, then I’ll start changing BIOS settings like enabling XMP at least.

Would you still recommend doing a negative curve optimizer of 10-20 and +200 if it’s stable?

 

[JayGau]

That's great! I would not say I "recommend" the curve optimizer especially if the temperatures are good, but it's always nice to get a significant temp improvement without any performance loss. I gained 15c with a -10 and it runs even better. For the +200 you can do it if you want to but this chip is already the fastest gaming CPU on the market so I don't think it's necessary.

Did you install CAM? I would really recommend to set your radiator fan curve with the liquid temperature. It should never go above 10 degrees relative to the room temperature. So if your room is at 25c, set the fans so the liquid doesn't go beyond 35c.

 

[Murkxx2]

That's great! I would not say I "recommend" the curve optimizer especially if the temperatures are good, but it's always nice to get a significant temp improvement without any performance loss. I gained 15c with a -10 and it runs even better. For the +200 you can do it if you want to but this chip is already the fastest gaming CPU on the market so I don't think it's necessary.

Did you install CAM? I would really recommend to set your radiator fan curve with the liquid temperature. It should never go above 10 degrees relative to the room temperature. So if your room is at 25c, set the fans so the liquid doesn't go beyond 35c.

I just installed CAM and did the Firmware update.

But I’m a bit confused on what you mean by radiator fan curve. CAM is showing two different settings, “Pump” and “Fan”.

https://imgur.com/a/vUYzpyU

View: https://imgur.com/a/vUYzpyU


Both are set to liquid temp source, is that what you mean? Or should I change them? There’s CPU, GPU, and Liquid temp sources.

Also, should I run the pump on a custom setting at 100%? And leave the fans on just one of the preset curves?

 

[JayGau]

I just installed CAM and did the Firmware update.

But I’m a bit confused on what you mean by radiator fan curve. CAM is showing two different settings, “Pump” and “Fan”.

https://imgur.com/a/vUYzpyU

View: https://imgur.com/a/vUYzpyU


Both are set to liquid temp source, is that what you mean? Or should I change them? There’s CPU, GPU, and Liquid temp sources.

Also, should I run the pump on a custom setting at 100%? And leave the fans on just one of the preset curves?

Yes this is exactly what I meant. The fan speed should be set as a function of the liquid temperature. You can try a preset and see how it does. If the curve is not aggressive enough and the liquid gets too warm then make a custom one.

For optimal performance the pump should always be set to 100% no matter the temperature.

 

[Murkxx2]

Just did a stress test in AIDA64 for 12 mins at 100% CPU Usage.

Is this pretty decent for stock settings?

Max CPU temp was 75.

Also, I can safely say during this test, there was absolutely no “winding up” from my fans. The constant up and down with fan RPMS seems to be pretty stable. I hope it stays this way while gaming.

https://imgur.com/a/vLlPKMj

View: https://imgur.com/a/vLlPKMj

 

[JayGau]

Yes, 75c at stock with AIDA64 looks pretty good. I think you are in a good shape.

With the fan curve set with the liquid temp, you are never going to hear the fan bursts anymore for the simple reason that the liquid temperature changes very slowly over time. Your fans will slowly speed up as the temp increase while gaming (so slowly that you will barely notice the change) until everything ramps off (stabilize). It's the main reason why I prefer AIO to air coolers. No annoying bursts when I just launch Firefox or install something.

 

[Murkxx2]

Yes, 75c at stock with AIDA64 looks pretty good. I think you are in a good shape.

With the fan curve set with the liquid temp, you are never going to hear the fan bursts anymore for the simple reason that the liquid temperature changes very slowly over time. Your fans will slowly speed up as the temp increase while gaming (so slowly that you will barely notice the change) until everything ramps off (stabilize). It's the main reason why I prefer AIO to air coolers. No annoying bursts when I just launch Firefox or install something.

Thanks so much for your help man. I’ll keep monitoring these temps. I think I’m gonna try XMP and a negative 10 curve optimizer like you said and check for stability.

Do you suggest switching PBO limits to Motherboard or Auto?

What about Platform Thermal Throttle Ctrl? Auto or 95?

Overall, Switching to this AIO and getting rid of the Artic Freezer seemed to be the difference and solved my issue. Really appreciate the recommendation on the Kraken. I’m loving the CAM software too.

