3900x with 360 Aio 81 degrees?

Epyon

[H]ard|Gawd
Joined
Oct 25, 2001
Messages
1,185
So I got a 3900x running in a maingear case with the side panels off and i get 81.5 degrees in Hwin64 latest update. Now I am pounding the shit out of it with Keyshot but this does not seem right at all.
see pics:



and let me add that this prorcessor is Fast as hell. 1800X<3900X holy crap guys. I am rolling fast and hard with these renders.
 
Thank you for the 2nd look. I don't over clock or play with any settings. I am really really happy that i took a 7 hour render (old Pc) to a 90 minute one. I mean fucking hell this thing is rolling. Fuck yeah i love this thing.
 
What kind of 360 AIO are we talking here? They are not created equally. It could be seated poorly, could need a pump speed adjustment or a fan speed adjustment.
 
https://www.amazon.com/Enermax-LIQTECH-Liquid-Cooler-ELC-LTTO360-TBP-W/dp/B07N8XDKQ9

I don't know how to check fan speed sorry. I am letting it spin at default i guess. How can i see or change it?

My motherboard has a dedicated AIO pump header, but you can plug the pump into any fan header and see the speed in BIOS where you can most like set fan curves. Same goes for the fans. Your motherboard probably also has Windows software for controlling all that from within Windows. I looked at the manual for your cooler and it's very simple, just plug in the pump and fan header. You did remove the protective film from the cold plate, right? ;-)

My Corsair AIO connects to the mobo via USB, much more complicated.
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
My motherboard has a dedicated AIO pump header, but you can plug the pump into any fan header and see the speed in BIOS where you can most like set fan curves. Same goes for the fans. Your motherboard probably also has Windows software for controlling all that from within Windows. I looked at the manual for your cooler and it's very simple, just plug in the pump and fan header. You did remove the protective film from the cold plate, right? ;-)

My Corsair AIO connects to the mobo via USB, much more complicated.

Generally, the AIO's connect via both USB and the fan header. The temps are about right though. A 360 AIO may be better than most air or 240 AIO coolers, but its a far cry from a custom loop and 12c/24t CPU's produce a ton of heat as stated earlier.
 
yeah i did remove the protective film on the copper block. I am in texas and its 105 degree or higher outside my car read 111 yesterday and right now the house is 76 degrees at 10:30 at night. i did remove the dust filter and its down to 78 degree or so. i would say from the noise of the fans they are going full speed lol.
 
I'm running a 280mm AIO (Corsair H115i Pro) and my temps are between 80-82C when doing a h265 encode which loads all 24 threads. That's with the pump set to "extreme" and a custom fan curve.
 
I have a fully custom looped PC.

Xspc raystorm peak op temp is 84c

Corsair XC7 waterblock op temp is 80c avg

These are full load temps. Like if I'm hand braking x4 programs at same time etc...

It's a really small IHS for so many cores. Threadripper 12 core runs much cooler due to having the heat dissipated more evenly and less concentrated.
 
I have a fully custom looped PC.

Xspc raystorm peak op temp is 84c

Corsair XC7 waterblock op temp is 80c avg

These are full load temps. Like if I'm hand braking x4 programs at same time etc...

It's a really small IHS for so many cores. Threadripper 12 core runs much cooler due to having the heat dissipated more evenly and less concentrated.

Not on an AIO it doesn't. At least mine didn't.
 
Not on an AIO it doesn't. At least mine didn't.

Well much is very vague. What I mean is with my raystorm threadripper block I was seeing temps around 70c full load and that was a 2950x. But you could run lower voltage on those and not take such a penalty to performance as these new 7nm chips.
 
Well much is very vague. What I mean is with my raystorm threadripper block I was seeing temps around 70c full load and that was a 2950x. But you could run lower voltage on those and not take such a penalty to performance as these new 7nm chips.

I tried both the 1950X and the 2920X on an AIO and under full load I would see temperatures around 78c using a Floe Riing RGB 360 TT Premium Edition, which at the time of its release was one of the best ones out there. Using a custom loop, the results were far better obviously. I think my temps were about what yours were.
 
