Is Power = Power?

[YoungBlade]

I sometimes see people talking about core count and CPU generation when discussing whether motherboard VRMs can handle a certain CPU. I will typically respond to this by saying that "power is power" and that core count and generation are irrelevant when it comes to whether a given VRM is sufficient for a specific CPU. But am I actually right about this?

 

Is there anything about core count or architecture that would make a motherboard be able to handle a 4 core R5 1500X overclocked and drawing 90W, but that would mean it wouldn't handle an 8 core R7 5700X also drawing 90W at stock?

[Hinjima]

5 minutes ago, YoungBlade said:

I sometimes see people talking about core count and CPU generation when discussing whether motherboard VRMs can handle a certain CPU. I will typically respond to this by saying that "power is power" and that core count and generation are irrelevant when it comes to whether a given VRM is sufficient for a specific CPU. But am I actually right about this?

 

Is there anything about core count or architecture that would make a motherboard be able to handle a 4 core R5 1500X overclocked and drawing 90W, but that would mean it wouldn't handle an 8 core R7 5700X also drawing 90W at stock?

I think you are right, power is power. 90w is 90w.  I might be mistaken of course.

[manikyath]

the problem here starts with the TDP ratings ffrom AMD and Intel being.. flawed.

 

but if one CPU is pulling 90W from the VRM's, and another is pulling 90W from the VRM's.. assuming it's at about the same voltage, it is identical. there's some slight variations in VRM efficiency at differing voltages.. but honestly, if you're that close in the margins that this difference matters, you're flying too close to the sun already.

[tkitch]

Yes and no:

 

Yes 90W at 12V is 90W at 12V.  That's absolutely irrefutable.

But what a CPU is rated at, especially today with self-boosting and similar tech?  Wattage has little to do with what something's spec'd to be.

 

Intel sure as hell doesn't list the spec for the 13900K to be nearly 300W, and yet it does draw nearly 300W when turboing.  (Or more, because who cares about power!)  Similarly a 7950X is listed at 170W, and will easily clear 225W at stock settings with self-boosting.  

 

So you have to compare actual draw for these things, not some listed specs.

[podkall]

7 minutes ago, tkitch said:

But what a CPU is rated at, especially today with self-boosting and similar tech?  Wattage has little to do with what something's spec'd to be.

ah yes, AMD's 65W TDP on AM5 CPUs lmao

 

I think they mistook 65W TDP with 65W when it's just sitting in motherboard with powered off PC

[tkitch]

3 minutes ago, podkall said:

ah yes, AMD's 65W TDP on AM5 CPUs lmao

 

I think they mistook 65W TDP with 65W when it's just sitting in motherboard with powered off PC

you say that like intel doesn't do the exact same thing.

[YoungBlade]

11 minutes ago, tkitch said:

Yes and no:

 

Yes 90W at 12V is 90W at 12V.  That's absolutely irrefutable.

But what a CPU is rated at, especially today with self-boosting and similar tech?  Wattage has little to do with what something's spec'd to be.

 

Intel sure as hell doesn't list the spec for the 13900K to be nearly 300W, and yet it does draw nearly 300W when turboing.  (Or more, because who cares about power!)  Similarly a 7950X is listed at 170W, and will easily clear 225W at stock settings with self-boosting.  

 

So you have to compare actual draw for these things, not some listed specs.

I am fully aware that the TDP numbers are garbage, which is why I said that the 5700X draws 90W, because its TDP might be 65W, but its PPT is 88W, which is what actually determines the power limit.

 

So yes, a 1500X OC'd that runs at 90W is not going to be even close to what a 5900X with a 105W TDP will draw, as its max is actually 142W. But I'm assuming that the CPUs are actually drawing the same amount - not that they have the same "TDP" or "PL2" or whatever flavor-of-the-week name Intel and AMD want to give to their power draw figures.

[podkall]

1 minute ago, tkitch said:

you say that like intel doesn't do the exact same thing.

