You’re right about Core Temp. I switched to HWinfo 64 and coretemp seems to be reading the cores highest temp verses the average temps over the cores. I’ve been using Ryzen Master and HW for reading lately. Definitely a difference. Mount for cooler is offset (still strange to me but it did help a little) and I did look into undervolting PBO2. Seems with a 5950 you cant go down that much especially compared to the other 5000 cpus so it doesnt seem worth it imo. I’m thinking of trying some different fans for the radiator that have better pressure.
went through this with my 6900xt, simple change made significant difference…
set up your CPU rad as an intake rather than exhaust. i.e., relocate your front case fans to the top of the case, put the rad in the front.
your 3080 will be heat soaking the shit out of your exhaust-radiator (via its air cooler dumping a shit tonne of heat into your case) if it is exhaust.
think about it:
your 3080 is dumping ~300-400w of heat into your case. your cpu is 100w or so worth of heat. have the cpu cooling as intake, don’t run the GPU-heated air through your rad as an exhaust.
cool the CPU first with fresh air into the rad, and then let the GPU dump 4/5 of the total system heat into the case on the way out, AFTER the CPU has been cooled.
If your radiator is exhaust, it will be taking the air inside your case, after the GPU has dumped 300-400 watts of heat into it, to cool the CPU….
IF you feed your radiator hot air, it gets hot. right now the bulk or your heat in your CPU cooling system will be absorbed from your GPU.
yes, strictly speaking this will be heating your GPU with your CPU. but unless you’re over clocking the living shit out of the CPU, the heat increase will be negligible compared to what the GPU can output in terms of heat.
While in theory this isn’t a bad idea, especially in a custom loop, in practice it’s a bit worse than you might think.
I have the 360 version of this AIO in a Phanteks P500A. When mounted in the front, both CPU and GPU at idle, the GPU was 48-50C. Moving it to the top dropped the GPU back down to 32-34C. That is NOT negligible. Not to mention that I’m only using a 3700X. Also, since OPs case is much more compact than mine, that could make it even worse.
What about load temps with the AIO in top exhaust, you ask? 10 minutes of Cinebench multi (no PBO) and Furmark 1080: GPU 78C, CPU 61C. Room temp is roughly 22C.
You’re welcome. Also, for easier cleaning maintenance, you may want to make the radiator fans pull instead of push. There won’t be any cooling performance difference.
Temps seem normal tbh. I wouldn’t worry about it too much.
What happens when you are gaming is the cpu has 16 cores of power budget that it can in some scenarios spend on only a few cores. because its putting 16cores worth of power into a smaller die area the temps will be higher even though its the same power limit and it will keep boosting till it hits its 90c temp limit or the current limit.
There also seems to be some variance in the die to IHS interface quality with these CPUs (or maybe its just silicon lottery) some people have examples that wont go over 80 and some behave more like yours and some even worse. For example there are 5800x out there that are able run up to the 90c limit under pretty much any cooling solution you throw at them.
For the typical (as listed above) AIO with a typical CPU - you will be getting a LOT less heat from the CPU than you will from the GPU, especially high end RTX3000 series.
The GPU is going to dump 300-400w of heat into the case.
You can either run that heated air through your CPU radiator (when set up as exhaust), or you can run the CPU heated air (nominal 100 watts or so give or take - which with a decent AIO will make it slightly above ambient temp) into the case to then hit the GPU (along with other air intakes to dilute it) over the GPU cooler which (if its a decent cooler) is capable of cooling a GPU down with barely turning the fan in regular ambient air.
Your choice.
I’ve run both setups (with a 6900 in place of the 3080ti, so a GPU that’s dumping a bit less heat) and no, my GPU is not throttling, but my CPU is sustaining much higher clocks with ZERO fan noise from my system, vs significant noise from my AIO that was being pre-heated by my GPU and spinning the rad fans like crazy trying to exhaust heat from the CPU cooler that the CPU never put into it (was absorbed from the ambient air inside the case, heated by the GPU).
For me, switching to the suggested setup (rad as intake in front, front case fans relocated as exhaust in roof) was worth 5-10% improvement on Firestrike ultra - with much less noise. I also pass stability tests in 3dmark now, which couldn’t do before with the setup of rad as exhaust.
My 5950x came in today, so for OP, I’ll show numbers some time during the weekend, though we do have a case and size of AIO for a variable.
In the meantime, and it’s a pain in the ass, but turn off PBO unless doing any productivity kind of work, as it means little to nothing in gaming performance.
