90c+ CPUs

[Deleted member 185158]

The link you provide its inaccessible to EU.
And this picture only shows that a CPU has an input of electric power and outputs the power as a heat, most of it through IHS and some of it though the socket and mainboard which is what AMD is saying indirectly with their definition of TDP against PPT and with their formula. This picture does not saying anything about efficiency.

----------------------------------

For the last time...
According to conservation of energy:
Energy cannot be created or destroyed, only can change form or be transferred to a different location or object.


In order to calculate the electric/power efficiency of a device you have to have 3 different values. 1 input and 2 output (1 input=2 output)
Out of the 2 output 1 is useful the other is considered lost.

Incandescent light bulb:
1. Input electric power (100W)
2. Light (2W) = useful
3. Heat (98W) = lost
This device has an efficiency of 2% since its used for light. The rest 98% is "lost" as heat from the conversion. Lost as not transformed into light, not really lost.

PSU:
1. Input electric power (100W)
2. Output electric power (90W) = useful
2. Heat (10W) = lost
This device has an efficiency of 90% since its used for supplying power of different voltage from the input. The rest 10% is "lost" as heat from the conversion to a different voltage. Lost as not transformed into the new voltage, not really lost.

Combustion engine with 35% efficiency:
1. Fuel+Air (Input mass)
2. Kinetic energy = useful 35%
3. Hot exhaust gasses (mass + heat) = lost 65%

Please fill the gaps below if you can
I can't...

CPU:
1. Input electric power (100W)
2. ____________ = useful)
3. ____________ = lost
This device has an efficiency of __% since its used for __________________________________. The rest __% is "lost" as heat from the conversion to _____________. Lost as not transformed into ______________, not really lost.

What 2 forms of energy are you going to put in there on 2 and 3?


I'm not convinced yet that it is really or not.

.

But that's my point.

Transistors do not USE up the current.
The wattage is transformed into another energy, basically through resistance. (Conservation of energy)
100w in = 100w out. (Or very damn near it.)
I'm no engineer, but that's the way I've always thought how processors really work.

Maybe I'm mistaken?

Here's a small forum post with a similar question.

"Where does all the power consumed by a cpu go?"

Answer
"In the CPU it's all heat. It's the changing from 0 to 1 and back (which ultimately is what a computer does) which consumes the energy, because charge has to be moved from one place to another, and it's this current (moving charge) through resistance which causes heat. P=I2×R"

Again, I have no idea how reputable this link is. But I'm trying to back my statements.
And since you cannot open links that I can, I quoted it above for you.
Source https://electronics.stackexchange.com/questions/79166/where-does-all-the-power-consumed-by-a-cpu-go

 

[Night]

Let's say that base clock current use is 64A, and that the CPU uses 1.25V, power would be 80W so I guess that's a possibility. P = I^2 * R (formula for direct current), R = U / I, which is 0,01953125 ohms so
P = 4096 * 0,01953125 = 80 W. Or more simply P = U * I.

CPUs/GPUs/etc don't work by directing the energy elsewhere, so it's not clear how you would rate their efficiency.

It's actually possible to rate their efficiency with performance per watt, and could also be expressed as FLOPS per watt.
Also like this:

Just looking at power draw in watts won't paint the whole picture for any given processor. It's not only important how much power is consumed, but also how quickly a task is completed—taking both into account results in "efficiency." Since a faster processor will complete a given workload quicker, the total amount of energy used might end up less than on a low-powered processor, which might draw less power, but will take longer to finish the test. In this section, we divide the achieved performance by the power usage, to get a Cinebench points per watt single-threaded and multi-threaded result.

 

[freeagent]

I can see your brains pulsating from here

I hate math.

 

[Deleted member 185158]

Let's say that base clock current use is 64A, and that the CPU uses 1.25V, power would be 80W so I guess that's a possibility. P = I^2 * R (formula for direct current), R = U / I, which is 0,01953125 ohms so
P = 4096 * 0,01953125 = 80 W. Or more simply P = U * I.

So is that accurate then?

The wattage in (electrically) is converted and then dissipated as a "Thermal" (Missing here: Design Point) In total TDP?

 

[Night]

So is that accurate then?

The wattage in (electrically) is converted and then dissipated as a "Thermal" (Missing here: Design Point) In total TDP?

