9800X3D, hotter than the first day, pulling more wattage?

[Pets4Ever]

When I first installed the 9800X3D I ran Cinebench 2024 and my temps were ~41°C idle, and maximum 82°C in Cinebench. I think I remember the CPU package wattage maxed out at 120 or 130 W. Even with PBO turned on, I got higher frequencies but I don't think the wattage went up that day.

Fast forward to yesterday, idling at about 50°C, averaged 90°C in Cinebench, and peaked at about 92°C. 140 watts package power now. So I think, PBO must finally be working? I turn it off. Cinebench score goes down 30 points, temps stay the same. Later played a game where temp spikes hit the thermal limit of 95°C once or twice briefly before dropping down again. (Helldivers 2 is an unoptimized mess)

My cooling setup: Noctua NH-D15, 3 front intakes, 1 rear exhaust, 1 top exhaust(directly above CPU), Case is Fractal Define S, Arctic MX-6 paste sold and fulfilled by Amazon (fixed)


The ambient temp was definitely higher than the first day, but I am unsure of how much higher. I know it was 71°F outside when I checked before bed but my room was warmer. That doesn't explain the wattage thing though. And I don't think this CPU should be getting so close to it's thermal limit with this cooler. The paste I used was MX-6 shipped and sold by amazon.

When I get home I plan to reset the bios, maybe reflash it (it's already up to date). I ordered an Arctic Freezer III Pro, but I'm not convinced that my cooler is entirely to blame here.

What do you guys think is going on?

 

[phanbuey]

So noctua air cooler relies on cold air being available - Cinebench average at 90C and then helldivers at 95C means you're cooling the CPU with graphics card heat (helldivers is 50% of the heat of a cinebench except for shader compilation).

Test with the side panel off and let us know what happens to the temps.

You also seem to have all of your fans intaking? How does the air leave the case?

 

[JustBenching]

When I first installed the 9800X3D I ran Cinebench 2024 and my temps were ~41°C idle, and maximum 82°C in Cinebench. I think I remember the CPU package wattage maxed out at 120 or 130 W. Even with PBO turned on, I got higher frequencies but I don't think the wattage went up that day.

Fast forward to yesterday, idling at about 50°C, averaged 90°C in Cinebench, and peaked at about 92°C. 140 watts package power now. So I think, PBO must finally be working? I turn it off. Cinebench score goes down 30 points, temps stay the same. Later played a game where temp spikes hit the thermal limit of 95°C once or twice briefly before dropping down again. (Helldivers 2 is an unoptimized mess)

My cooling setup: Noctua NH-D15, 3 front intakes, 1 rear intake, 1 top intake (directly above CPU), Case is Fractal Define S, Arctic MX-6 paste sold and fulfilled by Amazon


The ambient temp was definitely higher than the first day, but I am unsure of how much higher. I know it was 71°F outside when I checked before bed but my room was warmer. That doesn't explain the wattage thing though. And I don't think this CPU should be getting so close to it's thermal limit with this cooler. The paste I used was MX-6 shipped and sold by amazon.

When I get home I plan to reset the bios, maybe reflash it (it's already up to date). I ordered an Arctic Freezer III Pro, but I'm not convinced that my cooler is entirely to blame here.

What do you guys think is going on?

Similar experience with my u12a - it was scorching hot for me as well. I have a feeling that noctua just doesn't like the x3d chips

 

[Pets4Ever]

So noctua air cooler relies on cold air being available - Cinebench average at 90C and then helldivers at 95C means you're cooling the CPU with graphics card heat (helldivers is 50% of the heat of a cinebench except for shader compilation).

Test with the side panel off and let us know what happens to the temps.

You also seem to have all of your fans intaking? How does the air leave the case?

IDK why I wrote that, the rear two are exhaust. Isn't 92° C peak kind of high for just Cinebench though? The GPU doesn't figure into that situation really

Similar experience with my u12a - it was scorching hot for me as well. I have a feeling that noctua just doesn't like the x3d chips

This makes some sense as I notice that Noctua now offers three variants of the NH-D15 G2 with different levels of convexity or concavity to the heatsink.