I mean they probably do, but it's like, the CPU sucks so much power by the design to push as much power and handle constant 95°C, and then tell you it's 65W TDP

 

12 cores with hyperthreading, 5,4Ghz boost, has iGPU, apparentaly is 65W TDP:

 

[tkitch]

1 minute ago, podkall said:

I mean they probably do, but it's like, the CPU sucks so much power by the design to push as much power and handle constant 95°C, and then tell you it's 65W TDP

 

12 cores with hyperthreading, 5,4Ghz boost, has iGPU, apparentaly is 65W TDP:

 

So are you just intel fanboy-ing or are you trying to make a point?

Because so far, you haven't made any cogent points. 

[podkall]

2 minutes ago, tkitch said:

So are you just intel fanboy-ing or are you trying to make a point?

Because so far, you haven't made any cogent points. 

I mean my point is how ridiculous it is to think that such electricity draw that heats up CPU to 95°C is just 65W TDP, kind of my point

[igormp]

3 minutes ago, YoungBlade said:

So yes, a 1500X OC'd that runs at 90W is not going to be even close to what a 5900X with a 105W TDP will draw, as its max is actually 142W. But I'm assuming that the CPUs are actually drawing the same amount - not that they have the same "TDP" or "PL2" or whatever flavor-of-the-week name Intel and AMD want to give to their power draw figures.

In this case, then you're right, yes. Power is power as long as we're talking about the actual power and not the bogus TDP numbers from the manufacturers.

 

On the other hand, when talking about heat, even if both CPUs are using the same power ,a newer 7000 CPU is going to run hotter and be harder to cool than a 1000 series one due to its higher thermal density (smaller transistors = less area, so you get a smaller point packing the same amount of heat).

[tkitch]

1 minute ago, podkall said:

I mean my point is how ridiculous it is to think that such electricity draw that heats up CPU to 95°C is just 65W TDP, kind of my point

Do you realize how much heat 65W is in something that is less than 1 cubic inch?

That's a LOT of heat in a very very small thing.  

[YoungBlade]

2 minutes ago, podkall said:

I mean my point is how ridiculous it is to think that such electricity draw that heats up CPU to 95°C is just 65W TDP, kind of my point

The 65W versions don't get anywhere close to the 95C limit, even with the stock coolers, unless your case has no airflow or you're living in the jungle.

 

The 65W TDP isn't 65W, but it's not anywhere near the 250W+ that the X series ones can draw.

[podkall]

Just now, tkitch said:

Do you realize how much heat 65W is in something that is less than 1 cubic inch?

That's a LOT of heat in a very very small thing.  

yes, and law of thermodynamics:

 

6c12t 4,4Ghz boost - 65W TDP (AMD 5 5600)

 

12c24t 5,4Ghz boost - 65W TDP (AMD 9 7900)

 

1 minute ago, YoungBlade said:

The 65W versions don't get anywhere close to the 95C limit, even with the stock coolers, unless your case has no airflow or you're living in the jungle.

 

The 65W TDP isn't 65W, but it's not anywhere near the 250W+ that the X series ones can draw.

That is true, and I know Moore's law too,

 

but Moore's law isn't always exponential increase of one value without the increase of another value...

[igormp]

1 minute ago, podkall said:

yes, and law of thermodynamics:

 

6c12t 4,4Ghz boost - 65W TDP (AMD 5 5600)

 

12c24t 5,4Ghz boost - 65W TDP (AMD 9 7900)

7000 series is on a smaller node, requiring less power to reach the same performance targets of previous gen.

 

Boost is mostly for a single core for a short duration, hammer both those CPUs on an all-core load and they won't be getting much near those boosts while trying to cap themselves in order to be within their PPT limits (which are indeed higher than the TDP values, fwiw, but less than 120W iirc).

[tkitch]

8 minutes ago, podkall said:

yes, and law of thermodynamics:

 

6c12t 4,4Ghz boost - 65W TDP (AMD 5 5600)

 

12c24t 5,4Ghz boost - 65W TDP (AMD 9 7900)

Except you're wrong?  And objective testing proves it?

 

 

[podkall]

3 minutes ago, tkitch said:

Except you're wrong?  And objective testing proves it?

 

 

3:35 "enabling PBO blasted the power and pushed the power to 200Watts"

[YoungBlade]

14 minutes ago, podkall said:

That is true, and I know Moore's law too,

 

but Moore's law isn't always exponential increase of one value without the increase of another value...

