[SOLVED] Why is a stock i7 9700k reaching 180 watts during prime

Mezoxin

Reputable
Nov 3, 2019
493
97
4,840
When i run prime 95 everything is running normal for up to approximately 5mins as seen here , with system total consumption at 176w as seen in the corsair link and cpu at 114 as seen on hwmonitor
0S863S5.png


all of the sudden the wattage starts ramping up to 252w system and 184 cpu and ofcourse temperature is at throttling values , i dont understand this sudden ramp in power although the cpu load is the same at 100 %
7W4dVwS.png


When i checked reviews all of them had the maximum wattage for the whole system was at around180w , sp why is my processor having this behaviour ? Is this normal for a stock i7 9700k ? btw gaming load temp for the cpu never exceeded 70c
 

Mezoxin

Reputable
Nov 3, 2019
493
97
4,840
Maybe Prime95 is running an AVX load at that moment. You can adjust AVX limits in BIOS. They are very stressful workloads and really heat up the CPU. You should also be able to exclude AVX tests in Prime95, too, or just use Prime95 version 26.6

disabled avx
PiIfloh.png


lower wattage but same thermals , but is it normal for this cpu to reach that kind of wattage under any circumstances ?
 
One thing you have to understand with current Intel chips is that their TDPs are more or less outright lies if you run your system on default setting with TurboBoost enabled. The 9700k may be listed as a 95 Watt processor, but it will only achieve that if you turn off TurboBoost and only run at the 3.6GHz base clock. With TurboBoost enabled, the CPU runs at 4.6GHz on all cores by default, and consumes a lot more power and generates a lot more heat.

Prime95 (especially newer versions with AVX support) is basically a worst case scenario where your CPU power draw will be maxed out and you are likely to see uncomfortably high thermals regardless of cooling. Gaming stresses the CPU a lot less, so you'll see much lower temps doing that. Prime with AVX isn't a realistic workload on the CPU and is more for stress testing if you want to try to verify stability on an overclock when running AVX workloads.

If you ever plan to overclock your 9700k or plan on running lots of AVX workloads on it with stock Turbo behaviour, I would recommend a stronger CPU cooler than the Hyper 212+. As you've seen, higher end Intel chips can generate a lot of heat, and the Hyper 212 is at best a midrange air cooler. A high end air cooler or 240+mm water cooler would dissipate more heat and keep temps lower if you're going to run your CPU at that kind of load.
 

Mezoxin

Reputable
Nov 3, 2019
493
97
4,840
I knew that TDP isnt applicable at turbo boots but never have i imagined that its to the extent of reaching double its TDP at certain load scenarious, thats ridicoulus
I use the pc for games mainly , no avx loads as far as i know
I had that hyper 212 since my old Q9550, it due for an upgrade , i think i will be sticking with an aircooler they are much more reliable
 
If you are measuring from the wall,multiply that by .9 for 90% eff. Thats what your system is consuming.

Tdp is also not a power load rating but heat load at base clock all cores loaded (with intel). For example a 7w led bulb had a low power rating and very low tdp of 1 watt. A 7W night light bulb has a low power rating but a high tdp (like 6 watts heat)
 

Mezoxin

Reputable
Nov 3, 2019
493
97
4,840
If you are measuring from the wall,multiply that by .9 for 90% eff. Thats what your system is consuming.

Tdp is also not a power load rating but heat load at base clock all cores loaded (with intel). For example a 7w led bulb had a low power rating and very low tdp of 1 watt. A 7W night light bulb has a low power rating but a high tdp (like 6 watts heat)

for recent intel CPU's they match TDP values with their PL1 values , So basicly TDP is the same as load power consumption without turbo boost
 

TJ Hooker

Titan
Ambassador
Ill respectfully disagree with you both. Dont get me wrong. Silicon is silicon between amd and intel. But not 100% of the power going through a cpu is lost as heat. It is a very high percentage. Silicon is one of the most effecient thermal heaters.
Maybe not precisely 100%, but close enough as to make no difference. Where do you think the rest of the power is going if not to heat?
 
Maybe not precisely 100%, but close enough as to make no difference. Where do you think the rest of the power is going if not to heat?

You have to remember all power that goes into a processor may not necessarily be consumed inside the processor. For example, your voltage signals which drive memory chips and the PCIe bus get sent outside the processor. Also magnetic field lines can cause a change in potential of electrons. Then there's the inherent efficiency of the power delivery phases which sit outside the chip itself. If I crank load line calibration to high, the efficiency goes to hell as it starts to clock up. That's not the chips fault, but the power phase trying to compensate for voltage sag.
 

TJ Hooker

Titan
Ambassador
Ah, I see what you're saying. I was only talking about the power being consumed in the CPU itself, so after VRM efficiency and other losses have been taken into account.

I'm not sure exactly where the values being presented in hwmonitor are being measured, but you're right that the heat being dissipated solely by the CPU is likely going to be lower than what is being reported there.
 

Mezoxin

Reputable
Nov 3, 2019
493
97
4,840
Ill respectfully disagree with you both. Dont get me wrong. Silicon is silicon between amd and intel. But not 100% of the power going through a cpu is lost as heat. It is a very high percentage. Silicon is one of the most effecient thermal heaters.

Ofcourse its not 100% , but intel has decided to match both PL1 and TDP values as differences where insignificant and probably to add a safety factor
Rpgtzsz.png

At the end for the user, what matters really is to know that if his cpu consumes X power for his specefic use case scenario , he should get at least the same TDP rating when considering the cooler