I'm still a little nonplussed by the wattage thing though. I remember being kinda surprised it didn't pull more juice when I first got it, but figured I just had to work on the PBO config. Now it seems to pull more regardless of PBO config. I did monitor the Vcore voltages during the test though and they never exceeded safe levels.

 

[rv8000]

NH-D15 shouldn’t have any issues keeping temps well under control.

When was the cpu/temps first installed or recorded? If you didn’t write it down or take a SS how do you know there’s actually a difference in TDP?

Have you updated the bios since then? What
Motherboard? What settings did you change in PBO? Do you have any CO settings? Increasing the scalar or frequency limit will add voltage and thus power. Newer bios can change a number of voltages that will change TDP and PBO.

Your rear fan should be set to exhaust.

Zen 5 CPUs are designed to boost until their thermal limit (95c). It’s more normal to see games put the cpu in the 60-70c range, but heavier CPU games can definitely have temps in the 80s.

*Edit

92c is not high for CB, especially if you keep hammering MT tests in warmer/summer weather with no AC.

*Edit 2

Fractal Define S is kind of a hotbox, and is not relatively good for airflow/aircooling.

 

[JustBenching]

This makes some sense as I notice that Noctua now offers three variants of the NH-D15 G2 with different levels of convexity or concavity to the heatsink.

I'm still a little nonplussed by the wattage thing though. I remember being kinda surprised it didn't pull more juice when I first got it, but figured I just had to work on the PBO config. Now it seems to pull more regardless of PBO config. I did monitor the Vcore voltages during the test though and they never exceeded safe levels.

I was hitting 95c running CBR23 at around 140-150 watts too. They are a bit hard to cool but since you already ordered a new cooler no need to puzzle yourself, just wait for your cooler to arrive and see if there are any improvements

EG1. PBO doesn't make a difference when you are running CBR. The CPU already hits the power limit (and the temperature limit in our case) without PBO, so PBO ends up doing nothing.

 

[_roman_]

In my point of view: Please get a proper new Thermal paste and put on the new thermal paste please.

NH-D15 shouldn’t have any issues keeping temps well under control.

I have a second hand Noctua NH-D15 lying around. I bought it for 45€

Any arctic Liquid freezer II 360 keeps any processor better under control. I doubt the influence from my previous MSI B550 gaming edge wifi and fractal design mesfhify 2 with nocuta nh-d15 does influence much the result with the same ryzen 5800x a be quiet pure base 500 and an arctic liquid freezer II 360 AIO. The be quiet case has much worse air flow by design, cheap, bad case.

In my point of view, 3 or 4 years ago many sold NH-D15 and similar air coolers and went for an aio. The invoice from teh seller has shown me, he paid on amazon.de in austria 120€ for that cooler. At that time there were more than 5 nh-d15 for sale.

--

Regarding post #1

I miss which time frame.
How many weeks and how many monts please? e.g. 5.4 months and 47 weeks

Which thermal paste? When was it the last time changed?

Arctic MX-6

I bought MX-6 used it once also on the NH-D15 and sold the package than. Igor claimed Arctic may had issues. Much worse as my old package of Noctua Nt-h1

--

I would not spend 50 € to upgrade the NH-D15 with the new Fan set to get a similar NH-D15 Generation 2 tower cooler. I think they claimed they changed a bit the tower cooler also.

I ordered an Arctic Freezer III Pro, but I'm not convinced that my cooler is entirely to blame here.

Good Choice. I assume you bought 360 or 420mm, right?

 

[Pets4Ever]

NH-D15 shouldn’t have any issues keeping temps well under control.

When was the cpu/temps first installed or recorded? If you didn’t write it down or take a SS how do you know there’s actually a difference in TDP?

Have you updated the bios since then? What
Motherboard? What settings did you change in PBO? Do you have any CO settings? Increasing the scalar or frequency limit will add voltage and thus power. Newer bios can change a number of voltages that will change TDP and PBO.

Your rear fan should be set to exhaust.

