How do you determine appropriate thermal expectations?

I have 2 questions related to CPU thermals:

  1. How do you determine appropriate base thermal expectations for a new build (in this case 3950x see below)?
  2. What are (or how do you find) thermal limits beyond which you are risking shortened lifespan?

My Use Case:

  • General productivity power user (Office, a dozen chrome tabs, trading software, video recording/rendering, etc.)

My Rig:

  • Aorus Elite x570 m/bd (supposedly pretty good VRM)
  • 3950x
  • 32Gb (2x16) G.Skill Ripjaws V Series 360016-91-19-39 (XMP enabled)
  • 1 Tb Sabrent Rocket PCIE 4.0 NVME m.2
  • 1 Tb Intel 660p PCIE 3.0 NVME m.2
  • 3 Tb Seagate Constellation SATA
  • PowerColor rx590 8Gb Red Dragon (not a gamer)
  • Cooling:
    • Thermal Grizzly Conductonaut compound
    • NZXT X72 Kraken 360mm AIO
    • Intake = 3 x 120mm be quiet Silent Wings 3 (side mount)
    • Intake = 3 x 120mm be quiet Silent Wings 3 (bottom mount)
    • Exhaust = 3 x 120mm NZXT through the radiator ( top mount)
    • Exhaust = 1 x 80mm PCcooler F92 (top rear grill mount)
      So I’ve got great airflow and I thought I should have plenty of liquid cooling on he CPU.

Findings at this point:
I use Blender Benchmark (all scenes) 2.81a for he CPU testing/benchmarking and CPUID HWMonitor for Temps and load checking.

  • Out of box (3.5Ghz base 4.7Ghz boost never ever close to it):

    • 40C idle
    • mid-70sC under load
  • OC to 4.2Ghz all-core (1.4v VCcore)

    • 43-45C idle
    • 83C max under 100% load 30min
  • OC to 4.3Ghz all-core (1.43v VCore)

    • 45C idle
    • 86C max under 100% load 30min

So, anyone have ideas how I should set my expectations on thermals and how far I can/should push it?

I believe Tjmax is 95C on Ryzen processors, you don’t want to get above that. The CPU should also automatically start throttling itself at that temp so it doesn’t get damaged.

A very general rule of thumb is that every 10C increase in temperature halves the life of a component, and vice versa every 10C decrease in temperature doubles the life of a component.

Yea, I like to find cooling devices that are capable of cooling at at least 33% higher spec than the maximum operating temperature of the CPU.

Ideally, 50% more… and put these things in a basement or someplace that’s always a little cooler than other places in the house. You sorta get “something for nothing” doing that.

Thanks. That makes sense.

While I’d very likely avoid Tjmax and throttling, I didn’t have any idea if 70c or 90c was typical. (My old rig was a 6-core.12-thread Intel that ran in the high 60sC even when overclocked.)

I guess since I tend to get sucked into tech porn every 2-3 years and build a new one cuz I gotta have it, it won’t matter too much. I was just a little concerned to be in the mid-80s with all the fans and AIO in place.

Yeah, temperature kind of depends on a lot of stuff.

Different architecture will have different normal temperature. My Skylake Xeon at full load barely gets over 60C, but my 1700X hits 60C at the drop of a hat. An old 3770K I had would hit 80C with only a minor overclock. That’s using the same cooler across all three CPUs.

Temperature even depends on ambient room temperature. For example, my PC has an idle temp nearly 10C hotter in Summer compared to Winter simply because of the temperature in the room. Likewise my load temperature also increases.

I’d also suggest playing with your fan settings when you have time, even shutting some of them off if you can to see what happens. Gamers Nexus has found that in some scenarios fans can make temperatures worse compared to just leaving the fan slot empty. Just don’t shut your radiator fans or pump off.

I think your higher than expected temperature issues are related to the voltage i can easily get 4.3ghz on 1.23750 volts. Try lowering your voltage incrementally to see what you chip is capable of doing. The voltage these x570 boards give out of the box can be extremely excessive.

I’m absolutely willing to learn/try. I started out using the Tweaktown Gigabyte x570 OC Guide which gave this advice (which is why I didn’t even try below 1.3):

CPU VCore: The main voltage to change is the VCore, in fact when we overclocked it’s the only voltage we had to increase to achieve our CPU and memory overclock. It is mainly just for the CPU, and you can move your way up from 1.3v to 1.4-1.45v depending on your cooling capability.

Read more:

However, quite frankly there are so many sub-settings and different voltages that if I were more expert (complete and admitted OC newb) I could probably wring more out of it.

1 Like

Guys, thanks. I think I’ve got a better handle on this now. I’ve managed to tune it to where I get max 77c after a 30min at 16x 100% test on 4.35ghz on 1.35v Vcore at this point.

So, I’m more comfortable now. I suppose I might be able to wring a tiny amount more upclock (4.375ghz ?) and/or a tiny bit more down-volt and may play more as time permits. However, at 45-50c normal temp in daily driving I’m a happy camper now.

Again thanks

1 Like

1.35V vcore All Core OC on Zen 2 is high and might lead to degradation.

As for your temps they seem very in line with Zen 2.

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.