ncase m1, 5900x, 3080 FE thermals.

raz-0

Minimal Tinkerer
Original poster
New User
Oct 11, 2020
3
0
So before upgrading my CPU I went searching to see if the 5900x was viable with the C14S. I did not get a lot of useful results. Actually I found none. Just people asking the same question. I also didn't get a lot of useful results on temps with a 240AIO either, but most seemed to indicate they were not holding it below 80C. Doing what? Not really specified. So to help the next guy I figured I'd post some numbers. (cross posted at reddit as well since reddit and here were the top source of the very little info I could find.)


System Specs
------------------------------------------------------------
ncase m1 2x scythe 120x15 intake on bottom, 2x scythe 120x15 exhaust on side. NH-C14S with an noctuea chromax 120x25 hooked up to the bottom pushing out to the side and the scythe above it pulling to exhaust out the side. 3 of the scythes are hooked up to throttle up and down with chipset temp and the one in push-pull on the C14S is set up to throttle with CPU temps. CPU and case fans are all set to minum 1000RPM as it is pretty quiet.
32GB CL-18 3200.
silicon power 2TB NVME gen 4.
ADATA SX8200PNP 2TB (seems to be pre controller downgrade).
ASUS B550-i
3080FE.
kryonaut paste.
23C ambient.
No real OC or undervolting of CPU or GPU
1440p display.

OLD 3700x 7 months post install (All numbers as Tdie/Tctl as reported by HWINFO64. )
------------------------------------------------------------
Idle - 48C
Gaming would fall somewhere between 72-78 depending on how taxing it was on CPU with a typical max all core clock of 4050mhz
Linpack 16 threads 4096 - 86.1C Max and you would see it bounce off 4050mhz.
Cinebench R23 on a loop hit a max of 82.8C but mostly sat between 81 and 82.
I have my fan curves set to wait 12 seconds prior to ramping down to avoid RPM hunting which is annoying. from the 86C peak, it'd hit 55C before ramping down and slowly returning to idle temps.
(side note, when new, prime95 with avx would hit 92C+, and I don't do it anymore as it seems rather abusive and non-representative of typical heavy loads)
GPU idles at 27-30C unless doing something that actual does GPU stuff. In gaming it typically doesn't go over
Max clock I would ever see was 4150mhz.

NEW 5900x (All numbers as Tdie/Tctl as reported by HWINFO64. )
------------------------------------------------------------
Idle 48C
Linpack - 90.4C max. Initially it would bounce between 4321 and 4419mhz all cores, but after a while it would bounce between 4296mhz and 4321mhz.
Cinebench R23 - 90.3C max. But it takes almost a whole 10 minute loop to hit it. It spends a whole lot of time at 88.xC.
At the 12 second fan ramp down time, it reaches 55.7
GPU idles as 27-30C and in fact behaves exactly the same as with the 3700x setup.
Max clock I would ever see is 4.8Ghz just like it is advertised.
I played some random games for about an hour and my max CPU temp was 70.2C with an average temp of 68.xC. I was a bit shocked this was lower than the 3700x. Figuring something might be up, I ran a 3dmark timespy run. It scored 16,631, hit 80.25C in the cpu stress test portion, and hit a brief peak of 74.5C in the two GPU tests. Mostly it was between 68-69C.
ADDITIONAL INFO
------------------------------------------------------------
Both CPUs used kryonaut and it was applied using the frosting method. However, for the 3700x I bought the 1 gram tube from amazon and for the 5900x I bought the 5.5 gram tube from a thermal grizzly authorized seller. Color was identical, but they did not spread remotely the same. the amazon tube spread like tacky cake frosting. The 5 gram sample was WAY stickier and was a huge PITA to apply. It liked sticking to plastic way more than metal, and would rope up an glob real easily. Upon removal of the 3700x, the paste was not dry and crusty, but had dried out some. Either formulation has changed or one of them was not genuine, because they behaved way differently.
The 5900x is spread over two chiplets, and just seems to make better use of the heat spreader and/or heatsink.
 

evadne

Trash Compacter
Jan 15, 2020
45
47
Hi

Regarding the 5900x

Tdie/Tctl is not the real temperature and can be quite a bit higher than actual. It is a synthethised value that AMD uses to ensure consistent cooling. Best to watch Tdie for CCD1 / CCD2 to get the real values.

Also. The chip will boost right to the thermal limit. If you want it to run cooler you can enable PBO and set a lower temperature. Otherwise as long as there is thermal and electrical headroom your CPU will boost until it hits one of the limits. This behaviour is specific to Ryzen and can be tricky to reason about. You can also consider using Curve Optimiser and setting a slight undervolt which should also help your temperatures.

Thanks
 

raz-0

Minimal Tinkerer
Original poster
New User
Oct 11, 2020
3
0
I'm not too concerned. Outside of synthetic things it seems to be staying well under the 90-95C AMD says is normal. More so if you look at CCD1/CCD2.

The above combo seems to be holding everything in spec, including the memory hotspot on the 3080FE which hits 90C, but I haven't seen it go over.