Zen 5 CPUs are designed to boost until their thermal limit (95c). It’s more normal to see games put the cpu in the 60-70c range, but heavier CPU games can definitely have temps in the 80s.

*Edit

92c is not high for CB, especially if you keep hammering MT tests in warmer/summer weather with no AC.

*Edit 2

Fractal Define S is kind of a hotbox, and is not relatively good for airflow/aircooling.

I've only had the CPU for a couple of weeks. I didn't write the original TDP down, but maybe I took a screenshot, I'll check later. I do remember being disappointed it wasn't drawing more power though, and I connected the second CPU power cable to the motherboard even though the nominal TDP of the chip is lower than the amount that one of those cable scan deliver. I did not see a difference from installing the cable.


The motherboard was updated to the latest bios immediately after boot, I had it downloaded ahead of time. It's the MSI MAG X870E TOMAHAWK WIFI. I limited the scalar to 1x because people told me it was generally not needed, and perhaps inadvisable. I did a negative offset of 20 all-core which didn't seem to have a big effect. I allowed an extra 200Mhz of boost clockspeed too. However, note that when my temp issues arose yesterday I disabled PBO completely and did not notice any reduction in heat.

Part of me wonders if certain BIOS settings are "sticking" even after they are turned off, which is why I'm going to flash the BIOS this evening. Maybe the scalar was active when it said it wasn't, for example.

The rear and top fans are exhaust, I misspoke earlier.

Regarding boosting to thermal limit, is this different than how it's implemented on the intel side? In the past my 12700k never got that close to it's thermal limit in Cinebench

I don't just sit there running benchmarks back-to-back, we're talking 90 degrees almost immediately in CB

Regarding the Fractal S, I'm looking at replacing it because the airflow could be a lot better on the back panel, and I think the top grate is perhaps too restrictive for good liquid cooling too. Open to suggestions for a case that is least restrictive on a top radiator.

PBO doesn't make a difference when you are running CBR. The CPU already hits the power limit (and the temperature limit in our case) without PBO, so PBO ends up doing nothing.

Maybe not if you just turn it on, but if you use it to increase the boost clock it does have some effect. Last night when I disabled it I lost about 60 points in CB R24 although my temps did not improve.

In my point of view: Please get a proper new Thermal paste and put on the new thermal paste please.



I have a second hand Noctua NH-D15 lying around. I bought it for 45€

Any arctic Liquid freezer II 360 keeps any processor better under control. I doubt the influence from my previous MSI B550 gaming edge wifi and fractal design mesfhify 2 with nocuta nh-d15 does influence much the result with the same ryzen 5800x a be quiet pure base 500 and an arctic liquid freezer II 360 AIO. The be quiet case has much worse air flow by design, cheap, bad case.

In my point of view, 3 or 4 years ago many sold NH-D15 and similar air coolers and went for an aio. The invoice from teh seller has shown me, he paid on amazon.de in austria 120€ for that cooler. At that time there were more than 5 nh-d15 for sale.

--

Regarding post #1

I miss which time frame.
How many weeks and how many monts please? e.g. 5.4 months and 47 weeks

Which thermal paste? When was it the last time changed?



I bought MX-6 used it once also on the NH-D15 and sold the package than. Igor claimed Arctic may had issues. Much worse as my old package of Noctua Nt-h1

--

I would not spend 50 € to upgrade the NH-D15 with the new Fan set to get a similar NH-D15 Generation 2 tower cooler. I think they claimed they changed a bit the tower cooler also.



Good Choice. I assume you bought 360 or 420mm, right?

I was under the impression that Arctic MX-6 was good paste, and I used a freshly opened tube. Going by this article, it's within a degree or so of most of the top pastes: https://www.tomshardware.com/best-picks/best-thermal-paste


I've only had the CPU installed for a week or two so the paste is quite fresh. My usage has been light because I was dealing with a hand injury so I haven't been working the CPU hard every day since I installed it either.

I believe the NH-D15 G2 is significantly better based on the Gamers Nexus tests. It approaches AIO levels of performance, but it costs twice what the Arctic AIO cost me. I did choose the 360 version and it is the updated PRO with thicker rad and better fans.