From Wikipedia:

The thermal design power (TDP), sometimes called thermal design point, is the maximum amount of heat generated by a computer chip or component (often a CPU, GPU or system on a chip) that the cooling system in a computer is designed to dissipate under any workload.

 

[Deleted member 185158]

From Wikipedia:

Oh, right.

Then what is put into a cpu is converted and simply dissipated as a different type of energy. Nothing more to it then. That's all I was trying to say this entire time.

Put in 100w, get out 100w.
Efficiency of this transaction is simply measured by calculations per watt.

I think I better understand the thought a CPU is 100% efficient where all processing is done on all 100w now.
Figuring you can process a lot or a little, the energy USE is almost totally wasted because we get only but digital currency from the transaction of changing electrical wattage to a thermal wattage.

Cool.

 

[Zach_01]

It's actually possible to rate their efficiency with performance per watt, and could also be expressed as FLOPS per watt.
Also like this:

Yes I’ve said that several times actually.
This is different kind of efficiency though (than a PSU, or a light bulb efficiency) and has a meaning by comparison to another CPU. It’s not a percentage. That’s what I was saying.

To sum up

Every CPU will convert all its input power into heat at any given time. The difference is that some do more work and others less.
Depending on the characteristics of their process node, their architecture and of course their operating temp.

 

[SL2]

A single tower heatsink managed to run cooler, with higher clock speed, and higher CBench score than with the 360 mm Corsair H150i thanks to PBO2 undervolting Curve optimizer in BIOS.

 

[Zach_01]

A single tower heatsink managed to run cooler, with higher clock speed, and higher CBench score than with the 360 mm Corsair H150i thanks to PBO2 undervolting in BIOS.

View attachment 263900

True…
Still if you do the same with the H150i 360 you may get even better results.

 

[SL2]

True…
Still if you do the same with the H150i 360 you may get even better results.

Of course, but I think the whole point with that video was to show that you don't have to go liquid cooling, and you can still stay under 90 degrees, most likely.

He really missed the point tho with the video title, I mean how many are interested in seeing a Wraith cooler being tested?

 

[Zach_01]

Of course, but I think the whole point with that video was to show that you don't have to go liquid cooling, and you can still stay under 90 degrees, most likely.

Yes
Tuning and tweaking tends to be the real new normal with every new generation.

 

[SL2]

Yes
Tuning and tweaking tends to be the real new normal with every new generation.

Yup, it wasn't the same kind of necessity with the 5000s.

A 31 degrees drop when combined with lower power limit is more than what I expected. (OptimumTech)

 

[Zach_01]

Yup, it wasn't the same kind of necessity with the 5000s.

A 31 degrees drop when combined with lower power limit is more than what I expected. (OptimumTech)

View attachment 263907

Yeah I saw that before and two things impresses me...

1. The level of performance Ryzen 7000 offers below 100W
2. The level of inefficiency AMD chose to default them for the sake of competition

Most likely if you set this at half the wattage of the 5000 equivalent (5800X 142W) you get at least the same performance, if not more.
I guess AMD is on the edge with the Intel 13th and didn't wanted to be left too far behind until 3D V-Cache arrives at 2023.

EDIT: typo

 

[R0H1T]

A single tower heatsink managed to run cooler, with higher clock speed, and higher CBench score than with the 360 mm Corsair H150i thanks to PBO2 undervolting in BIOS.

View attachment 263900

Did he do the curve optimizer or just the undervolt? There's still lots of room for tuning if it's the latter.

 

[SL2]

Did he do the curve optimizer or just the undervolt? There's still lots of room for tuning if it's the latter.

Curve optimizer. Sorry, just a typo

 

[AusWolf]

A single tower heatsink managed to run cooler, with higher clock speed, and higher CBench score than with the 360 mm Corsair H150i thanks to PBO2 undervolting Curve optimizer in BIOS.

View attachment 263900

That's proof of what reviewers have been saying, and of what some of us have been saying in the last 10 or so pages:
1. Zen 4 will always run hot. Your cooler choice will have little to no effect on temperatures. What your cooler choice will affect, is your clock speed that goes with that temperature.
2. It also proves that the Zen 4 is massively overvolted by default.