I tried both the 1950X and the 2920X on an AIO and under full load I would see temperatures around 78c using a Floe Riing RGB 360 TT Premium Edition, which at the time of its release was one of the best ones out there. Using a custom loop, the results were far better obviously. I think my temps were about what yours were.

I'm guessing you were ocing as they throttle clocks at 68C when stock, in which case it's not really a like for like comparison. A 240mm aio (h100i platinum) keeps my 1920x at 68C with clocks dropping from max all core of 3.7 to after about 5 minutes to 3.65 (so only 50 mhz throttle). If I oc it obviously can go way higher, bit I think stock for stock the tr's run way cooler, especially given their 180W tdp rating.
 
I think his temp looks fine. It could improved with a good loop sure. For ex. my 3900x tops out at 77c under prime avx but otherwise maxes around 55c-60c. Two blocks with two 480mm rads, a lot of rad per block.
 
let pbo do it's thing. undervolt about 0.1 and you'll see a 10% drop in max and average temps and if you do the undervolt right, you'll see no drop in performance.

i have a 240mm aio and it tops out in the mid 70's with a 0.1uv (though i should say I haven't run something that maxes it for hours at a time yet to see if it gets back into the 80's on the current setup)

I am adding another 240mm radiator and a 200gph pump (though likely run at about 70% full power) to my system this weekend. I'm hoping to get the temps to fully max out at around 70c. So two pumps, two radiators 1 cpu block.
 
let pbo do it's thing. undervolt about 0.1 and you'll see a 10% drop in max and average temps and if you do the undervolt right, you'll see no drop in performance.

i have a 240mm aio and it tops out in the mid 70's with a 0.1uv (though i should say I haven't run something that maxes it for hours at a time yet to see if it gets back into the 80's on the current setup)

I am adding another 240mm radiator and a 200gph pump (though likely run at about 70% full power) to my system this weekend. I'm hoping to get the temps to fully max out at around 70c. So two pumps, two radiators 1 cpu block.

PBO or PB2?
 
Both. Why would you disable one if you're going to use either at all ?

enable both options (dont leave it to auto since playing with core voltage may disable the feature in auto mode), do the undervolt ... You may need to also adjust vrm settings when doing an undervolt but either way. You do the undervolt and you will drop 10% of the heat output compared to stock. Meaning you should see temps fall 10% unless your cooling system is still not capable of dissipating the wattage.

edit : by both, I mean in bios's you'll usually see performance boost in multiple places under settings and they appear to have independent flags even though they're labeled the same. Not sure if that equates to the difference between precision boost and precision boost overdrive settings or just the same setting seen from two submenus but everywhere that talks about precision boost, change it from auto to enable.
 
You can run the thing to 95c.

Anything under and it's fine. People have this unfounded internet mass hysterical belief that anything over 70c is going to detonate like a nuclear weapon. I've ran my server to 87c for 24/7 a xeon e5 quad core for a month straight once. It's still running 7byears later 24/7 365. Plex and all works fine, it's my nas and host VMs. Its air cooled on a supermicro board. Hits 87 to 90c for hours on end sometimes doing work.

Your aio is working fine if its keeping temps under thermal absolute limit.

Stop worrying if you ever were.

As far as maintaining PBO high clocks, sure 70c is going to be better than 80c.

As far as full load on a 3900x at 50c?? I'm going to call total chocolate sauce on that Unless its running like 1.0v core. Remember that even if a processor is at 100% usage it doesnt mean its 100% maxed out. It's just representative of all the cores being engaged. You run some prime 95 small fft and now were talking max load to the hilt.

So saying 50c full loaded? That's not true in the least unless your running a loop chiller.

My Athlon 200ge dual core with an AIO even hit 60c max load.
 
Last edited:
YYou run some prime 95 small fft and now were talking max load to the hilt.

So saying 50c full loaded? That's not true in the least unless your running a loop chiller.

My Athlon 200ge dual core with an AIO even hit 60c max load.

Yea, and with regards to P95, there's a big difference between running it w/ or w/o AVX. P95 w/ AVX will make crumble lotsa overclocks and cooling.
 