In defense of the old NH-D15, it always kept my 12700k cool. It idled at 30°C and basically topped out at 80°C. Nominally, the 12700k has a higher TDP than the 9800X3D.

 

[JustBenching]

Maybe not if you just turn it on, but if you use it to increase the boost clock it does have some effect. Last night when I disabled it I lost about 60 points in CB R24 although my temps did not improve.

Even with the +200 option it didn't really do much for me - was already thermal throttling. And regardless of your cooler, PBO +200 is definitely not worth running, neither for games or for other workloads, power draw goes 50% up while you gain around 4% performance.

In defense of the old NH-D15, it always kept my 12700k cool. It idled at 30°C and basically topped out at 80°C. Nominally, the 12700k has a higher TDP than the 9800X3D.

Same thing, my u12a is handling 12900k and 14900k like a champ but it was scorching with the 9800x 3d.

 

[Dorek]

I have the same cooler but mine tops out at around 80+ which isnt really super high for such a tough bench but i used some ancient arctic silver 5 paste whatever.

 

[Pets4Ever]

Even with the +200 option it didn't really do much for me - was already thermal throttling. And regardless of your cooler, PBO +200 is definitely not worth running, neither for games or for other workloads, power draw goes 50% up while you gain around 4% performance.


Same thing, my u12a is handling 12900k and 14900k like a champ but it was scorching with the 9800x 3d.

That's part of what was so weird last night, I disabled PBO (which reverted by boost clock change) and yet the power draw remained the same.

 

[Carlyle2020hs]

Yo OP, how about more facts ?

PBO 200?
Skalar 10?
LLC ?

And yes, balls to the wall a 9800x3d can hit over 90C .

Even with a G2, 3 fans, PTM and frame.

Its just that most sane people run the 9800x3d shackled.
And then they think that small air coolers are fine for any usecase imaginable ...

 

[Pets4Ever]

I have the same cooler but mine tops out at around 80+ which isnt really super high for such a tough bench but i used some ancient arctic silver 5 paste whatever.

Huh, maybe MX-6 has manufacturing inconsistency or something. I don't think i used too little either since I manually spread it out.

Yo OP, how about more facts ?

PBO 200?
Skalar 10?
LLC ?

And yes, balls to the wall a 9800x3d can hit over 90C .

Even with a G2, 3 fans, PTM and frame.

Its just that most sane people run the 9800x3d shackled and then tell everybody that small air coolers are fine...

This is sprinkled through the thread but:

PBO +200 boost clock
Scalar x1
LLC - don't know, didn't touch it

I disabled PBO last night and it didn't change temp or power draw which seems wrong. I plan to flash the BIOS tonight because I am suspicious, but I also ordered an Arctic Freezer III Pro 360.

I do find it funny how I keep hearing that the 9800X3D is easy to cool (elsewhere). Hasn't been my experience so far. In the beginning I thought 40°C was a bad idle temp. Now I'd be happy to see that again...

I have the same cooler but mine tops out at around 80+ which isnt really super high for such a tough bench but i used some ancient arctic silver 5 paste whatever.

Hmm interesting. And our cases are pretty similar for airflow too.

 

[JustBenching]

Huh, maybe MX-6 has manufacturing inconsistency or something. I don't think i used too little either since I manually spread it out.


This is sprinkled through the thread but:

PBO +200 boost clock
Scalar x1
LLC - don't know, didn't touch it

I disabled PBO last night and it didn't change temp or power draw which seems wrong. I plan to flash the BIOS tonight because I am suspicious, but I also ordered an Arctic Freezer III Pro 360.

I do find it funny how I keep hearing that the 9800X3D is easy to cool (elsewhere). Hasn't been my experience so far. In the beginning I thought 40°C was a bad idle temp. Now I'd be happy to see that again...


Hmm interesting. And our cases are pretty similar for airflow too.

It's not your paste, I was using Liquid metal and results were the same.


