TDP / Temperatures / Heat Efficiency; AMD vs INTEL; Sanitycheck and Sources


a lot of people are PC enthusiast, but do not really follow hardware.
So they have old opinions or the opinions of the rumor mill lets say.

So i was talking to a friend several weeks ago, and he said “i am not buying AMD, they run way too hot.”. I replied with, “Have you checked that recently? As far as i know here intel and amd are basically the same on that subject. Might even be that intel is the one running hot now”.
Long story short he wants to build a new PC and considers AMD, but asked me for source material for my statement.
My statement was just the general tendency i got from all the hardwareChannels’ videos i watched.

So now i’m asking you if my statement, that when it comes to heat efficiency amd and intel are basically the same, is correct?
And of course if you have sources for that, please post them here :slight_smile:

1 Like

TDP Is Essentially Meaningless

TDP isn’t a good way to measure heat generation or even a good indicator of power useage. It’s a marketing term that serves as a ballpark of what the CPU might draw under moderate loads at stock settings.

To demonstrate how wrong using this metric as a benchmark can be, parts like intel’s 10+ core parts have been measured pulling 650+ Amps during extreme overclocks.

The TIM affects Percieved Temperatures

AMD uses a soldered HS, and indium has much lower thermal resistance than whatever interface material intel is using at present. The thermal resistance between the die and the first conductive material will always be the bottleneck in any cooling solution (to explain this fully you have to go into thermal resistance, R values, and cross sectional area measurements, but suffice to say that bare die > soldered die > pasted die.)

AMD Doesn’t Report Temps Accurately

Also note that AMD runs a 20 degree C offset on their X line of processors, to force fan control software to cool them more agressively. While this may give users higher boost frequencies, it also has the side-effect of falsely reporting temperatures to the user on software that isn’t accounting for it.

Current Gen AMD is Forced To Run Cooler

That said, AMD does lose stability at high clocks at a much lower temperature than intel parts. Intel can usually sustain moderate overclocks at up to 85 C before stability is completely lost, whereas some AMD parts can lose stability for the same % of frequency bump as low as 65-70 C. This means that current-gen AMD parts can’t actually afford to run hotter, because they a) scale back frequency more aggressively to compensate, and b) can’t maintain boost or OCs stable at even “moderate” intel OC temperatures.

Right now the chief factor in purchasing a new platform is more about your connectivity and needs for certain features.

Select a Product That Fits Your Needs

For example, let’s say your friend needs a workstation:
Buying a used HEDT intel chip or a xeon and putting it in an X99 board (older premium chipset that isn’t that hard to find at a discount or used nowadays) may be more attractive than ryzen + or coffee lake because you get much more connectivity for similar money (more pcie lanes, generally more generous features on the motherboard)

Or maybe RAM prices are sky high and you want a current-gen platform that won’t be picky or lose much performance if you stick value DIMMS in it. In which case, Zx70 and a coffee or kaby lake (at a discount by now) part is probably the way to go, especially if he needs his machine for gaming to the exclusion of other uses.

There’s No Such Thing as ‘Heat Efficiency’

For layman purposes, essentially all of the energy put into silicon ends up as heat. The concept you’re likely looking for is thermal resistance, a function of the combined conductive ability of the TIM, IHS, Paste, and HSF to move thermal energy away from the silicon. This value doesn’t scale 1;1 with the size of any single component, and the weakest link will always cap your wattage on cooling capacity, especially if it is early in the stack.


At that point your previous sentence about stock settings doesn’t apply anymore though :wink:

Eeehhhh, probably tooth paste or shoe polish :stuck_out_tongue_winking_eye:

That might have been a good plan, however most fan controls factor this in as far as I’m aware, might be wrong though.

That being said I agree with you on your points. The regular user might even say Intel runs hotter at the moment, but that is mostly because they can afford to as stated above.

To get anything semi accurate you’d need to measure how much heat is actually being dissipated, and even then it’s questionable how useful that data is, because at the end of the day, what does it matter as long as it’s not thermal throttling?

There are plenty of CPUs that draw way more/less at stock than their TDP, too. That’s just the most extreme example of being outside the “thermal design”

They didn’t originally, which is why the offset was put in place, because AMD is very used to working around intel thermal stability points. Back in the day this was a problem with Phenom II, which had a much lower temp throttle/stability point than Equivalent intel parts. Now that most common softwares account for it, it’s less of a problem, but most raw bioses and simple reporting sensor readouts will still put out +20 C

I was more pointing out that the 2 companies don’t necessarily report accurate temps (particularly AMD)

Measuring how much power is being consumed across the vrm traces or the mobo 12v lines using an amp clamp would yield much more accurate power figures, as silicon is very efficient at turning power into heat.

Thank you for that in depth answer!
I think i get your point tkoham.
But what bugs me a bit is that the articles cited are older. The Puget-System article is from 4 years ago and with the old AMD chips.

I never experienced the “AMD is running hot” thing. I guess what that meant was, that under load it needed a really good cooler to make it not thermal throttle. Is that correct?

So given my assumption is correct, i guess my question is, does AMD still need a beefy cooler to not throttle when under heavy load.
No overclocking.

at stock, the included heatsink as actually much closer to adequate than what intel gives you out of the box, and if you aren’t overclocking, there’s no need to get a different cooling solution

Also, basic physics doesn’t change year to year, and the puget system article is just about throttle temps. as far as I know there’s no official design doc that states the OC stability temps for either manufacturer, just the thermal throttling

thank you :slight_smile:
I will look out for the design docs.