Both. Why would you disable one if you're going to use either at all ?

enable both options (dont leave it to auto since playing with core voltage may disable the feature in auto mode), do the undervolt ... You may need to also adjust vrm settings when doing an undervolt but either way. You do the undervolt and you will drop 10% of the heat output compared to stock. Meaning you should see temps fall 10% unless your cooling system is still not capable of dissipating the wattage.

edit : by both, I mean in bios's you'll usually see performance boost in multiple places under settings and they appear to have independent flags even though they're labeled the same. Not sure if that equates to the difference between precision boost and precision boost overdrive settings or just the same setting seen from two submenus but everywhere that talks about precision boost, change it from auto to enable.

Enabling PBO simply tells PB2 to get its PPT, EDC and TDC values from the motherboard and not the CPU. However, there is a seriously compelling reason not to use PBO at all. Enabling PBO can actually have a negative impact on your clocks compared to PB2. I've seen the 3900X essentially not boost with PBO or boost lower than it does with PB2 and I'm not alone there. Gamer's Nexus experienced the very same thing.

As for the BIOS, there are actually duplicate settings which are linked. This is due to AMD mandating a tweaking menu that's essentially the same on all the BIOS implementations regardless of manufacturer. However, motherboard makers left their own tuning menus intact.
 
maybe that distinction is only made in gamer motherboards or certain brands? I dont think i have distinct options for that or it's all just pb2 that I've enabled. I do know i hit the max frequencies on the golden cores (4.5-4.6Ghz) and I hit all core 4.2Ghz sustained. and all the 40+ benchmarks reflect that what I'm seeing is what is actually occurring (no false frequency readings).
 
PB2 is enabled by default. You don't have to do anything with it. PBO on the other hand requires you enter specific menus and enable it. There is also PBO+AutoOC, etc. The issue with the 3900X specifically is that the CPU's are already near the edge of what they can do that these things aren't likely to get you anymore than what PB2 already gives you. If you are on a 3600, 3600X, or even a 3700X, your experience may differ. But you won't get far with a 3900X.
 
Both. Why would you disable one if you're going to use either at all ?

enable both options (dont leave it to auto since playing with core voltage may disable the feature in auto mode), do the undervolt ... You may need to also adjust vrm settings when doing an undervolt but either way. You do the undervolt and you will drop 10% of the heat output compared to stock. Meaning you should see temps fall 10% unless your cooling system is still not capable of dissipating the wattage.

edit : by both, I mean in bios's you'll usually see performance boost in multiple places under settings and they appear to have independent flags even though they're labeled the same. Not sure if that equates to the difference between precision boost and precision boost overdrive settings or just the same setting seen from two submenus but everywhere that talks about precision boost, change it from auto to enable.

I just got Windows installed and I'm using an H100i platinum SE in push pull. I think I'm going to offset cpu to -0.1 like you do as this Asus board pushes the high volts at stock. Then I'll get my RAM stable, hopefully at 3800. I've got a b die kit that's XMP is 4000 CAS 19 1.35v, but I'm a little worried about as it won't boot at those settings on auto. Also, I can't boot with fclck of 1900 yet. However, I haven't messed with any voltage yet. I'm going to use the Ryzen RAM utility and see if I can dial in 3800 at CAS 16. I'll report back temps.
 
These temps are fine! I was worried about my temps being mid 80's to sometimes spiking to low 90's on a good 240 AIO (running custom AVX workloads) so i got a NHD15S thinking my AIO was dying. Well my temps are about the same maybe slightly worse now...

If i am running non AVX 100% load, i get mid to high 70's
This is with running -.1 offset in bios BTW and stock clocks on the 3900x

I can also report that my 1950x doesn't get nearly as hot running similar cooling and the same type of workloads so i agree it has alot to do with how the chiplets are connecting to the cooler.
while my temps are higher the actual power at the socket is much lower on the 3900x than my 1950x so that is the only variable that makes sense.

I have come to accept the high temps as the new norm. If AMD thinks its ok to run this hot then who am i to argue, and i can't judge this generations temps by the previous one.
 
These temps are fine! I was worried about my temps being mid 80's to sometimes spiking to low 90's on a good 240 AIO (running custom AVX workloads) so i got a NHD15S thinking my AIO was dying. Well my temps are about the same maybe slightly worse now...