 

[JayGau]

Thanks so much for your help man. I’ll keep monitoring these temps. I think I’m gonna try XMP and a negative 10 curve optimizer like you said and check for stability.

Do you suggest switching PBO limits to Motherboard or Auto?

What about Platform Thermal Throttle Ctrl? Auto or 95?

Overall, Switching to this AIO and getting rid of the Artic Freezer seemed to be the difference and solved my issue. Really appreciate the recommendation on the Kraken. I’m loving the CAM software too.

That’s great! I'm glad you are happy with your new cooler.

Personally, I have curve optimizer set on all core -10, PBO set to enabled and all its settings set to auto except for the scalar factor that I cranked up to the max (10x). After doing that and changing the Windows power plan to "performance" it works great with cold temp. But you can make your own experimentation and see what works best. And you should definitely enable expo, especially that your RAM is straight on the sweet spot for the 9800X3D (6000 MHz CL30).

 

[Murkxx2]

That’s great! I'm glad you are happy with your new cooler.

Personally, I have curve optimizer set on all core -10, PBO set to enabled and all its settings set to auto except for the scalar factor that I cranked up to the max (10x). After doing that and changing the Windows power plan to "performance" it works great with cold temp. But you can make your own experimentation and see what works best. And you should definitely enable expo, especially that your RAM is straight on the sweet spot for the 9800X3D (6000 MHz CL30).

I tried your OC recommendation and ran a 10min stability test with no issues. Temps stayed around the same at 100% load. Average 63.5 and Max of 76.

I assume the temps are the same due to the 10x scaler enabled this time but with also a negative 10 curve?

It seems I gained more performance at the cost of running the same temps. I guess that’s a win

 

[Murkxx2]

Although one thing I’ve noticed while gaming- one of my GPU fans seems to be not activating like the other 2. And CAM software isn’t reporting any GPU Fan Speed. Do you know what might be causing this? GPU temp seems fine

https://imgur.com/a/4cg3IS4

View: https://imgur.com/a/4cg3IS4

 

[JayGau]

If you disable PBO and leave the curve optimizer on do you see a lower temperature?

Also, the biggest temperature drop I have seen with curve optimizer is when I run really heavy tests, like Cinebench. It was going to 90+ at stock and doesn't even reach 80 after curve optimizer. But for less demanding tasks the change was not so extreme.

About your GPU fan, do you physically see it not spinning while the others do? If there's only one not spinning it's likely a problem with the fan and you should contact Gigabyte.

 

[Murkxx2]

If you disable PBO and leave the curve optimizer on do you see a lower temperature?

Also, the biggest temperature drop I have seen with curve optimizer is when I run really heavy tests, like Cinebench. It was going to 90+ at stock and doesn't even reach 80 after curve optimizer. But for less demanding tasks the change was not so extreme.

About your GPU fan, do you physically see it not spinning while the others do? If there's only one not spinning it's likely a problem with the fan and you should contact Gigabyte.

Yeah only 1 fan is not spinning. I even set the GPU fan speeds to manual in the Nvidia App. Confirmed 1 fan is not spinning 🙁

Edit: The fan spins with all the others during system startup and boot process. But once Windows 11 loads and I’m in, all 3 fans stop, which is normal due to low temp, but when I switch it to Manual and rev the Fan Speed up, only 2 of 3 spin. I sent this question into Gigabyte. I’m hoping it’s a driver issue or something because I might not get this card back or new one for awhile if i gotta RMA it 🙁

 

[JayGau]

Yeah only 1 fan is not spinning. I even set the GPU fan speeds to manual in the Nvidia App. Confirmed 1 fan is not spinning 🙁

Edit: The fan spins with all the others during system startup and boot process. But once Windows 11 loads and I’m in, all 3 fans stop, which is normal due to low temp, but when I switch it to Manual and rev the Fan Speed up, only 2 of 3 spin. I sent this question into Gigabyte. I’m hoping it’s a driver issue or something because I might not get this card back or new one for awhile if i gotta RMA it 🙁

Yeah indeed, a replacement would be very hard to get with the current stocks. If the fan spins at startup but doesn't when the card warms up it might be the heat sensor that doesn't communicate with this particular fan. Hopefully Gigabyte will have a solution, but be prepared to the idea of having to ship it to them.