I was idling at 47C (but fans were off), your idle temp is probably high because of IO die temps. Check on hwinfo, your cores most likely are sitting lower (30c). As I've said before, PBO won't do anything for running Cinebench since you are already hitting power and thermal limits.

It's "easy" to cool if you keep it between 80-100w, but when you cross 100w yeah, it goes crazy. I mean it's just 8cores so 140w on 8 cores with bad IHS conductivity will result in high temps so...

You can check this thread with my experiences on the 3d - starting at page 6

Thread '9800x 3d vs 12900k - Battle of the Century'

Nov 12, 2024

So, since I'm "upgrading" to the 9800x 3d, im getting some footage together and ill end up doing a side by side video between 12900k and 9800x 3d. Depending on how much time I have, ill try to compare both stock and both OCed with different OSes. Sadly I had to restart the whole process a couple of times cause I figured out win 11 24h2 doesn't play nice with Intel, at all. Im seeing 20-25% performance difference between 24h2 and win 10.

If you got any game suggestions (no multiplayer games lads, they are really not easy to do) im all ears. I have already tested CP2077, TLOU, Hogwarts...

• JustBenching
• Replies: 479
• Forum: Overclocking & Cooling

• 

 

[freeagent]

I do find it funny how I keep hearing that the 9800X3D is easy to cool. Hasn't been my experience so far.

Its like when people say the 5800X3D was a hot running chip.. I could do 140w with it and no fans on my CPU cooler running semi passive in my old Torrent C

 

[_roman_]

My opinion.

Personally I would buy a 420 mm artic liquid freezer II. And get a proper case where it fits. Especially as you wrote you want to replace the case anyway. You may also read all those posts abut the arctic fans for the arctic liquid freezer III. There is something about the fans tehy use and the fan performance. I assue you want to use the stock fans.
I think, not sure, the arctic liquid freezer III pro version uses worse fans. Maybe the noise or perforamnce was worse according to hardwareluxx. I want to ask you to research that yourself please.

 

[rv8000]

Yo OP, how about more facts ?

PBO 200?
Skalar 10?
LLC ?

And yes, balls to the wall a 9800x3d can hit over 90C .

Even with a G2, 3 fans, PTM and frame.

Its just that most sane people run the 9800x3d shackled.
And then they think that small air coolers are fine for any usecase imaginable ...

An NH-D15 is not a small air cooler… and for the thousandth time, Zen 5 chips will continue to boost if there is thermal headroom up to 95c, entirely normal behavior and very different from intel. I’m unsure why people equate CB and the like to realistic scenarios for heat/temps when that’s far from the truth when using a 9800X3D for… ya know… gaming, or not slamming your pc with benchmarks/unrealistic loads.

Huh, maybe MX-6 has manufacturing inconsistency or something. I don't think i used too little either since I manually spread it out.


This is sprinkled through the thread but:

PBO +200 boost clock
Scalar x1
LLC - don't know, didn't touch it

I disabled PBO last night and it didn't change temp or power draw which seems wrong. I plan to flash the BIOS tonight because I am suspicious, but I also ordered an Arctic Freezer III Pro 360.

I do find it funny how I keep hearing that the 9800X3D is easy to cool (elsewhere). Hasn't been my experience so far. In the beginning I thought 40°C was a bad idle temp. Now I'd be happy to see that again...


Hmm interesting. And our cases are pretty similar for airflow too.

There is nothing wrong with mx-6, great paste; there are reports of counterfeits being sold by resellers however.

Funny you mention having an X870 Tomahawk. My friend is using one with a 9700X and temps have always been suspiciously high at completely default settings (I wonder if they have an issue with sensor reporting tbh).

There’s no reason to add/use/enable the scalar setting. Unless you’re doing hwbot runs and extreme OCs, it’s not going to provide any tangible benefit. Both the scalar and frequency offset push the voltage on the V/F curve higher (which is why it’s most common to use -CO to adjust the voltage back down).

I don’t have much hands on time with that board and it’s PBO settings, but it may be advisable to just set optimize defaults and/or clear cmos to make sure all the PBO settings have been cleared. Certain manf. also have multiple menus in which PBO settings can be changed, could be some remnant setting.

