How hot do CPUs REALLY get? - It's SCIENCE time!

[GoldenLag]

2 minutes ago, Tetanus said:

Fun video!

 

Just out off curiosity, why didn’t you just weigh the water instead of using the measuring cup? Seems like it would be much easier and more precise. 1 liter is one kilogram, so 250 ml would be 250 grams. 350 ml would be 350 grams etc. 

 

I have to admit that using the metric system seems far easier than imperial, but then again I live in Europe so that’s what I am used to. 

like if the video was super precise. 

 

a precise method would measure something along the lines of what me and the other guy is talking about. 

[spartaman64]

1 hour ago, poochyena said:

literally what even was this video.

 

If they wanted to measure heat and tdp, why not just measure the wattage being taken from the wall?

power draw doesnt directly translate into thermal energy

[Tetanus]

@GoldenLag I would be the first admit that I don’t have the technical knowledge to fully understand what you are discussing, but I appreciate you taking the time to share some of your knowledge. 

 

My question was based on the premise that the goal of the experiment was to measure how fast and how high the temperature would go in the liquid. As such I was a little bit confused as to why the amount of liquid seemed to be a little “random”.

 

Anyways, I thank you for the feedback. 

[GoldenLag]

1 minute ago, spartaman64 said:

power draw doesnt directly translate into tdp

it kinda does, but not really. 

 

they attempted to compare heat generated by the CPU that translated into the cooler. 

 

and since you dump your power into the CPU, that would work. 

[GoldenLag]

3 minutes ago, Tetanus said:

My question was based on the premise that the goal of the experiment was to measure how fast and how high the temperature would go in the liquid. As such I was a little bit confused as to why the amount of liquid seemed to be a little “random”.

mengda vatn var rimelig lik ettersom dei berre fylte loopen. 

[Tetanus]

@GoldenLag LoL

Vestlandet FTW

[spartaman64]

15 minutes ago, GoldenLag said:

it kinda does, but not really. 

 

they attempted to compare heat generated by the CPU that translated into the cooler. 

 

and since you dump your power into the CPU, that would work. 

not all of the electrical energy going into the cpu will be turned into heat by the cpu so you would be like a few 10Ws over

[poochyena]

21 minutes ago, spartaman64 said:

power draw doesnt directly translate into thermal energy

What happens to that electricity if its not being turned into heat?

[GoldenLag]

6 minutes ago, spartaman64 said:

not all of the electrical energy going into the cpu will be turned into heat by the cpu so you would be like a few 10Ws over

what else does it get turned into then? i guess you will have a few output signals. tho that still counts as operating the CPU. 

[spartaman64]

6 minutes ago, poochyena said:

What happens to that electricity if its not being turned into heat?

 

Just now, GoldenLag said:

what else does it get turned into then? i guess you will have a few output signals. tho that still counts as operating the CPU. 

some goes to other components and some goes back to the power supply

[GoldenLag]

12 minutes ago, spartaman64 said:

 

some goes to other components and some goes back to the power supply

other components inside the CPU. as i am assuming we are measuring EPS connectors here. 

 

and if there is leftover current going back to the PSU, then you can measure that aswell.

[Computernaut]

I feel like this is more of an Alex project, from the cooling to the numbers, it seems to be more his kind of thing.

[adammmmmm]

I joined the forum just to add to this discussion after watching the Youtube video. I really like this project!

 

A flow meter and two thermal couples - one immediately before and one immediately after the CPU water block would allow you to calculate real time power output of the CPU (or rather power input into the water loop). If all of these variables are logged simultaneously, you would not even have to measure out a specific loop volume or even care so much about what goes on after the CPU block.

 

Sorry for the US units but here is the math behind this: 500 x the flow rate in GPM x the rise in water temperature between the thermocouples in Fahrenheit (BTU = 500 x GPM x ΔT) would give you the heat output in BTU. Log all of those variables over time and you'd have some slick data to play with!

 

This exact setup is how we log HVAC equipment capacity at work!

[Arika]

Quote

IHS solder quality *wink*

 

cheeky

[JoostinOnline]

13 hours ago, GoldenLag said:

it kinda does, but not really. 

 

they attempted to compare heat generated by the CPU that translated into the cooler. 

 

and since you dump your power into the CPU, that would work. 