You realize that the boost if for lightly threaded situations, right? So only when one or two cores are in use does it actually get to the full boost value. Otherwise, it stays around 3.7GHz for all-core.

 

2 minutes ago, podkall said:

3:35 "enabling PBO blasted the power and pushed the power to 200Watts"

Overclocking the CPU by removing power limits does, in fact, make it draw more power.

 

News at 11

[HanZie82]

power = power * efficiency   (not really but think on it )

90W is 90W

But what a CPU can do with said 90W depends very much on many things.
a 20 year old CPU would still use up that 90Watts but do very little with it, compared to one that is new.

[YoungBlade]

2 minutes ago, HanZie82 said:

power = power * efficiency   (not really but think on it )

90W is 90W

But what a CPU can do with said 90W depends very much on many things.
a 20 year old CPU would still use up that 90Watts but do very little with it, compared to one that is new.

This is true, which is why someone would want to upgrade to the newer CPU. My confusion is when I see people saying "you shouldn't upgrade from your 1600X to a 5700X because that's more cores and the CPU is newer, so your motherboard won't be able to handle it" when the difference in power draw is actually the reverse - the 1600X draws more power at stock than the 5700X.

[podkall]

17 minutes ago, YoungBlade said:

Overclocking the CPU by removing power limits does, in fact, make it draw more power.

it had like near 90W wtihout the PBO so..

 

18 minutes ago, YoungBlade said:

You realize that the boost if for lightly threaded situations, right? So only when one or two cores are in use does it actually get to the full boost value. Otherwise, it stays around 3.7GHz for all-core.

so? new architecture doesn't mean it goes substantially faster without drawing more power, maybe slightly or moderately faster sure, but substantially? no

 

riddle me this:

new architecture CPU has 6c12t 4,7Ghz base clock - 105W TDP

 

another new architecture CPU has 12c24t 3,7Ghz base ckick - 65W TDP

 

how come the CPU with double the cores and threads, same architecture, not even quarter decrease in clock has 40 less TDP wattage?

[YoungBlade]

3 minutes ago, podkall said:

it had like near 90W wtihout the PBO so..

 

so? new architecture doesn't mean it goes substantially faster without drawing more power, maybe slightly or moderately faster sure, but substantially? no

 

riddle me this:

new architecture CPU has 6c12t 4,7Ghz base clock - 105W TDP

 

another new architecture CPU has 12c24t 3,7Ghz base ckick - 65W TDP

 

how come the CPU with double the cores and threads, same architecture, not even quarter decrease in clock has 40 less TDP wattage?

Because the second CPU by default draws about 90W at max load and the first CPU by default draws around 145W at max load.

 

The real riddle is why that makes the "TDP" 65W and 105W - you'd have to ask AMD.

[HanZie82]

5 minutes ago, podkall said:

riddle me this:

new architecture CPU has 6c12t 4,7Ghz base clock - 105W TDP

 

another new architecture CPU has 12c24t 3,7Ghz base ckick - 65W TDP

 

how come the CPU with double the cores and threads, same architecture, not even quarter decrease in clock has 40 less TDP wattage?

Chiplets

die lay out

die thickness

die style 8nm vs 10nm etc.
and more

So many reasons why that could be.

Also VERY IMPORTANT to note is all manufacturers have their own way of measuring what the TDP would be.
So besides giving a hint as to what cooling it needs, its totally arbitrary those numbers.

(Cant wait until some awesome people would put some unbiased actuall numbers out there. Looking at you LTT-lab. heheheh )


A CPU has so many parts in it, its insane to try and think of.
Theres billions of transistors in there, so many things that can be the culprit.

[podkall]

2 minutes ago, HanZie82 said:

A CPU has so many parts in it, its insane to try and think of.
Theres billions of transistors in there, so many things that can be the culprit.

lay out? if you layout same CPU differently it's same CPU, unless you change the overall surface size of course or put 2 things that get really hot next to each other

[HanZie82]

3 minutes ago, podkall said:

lay out? if you layout same CPU differently it's same CPU, unless you change the overall surface size of course or put 2 things that get really hot next to each other

Besides you sentence being really hard to read.
you didnt even quoted the lay-out you respond too. (apologies if i misunderstand)

But nowhere did i say it would be a different CPU, just that it could have different heating characteristics.