An AIO is going to provide minimal benefits and hurt VRM/Memory cooling, and unless you do a lot of productivity work, you’re just throwing money at problems you don’t have; if the noctua cooler is within it’s return period that’s a different story.

From personal experience, at least in regards to CB benching, +100 frequency offset provides more consistent clocks & better score (results will vary, silicon lottery, blah blah blah).

 

[JustBenching]

and for the thousandth time, Zen 5 chips will continue to boost if there is thermal headroom up to 95C

No, they will not. If I put one in the freezer will it keep boosting until it hits 95c? Come on man, Zen chips work exactly like any other CPU ever created in the history of mankind. They boost until they either hit a power, thermal or clockspeed limit. It's just that because they have kinda bad heat dissipation, they are more likely to hit the thermal limit rather than the power limit.

 

[rv8000]

No, they will not. If I put one in the freezer will it keep boosting until it hits 95c? Come on man, Zen chips work exactly like any other CPU ever created in the history of mankind. They boost until they either hit a power, thermal or clockspeed limit. It's just that because they have kinda bad heat dissipation, they are more likely to hit the thermal limit rather than the power limit.

If they’re not hitting power or frequency limitations they will.

 

[JustBenching]

If they’re not hitting power or frequency limitations they will.

Well sure, but that's the same way any CPU works.

 

[phanbuey]

AMD chips do have massive heat dissipation issues in comparison to intel - so amd 150W = 90-95C on an AIO. Intel 150W is like 70C on a $35 tower air cooler. The 9xxx series really like a cold coldplate -- so intake on the AIO is better than exhaust - they don't put out much heat but the chip runs hot.

@Pets4Ever - hitting 90C instantly CB is pretty normal tbh especially with PBO on and no curve offset. Hitting 95C during Gaming - if you're doing shader compilation and the case is already hot, also normal. If you're hovering at 95C in game then you have airflow problems in the case.

 

[rv8000]

Well sure, but that's the same way any CPU works.

No it’s not. Intel will limit boost by specific durations as well. I’m not gonna waste time with semantics, especially with you.

 

[JustBenching]

AMD chips do have massive heat dissipation issues in comparison to intel - so amd 150W = 90-95C on an AIO. Intel 150W is like 70C on a $35 tower air cooler. The 9xxx series really like a cold coldplate -- so intake on the AIO is better than exhaust - they don't put out much heat but the chip runs hot.

@Pets4Ever - hitting 90C instantly CB is pretty normal tbh especially with PBO on and no curve offset. Hitting 95C during Gaming - if you're doing shader compilation and the case is already hot, also normal. If you're hovering at 95C in game then you have airflow problems in the case.

You are playing POE 2 right? Well going in and out of town for example made my cooler go crazy cause cpu was hitting 100-120 watts. So even for gaming hitting 90C+ isn't that uncommon, but of course depends on the game. With PBO off it generally runs kinda cool for gaming - ~70C which is perfectly fine.

No it’s not. Intel will limit boost by specific durations as well. I’m not gonna waste time with semantics, especially with you.

Yes it is but okay, let's pretend otherwise

 

[Pets4Ever]

An AIO is going to provide minimal benefits and hurt VRM/Memory cooling, and unless you do a lot of productivity work, you’re just throwing money at problems you don’t have; if the noctua cooler is within it’s return period that’s a different story.

How would an AIO hurt VRM and Memory cooling? Shouldn't the overall flow of air in my case be improved with that big tower cooler out of the way?

Plus, the heat from the CPU will no longer be in the case so the GPU should stay cooler.

There is a VRM fan on this AIO I ordered but a lot of people say it isn't even necessary.

 

[JustBenching]

How would an AIO hurt VRM and Memory cooling? Shouldn't the overall flow of air in my case be improved with that big tower cooler out of the way?

An AIO doesn't hurt VRM and memory cooling per se. But the aircooler helps with that because you have airflow right next to the components, with an AIO you don't.