Doesn't the efficiency of the CPU affect the TDP though? I don't see how it would work with different architectures.

[GoldenLag]

3 minutes ago, JoostinOnline said:

Doesn't the efficiency of the CPU affect the TDP though? I don't see how it would work with different architectures.

They are testing the rating of the TDP/powerdraw not performance per watt

 

So i dont see of efficiency has anything to do with it

[_xJ_]

16 hours ago, GabenJr said:

We have two 8-core CPUs: One at 105W, the other at 95W – But those numbers are completely arbitrary. Let’s test them and see for ourselves which will heat up faster!

 

I was wondering why you guys didn't try to measure the heat flux produced by the CPUs directly?

 

If you manufacture a thick-ish copper slug, insulate it on the sides, and connect it to the CPU on the bottom with the water cooler attached to the top (better make a thicker slug and go chilled on the water to get higher delta-T). read the temps at the top and bottom of the slug (drill holes into the sides, does not have to be connected directly between the slug and the CPU/cooler).

 

Then one can compute the heat flux through the slug in the CPU-Waterblock axis, under assumption that there is no loss to the insulation.

 

https://www.efunda.com/formulae/heat_transfer/conduction/conduction_1d.cfm

 

I believe this would be an easier and more accurate way to calculate what you guys tried to do.

 

Surprisingly they use this method to very accurately determine body temperature for those "best time to get a baby" monitors.

 

Just to clarify - we are basically creating a thermal bottleneck by doing this, and for a homogeneous material the amount of power conducted through depends on the thickness, area, heat conductivity and delta-T between the sides of the material. So the thicker the bottleneck the more delta-T one would need to conduct the same amount of heat across. Or, the more heat there is being generated, the higher delta-T would develop.

 

Back of a napkin calculation:

385 (W/(m*K)) * 35 mm * 35 mm * 10 K / 40 mm  == 118 watts

 

for 35x35mm IHS size

Copper material 385 (W/(m*K))

40mm slug thickness

and 10C delta-T

 

Quite obviously we assume the water block is anchored to 0C via an ice-water bathtub. And you are running the fastest pump you have in store

 

 

[Video Beagle]

something that folks complaining about the methodology might have missed, they wanted something that would be visually appealing to film. Hell, for pure testing, they could have used their thermal camera, but doing water color change is both visual and "jankie"? (is that the word?)...ltt trademarks

[JoostinOnline]

2 hours ago, GoldenLag said:

They are testing the rating of the TDP/powerdraw not performance per watt

 

So i dont see of efficiency has anything to do with it

But doesn't greater efficiency mean less power is wasted as heat?

[GoldenLag]

Just now, JoostinOnline said:

But doesn't greater efficiency mean less power is wasted as heat?

no? it just means less power is used. almost all power put into a CPU is used up as waste heat. unless there is some magic current draw im unaware of that doesnt generate heat. 

[JoostinOnline]

18 minutes ago, GoldenLag said:

no? it just means less power is used. almost all power put into a CPU is used up as waste heat. unless there is some magic current draw im unaware of that doesnt generate heat. 

I've definitely read that before as an explanation of why power draw and thermal output aren't the same.

 

[GoldenLag]

6 minutes ago, JoostinOnline said:

I've definitely read that before as an explanation of why power draw and thermal output aren't the same.

 

Some power does dissappear through signals, but not a lot. PCIe lanecount is fairly similar aswell providing no PCIe 4.0 is used

[greenhorn]

Measuring the heat output is not an easy undertaking if you want to do it with an adequate precision. I would probably have used some shunt resistors and some probes to measure the CPU's power draw (similar to how der8auer measured the X570 chipset's power draw).
In order to compare heat transfer away from the CPUs, I would have locked both in such a way that they have the same power draw, built a setup similar to the one in the video and the looked at water temperature changes with probes (disregarding the color-changing ink). In case someone wanted a more "fancy" approach, one could run both systems side by side, putting the two water reservoirs side by side and using a Peltier element to measure the temperature difference

[Lowbuget]

I wanted to get this off my mind,no matter what your going to never get a clear result IMO,good video but useless to me especially when there is so many variables that could have gone wrong.

[adammmmmm]

1 hour ago, GoldenLag said:

Some power does dissappear through signals, but not a lot

This is incorrect. 100% of electricity that gets put into a computer gets turned into heat.