FanControl:
JayzTwoCents did a video about it. Can’t link (forum rules) but it was a week or so ago. The way that he instructs how to set it up isn’t necessarily correct for temp control, but still very informative. See my image for example. I don’t care about noise. It’s white noise to me, so whatever, but if you want to control fan speed by highest temp, you need to make a Custom Mix Sensor.
Given all variables, as stated previously by others, your temps are fine. As suggested, change radiator fans, set up a custom fan curve, turn off PBO unless doing productivity work and all should be good.
Thank you so much for going out of your way to do all of this! This helps a lot. As long as I dont suspect my fans are going to explode or launch my system into space, the noise doesnt bug me much either, haha. The data doesnt lie so this is the way.
I did forget to mention that the only settings in 3DMark that were “custom” was that I ran them in windowed mode with the demo turned off. Everything else was default.
6900xt on air
5900x cooled by h115i in front as intake
PBO on
No custom fan curve (all bios fans/voltages at default)
Define r6 - former front fans from where the h115i are moved to top as exhaust
All filters installed
GPU clocks above 2400
Cpu clocks boost up to 4850
Cpu temp under 60c during firestrike extreme (not water temp, temp from benchmark run)
I was getting heavy cpu throttling with the rad as exhaust (and much more fan noise from the radiator). GPU has not changed and benchmarks improved with rad as intake. By a lot. The GPU was clearly heat soaking the cpu rad with the heat it was dumping into the case.
Sure I’m dumping cpu heat into the case for the GPU to cool with but it doesn’t care. There isn’t enough cpu heat to matter.
Interesting. Though doesn’t Firestrike stress the GPU only? I find that the benchmarks don’t really represent the temps that you’ll see in gaming. If you have it, try running a game like Killing Floor 2 at 4k. That will tell you how good your thermals are. Took some tweaking but I managed to get my GPU temps to mid to high 70s and cpu to 60s with all settings maxed at 4k.
The GPU was heat soaking the air inside the case, which was then exiting via the CPU radiator and heat soaking it (and thus the CPU) on the way out.
People forget - radiators work both ways. You feed them hot air they take the heat into the water from the ambient air. They only cool things if the air hitting them is cooler than the liquid inside them. If the air is hotter, you just built yourself a water heater.
Also:
CPU Rad as exhaust - could not pass time spy stability test
CPU Rad as intake - reliably passes time spy stability test
System is also noticeably quieter, as only my GPU fans ramp a little bit now, rather than the radiator fans ramping like crazy due to heat soak from the CPU (Which just feeds it more hot air form the GPU).
You can say the same thing about a heatsink. Its limited by the temperature of the air its using in the same way.
At the end of the day its a trade off between lower gpu temps or lower cpu temps depending on what is getting the coolest air. Its going to depend on the individual setup as to what works out best from a thermals/noise level perspective.
Yes, definitely. And this is probably very much system dependent - what CPU and GPU you have and what the cooling capacity is of their attached coolers.
But its not always an EQUAL trade-off…
The difference is what you want to heat with what, and the difference in the ability to deal with the heat.
For me, with a 280mm CPU radiator (for example) shedding 100-140w of heat from the CPU into the case, I’m doing less to heat the GPU than that 280mm rad absorbing 300+ watts of heat from the GPU and ADDING to the 100-140 watts of heat the CPU is adding to that - prior to it all exiting the case.
My radiator does a lot better dealing with 100-140 watts of CPU heat than it does dealing with 300 watts of GPU heat plus the 100-140 watts of CPU.
Those figures aren’t exact because there is another exhaust fan, but the principle is the same.
Be aware of what is getting heated by what, and what capacity for cooling said components have. In my case the GPU cooler is well and truly capable of shedding 300-400 watts plus (its a 3 slot card). So the small amount of additional heat from the CPU doesn’t give it trouble. Most of the time the air coming off the CPU rad even under load is close to ambient.
The reverse is most certainly not true.
If i had say, an overclocked, higher Alder lake CPU and a low-mid-range 150-200 watt GPU things might be different.
Just my 2 cents… Your problem may not be anything physical… What I’ve found on my platforms is that Microsoft will run Winsat (windows user experience app) in the background at random times, but more when it’s late at night when a PC is typically unattended… Winsat will run a full benchmark suite on your CPU/GPU. I’ve had one of my platforms (a 3990x with an IceGiant Thermosiphon and a total of 23 fans in the case) all of a sudden sound like a passenger jet starting up. I’d go “WHAT THE HELL?” and start digging. Winsat. I disabled it. It’s in /Windows/System32