If i am running non AVX 100% load, i get mid to high 70's
This is with running -.1 offset in bios BTW and stock clocks on the 3900x

I can also report that my 1950x doesn't get nearly as hot running similar cooling and the same type of workloads so i agree it has alot to do with how the chiplets are connecting to the cooler.
while my temps are higher the actual power at the socket is much lower on the 3900x than my 1950x so that is the only variable that makes sense.

I have come to accept the high temps as the new norm. If AMD thinks its ok to run this hot then who am i to argue, and i can't judge this generations temps by the previous one.

Just remember your engine in your car is a literal fire filled HELL powered dimension. You stick your hand or head over a hot running engine and it feels like the domain of satan.

But to the engine as long as its within OP temp it's as happy as a fat rat eating honey. Its definition of HOT is extremely different than our fleshly tolerance.

So whilst 80c is scorching hot to our flesh and senses, a CPU is yawning and chilling in a cool breeze at the beach especially if that is its load temp.
 
3600 here... ~65C under 12T artificial load with H115i Pro, same as my old 1600 OCd @140W... a Zen 2 CCX chiplet has ~twice the power density as a Zen die, easy math. agreed Zen 2 just runs hot. very efficient though! What sort of power consumption are folks seeing on 3700X / 3900X?
 
3900 puts out around 118 watts according to tests that have measured it. But what really matters when comparing to previous gen cpu's and current gen intel's is that it's all that power put in a space that's 60% the size of what you're comparing to. A slab of copper just doesn't net you the results it does on larger sized cpu's.

I ran tests on mine with thermaltake's 240mm aio and a custom loop i just built (tests done same day and same ambient temp). The custom loop has a flow rate many times faster than the aio, and the custom loop has two radiators (1 a thicker 240mm radiator, 1 a 360mm radiator) and the temp difference betweet the aio and the custom loop is about 1-3C in most tests.

Your best option is to undervolt to stay out of thermal throttle temps until some new better way of moving heat out of the cores are found. 0.1 under volt will net you a 10% drop in temps, assuming your cooler is decent to begin with.

https://openbenchmarking.org/result/1908180-HV-1908170HV38
 
I have a 3900x cooled with an Enermax LiqTech II 360 and when running the AIDA64 stress test for an hour I was right around 83c and all cores at 4ghz.

For me, it's not the load temps that bother me. I can do a Cinebench R20 run and stay down near 70c. That's fine. I expect the fans to ramp up and the heat to rise when doing heavy tasks. What is frustrating me is the idle/low-load temps. My fans shouldn't be kicking up just because I opened Chrome or plugged in a thumb drive or whatever. So if there's any way I can keep this thing cooler when it's not doing much, I'd love to try it.

Definitely gonna give the slight undervolt a try once I'm home.
 
Under load I never see temps like that on the 3900X. Then again, I'm using a custom cooling loop.
 
Under load I never see temps like that on the 3900X. Then again, I'm using a custom cooling loop.

I think more depends on the voltage it needs to run at, perhaps your 3900X requires a bit less voltage to be stable at 4.1 all core.
 
I think more depends on the voltage it needs to run at, perhaps your 3900X requires a bit less voltage to be stable at 4.1 all core.

My test Ryzen 9 3900X goes to 4.3GHz on all cores at 1.35v with manual tuning. Under PB2, it settings between 4.1GHz and 4.15GHz'ish but still doesn't show temps like the OP is talking about.
 
I had alot of heat issues with my 3900x.

Then I remounted it and applied tim across the heat spreader manually not using the dot method and my temps are lower than an 8700k.

I havent touched my voltages. There is no need for it.

Wide open my chip hits 80 to 83 C.

Idle is 30s

Gaming is 40 to 50s

I have a custom loop but an AIO with 3 fans should be decent as well.

Moral of story, if it's hot dont fuck with bios. Remount and reapply TIM. Use a better brand tim instead of the junk they put on the CLC hot plate.
 
Last edited:
I'll give a reapply a try tonight, see what that gets me. Lowering the voltage brought temps down a bit, but brought performance down too.

EDIT: Gave it a go with the two line method. Saw no real difference. Guess I'll just live with my machine being a bit louder than I'd like and hope these idle/low-load temps can be figured out eventually.
 
Last edited:
Back
Top