[SOLVED] Rather high spiking and load and over all 5950x temperatures with dark rock pro 4

robthatguyx

Distinguished
Dec 20, 2011
1,155
0
19,310
Hey guys all intel guy finally switched over to amd and have some concerns about my cpu.
Dark rock pro 4 ,kryonaut paste, crosshair viii dark hero . Voltage under load shows to be 1.5v solid across most of my cores
Im seeing between 43 and 60c at idle/web browsing and spikes up to 70c . Under prime 95 for 10 minutes I went up to 85c then started to slowly thermal throttle back to 3.5-7ghz per core.Gaming mostly stays around 84c unless I shut off my case fans(Corsair7000x). With those off I hit 92c and alt f4'd the game to go put them back to balanced mode. What concerns me is how crazy fast this thing is spiking in temps it seems like within seconds I go from mid 40s to upper 70s sometimes just moving windows around. I tried disabling pbo and that didn't see to make a difference. Are these just pretty hot chips? I lose on the lottery or what? maybe because im somehow getting near 5ghz on most of my cores without touching my overclocking settings? :monkas: I have flashed the bios to the latest and that also made no difference.
Here is a cap of my hw monitor. All I have done is open this and type up this post. View: https://imgur.com/a/VpUQGyd

Any input is extremely appreciated
Edit, with pbo off and auto overclock off my cpu sat locked at 3.4ghz and never went above 70c doing anything.. so do I leave it running on border of spicy meatball problems or just disable all the auto overclock stuff? The cpu never went over 1v when it was all off
 
Solution
....when I was overclocking 2500ks back in the day 80c was like no go territory so its very concerning for me to see almost 90c peaks

Gotta forget what you know about Intel CPU's. Ryzen's completely different and not only because it uses 'hot spot' sensors. It all gets down to the 7nm geometry that leaves too small of a surface area to transfer out the heat, so they had to design the chip to be more heat tolerant too.

It also uses a boost algorithm that's very aggressive with light loads. But when the loads get heavy and it heats up it will pull back clocks and voltage to safe levels for the higher temps...if you leave it in AUTO. Better cooling means it will pull clocks back less.

Makes manual all-core overclocking very...
...
Any input is extremely appreciated
...
Ryzen reports a hot-spot temp from the hottest of several dozen sensors scattered across the die. Ryzen also boosts agressively to max clocks on light bursty loads. That results in a temperature spike on one of those sensors. Each spike isn't all that thermally significant until they start adding up with with heavier processing loads. It's exactly like lighting a candle: the room won't get warm until you light several dozen and keep relighting them as one burns out.

So the first thing is to get HWInfo64, not HWMonitor, for a proper monitoring utility. It works much better with AM4 motherboards and Ryzen CPU's.

In HWInfo look for a CPU Die (average) sensor. That's an average of the temps across the CPU die. Watch how it changes with light loads and heavier processing loads. There are several columns, watch the average column in particular as the average of the averages is far more important to get an idea of the actual thermal state of the CPU core over time and not get confused by the temp spikes.

In the HWInfo Sensor Configuration screens you can set both the polling period (how many mS between polls) and the averaging time...how long of a time period it accumulates sensor readings to calculate the average column. Default is OK, but reducing polling period to 500mS is helpful because Ryzen's algorithm works so fast.

 
Last edited:

robthatguyx

Distinguished
Dec 20, 2011
1,155
0
19,310
Ryzen reports a hot-spot temp from the hottest of several dozen sensors scattered across the die. Ryzen also boosts agressively to max clocks on light bursty loads. That results in a temperature spike on one of those sensors. Each spike isn't all that thermally significant until they start adding up with with heavier processing loads. It's exactly like lighting a candle: the room won't get warm until you light several dozen and keep relighting them as one burns out.

So the first thing is to get HWInfo64, not HWMonitor, for a proper monitoring utility. It works much better with AM4 motherboards and Ryzen CPU's.

In HWInfo look for a CPU Die (average) sensor. That's an average of the temps across the CPU die. Watch how it changes with light loads and heavier processing loads. There are several columns, watch the average column in particular as the average of the averages is far more important to get an idea of the actual thermal state of the CPU core over time and not get confused by the temp spikes.

In the HWInfo Sensor Configuration screens you can set both the polling period (how many mS between polls) and the averaging time...how long of a time period it accumulates sensor readings to calculate the average column. Default is OK, but reducing polling period to 500mS is helpful because Ryzen's algorithm works so fast.



Hey so I re did the polling like that guide suggests, I did not touch the averaging time.

Here is what I came up with from one game of enlisted beta
View: https://imgur.com/a/VyqlVFa


Seems like most of the time im in the mid60s low 70s with spikes of temperature that last like a 2 seconds then vanish again. when I was overclocking 2500ks back in the day 80c was like no go territory so its very concerning for me to see almost 90c peaks
 
....when I was overclocking 2500ks back in the day 80c was like no go territory so its very concerning for me to see almost 90c peaks

Gotta forget what you know about Intel CPU's. Ryzen's completely different and not only because it uses 'hot spot' sensors. It all gets down to the 7nm geometry that leaves too small of a surface area to transfer out the heat, so they had to design the chip to be more heat tolerant too.

It also uses a boost algorithm that's very aggressive with light loads. But when the loads get heavy and it heats up it will pull back clocks and voltage to safe levels for the higher temps...if you leave it in AUTO. Better cooling means it will pull clocks back less.

Makes manual all-core overclocking very unrewarding most times as just putting on better cooling gets as good of a performance return since it has to pull back less on the clocks as it heats up.
 
Last edited:
  • Like
Reactions: Phaaze88
Solution