 

[SL2]

That's proof of what reviewers have been saying, and of what some of us have been saying in the last 10 or so pages:
1. Zen 4 will always run hot. Your cooler choice will have little to no effect on temperatures. What your cooler choice will affect, is your clock speed that goes with that temperature.
2. It also proves that the Zen 4 is massively overvolted by default.

It's not really surprising, as that much higher clock speed must come from somewhere. AMD showcased the 7950X running a game at 5.5+ GHz in May.

 

[AusWolf]

It's not really surprising, as that much higher clock speed must come from somewhere. AMD showcased the 7950X running a game at 5.5+ GHz in May.

Exactly. Especially when you pair it with increased TDP and increased density of the chiplets.

 

[Upgrayedd]

I see you guys talk chips being hotter that are "denser" which to me, means packing more transistors.
That doesn't make sense to me.
Whether it has 5 transistors or 50bil, if using 100w there's 100w of heat. I don't think the silicon molecules are getting any more or less dense.

Or are you guys referring to actual die area just getting smaller? Like 100w in 200mm^2 vs 100w in 100mm^2.

 

[kapone32]

I see you guys talk chips being hotter that are "denser" which to me, means packing more transistors.
That doesn't make sense to me.
Whether it has 5 transistors or 50bil, if using 100w there's 100w of heat. I don't think the silicon molecules are getting any more or less dense.

Or are you guys referring to actual die area just getting smaller? Like 100w in 200mm^2 vs 100w in 100mm^2.

The 7950X has about 3 times the transistors vs a 5950X. The package is about the same size and you need to power every single one of them. That in turn means more gates are opening in the same space which means more heat in a defined space. Testing and time may prove that a cooler that has a base plate much larger than the actual CPU may be better able to spread that heat corresponding to lower temps. .

 

[Upgrayedd]

The 7950X has about 3 times the transistors vs a 5950X. The package is about the same size and you need to power every single one of them. That in turn means more gates are opening in the same space which means more heat in a defined space. Testing and time may prove that a cooler that has a base plate much larger than the actual CPU may be better able to spread that heat corresponding to lower temps. .

More gates but the same power draw just means each gate uses less power right? Still the same amount of energy used overall.

 

[Deleted member 185158]

I see you guys talk chips being hotter that are "denser" which to me, means packing more transistors.
That doesn't make sense to me.
Whether it has 5 transistors or 50bil, if using 100w there's 100w of heat. I don't think the silicon molecules are getting any more or less dense.

Or are you guys referring to actual die area just getting smaller? Like 100w in 200mm^2 vs 100w in 100mm^2.

The reduction of transistor size helps use less current.
The addition to transistor count adds to use more current.

90nm 50mil = 100w
5nm 50bil = 100w.

Transistor density from reduced node size, but more productive at the same wattage. Because there's more transistors.

Could you imagine how cool these chips would run if we never made it past dual core days?

In example, my 12400F is quite cool with 2 core running 5.2ghz or faster even at 1.30v. Which all core 6c 12t would be pretty hot.

Surface area would be smaller if the motherboard still housed the North Bridge chipset, but having I/O on die reduces latency quite a bit. And only a couple extra watts involved there.

 

[kapone32]

More gates but the same power draw just means each gate uses less power right? Still the same amount of energy used overall.

Even if the the smaller gates use less power the fact that there is more of them inherently means there is more power. We are basically discussing the mitigating factor of why 7000 is so much faster than 5000. Even LN could only get a 5950x to over 5 GHZ stable but that is literally the limit of the chip where as the 7950x hits 5.8 GHZ with regular cooling, even the base clock is 1.1 GHz higher. If they had done the same thing with the 5950x in terms of transistor count you would have a chip between the size of a TR4 and AM4 chip. This is why even in eco mode a 7950x will be "faster" than a 5950X. AMD has turned the chips to 11 so tuning one of these monsters could really show that these chips are one of the major benefits of competition as 13th Gen will be great too.

 

[AusWolf]

Or are you guys referring to actual die area just getting smaller? Like 100w in 200mm^2 vs 100w in 100mm^2.

Yes, though the main reason is still more power, higher clocks and voltage, I think.

 

[Upgrayedd]

Even if the the smaller gates use less power the fact that there is more of them inherently means there is more power.

Can you see how that doesn't make sense at all. Maybe processor performance but not watt power. 100w is 100w.