AMD Ryzen 9 7950X

[Count von Schwalbe]

Really, we are moving potential energy around (differential charge, that is how data is stored) by pushing it with more energy. The push is what is lost as heat.

How the heck did we derail this far?

 

[Valantar]

Really, we are moving potential energy around (differential charge, that is how data is stored) by pushing it with more energy. The push is what is lost as heat.

How the heck did we derail this far?

Because some people apparently need explanations to go all the way to these levels from time to time.

 

[zlobby]

In terms of energy, rendering (or other computing workload) is a side effect of the work done, not the work itself. Work in the physics sense is energy changing from one form to another. All the electrical energy that goes into a computer has to come out in another form to "do" anything; that is, perform work. That ends up being kinetic energy in the fans and the air they're moving (we can consider the sound generated by fans and HDDs as kinetic, but there's hardly enough of that to matter), a small amount of EM radiation from the various electrical components, a few milliwatts as signaling to your monitor or whatever, and maybe another mW or three as light from LEDs. The rest becomes heat. Those are the only options. Unless you've got some weird custom setup with a rechargeable battery, none of the energy becomes chemical, and if you're turning a bunch of energy into light, something has gone VERY wrong.

Long story short: >95% (est) of the electricity drawn by a PC becomes heat.

Energy conservation is universal. We don't get energy out of nothing. It is just converted from one form to another. In SI energy is measured in [J].
Work is the rate at which one form of energy is converted to another, i.e. the rate at which work is done. Unit would be [J/s].

And here is where our agreement ends.

Speaking of a CPU, you basically reduce it to a room heater! Last time I checked my 2kW room heater produced very low FPS in Crysis.

We have found clever ways to utilize the electrical work done (i.e. pushing electrons and holes around) by electrical energy to do some neat calculations for us. It is a direct result, not a byproduct.

Due to material imperfections, some of this very same energy is converted to heat.
Some of it is converted to EM radiation by variations in the electric field (electric and magnetic fields are coupled). Variations, i.e. distruptions by particles carrying electrical charge.

The sum of those different kinds of output energies (electrical, thermal, EM) would be exactly equal to the energy input of the system, in our case - a CPU.

So ... uh ... work is not energy. Energy is used to do work - to convert energy from one form to another. That is how that rendering (for example) gets done. In the process of doing work, electrical energy is converted to heat energy, which is a lower form of energy than electricity. Take a light bulb: electric energy is used to trigger some form of light-generating process (incandescent, LED, whatever). Some portion of said electrical energy is converted to electromagnetic radiation in the visible light spectrum. Some portion of it is turned directly into heat. As that visible light is spread out and hits other things, eventually it will all be absorbed by surrounding objects, and, again, turn into heat. CPUs do not give off light, or any other meaningful form of energy. As I said, some negligible amount of electrical energy is kept as that and transferred off-die for I/O - writing to RAM or storage, transferring data over PCIe, etc. This amount in watts is tiny compared to the heat output of the CPU.


Exactly. We've essentially found a bunch of smart ways of converting energy from one form into another that leave us with really useful byproducts. Like converting fuel to heat, while leaving us with rapid movement or cooked food; or converting electricity to heat and leaving us with lights, warm (or cool) houses, and RGB. Lots of RGB.

Eh...

Because some people apparently need explanations to go all the way to these levels from time to time.

Bruh, you were insisting that all electrical energy that enters in a CPU is dissipated as heat! Irony much?

 

[80-watt Hamster]

Energy conservation is universal. We don't get energy out of nothing. It is just converted from one form to another. In SI energy is measured in [J].
Work is the rate at which one form of energy is converted to another, i.e. the rate at which work is done. Unit would be [J/s].

1 J/s = 1 Watt. So far we're on the same page.

And here is where our agreement ends.

Judging by the below, we agree on every point but one...

Speaking of a CPU, you basically reduce it to a room heater! Last time I checked my 2kW room heater produced very low FPS in Crysis.

We have found clever ways to utilize the electrical work done (i.e. pushing electrons and holes around) by electrical energy to do some neat calculations for us. It is a direct result, not a byproduct.

Due to material imperfections, some of this very same energy is converted to heat.
Some of it is converted to EM radiation by variations in the electric field (electric and magnetic fields are coupled). Variations, i.e. distruptions by particles carrying electrical charge.

The sum of those different kinds of output energies (electrical, thermal, EM) would be exactly equal to the energy input of the system, in our case - a CPU

... and that is this: In regard to energy and power, calculations are not work. Let's compare a classic seismograph (I'm going to get the terminology wrong. You have been warned).

The energy from an earthquake is transferred to the scribe arm, which drives the stylus across the drum leaving a mark. Most of the energy is dissipated in the drag of the stylus, which is where the useful work happens. It is then up to a technician to interpret what's left. In a computer, the CPU is the arm and stylus, and your output device is the drum. Think of it this way: the energy to perform the calculations: where is it stored? A video file has no stored energy, nor does the image on a screen. The energy used needs to GO somewhere. And where is that if not heat?

And yes, a processor is basically a space heater that can do lots of logic operations per second.

 

[Valantar]

Speaking of a CPU, you basically reduce it to a room heater! Last time I checked my 2kW room heater produced very low FPS in Crysis.

And last I checked, room heaters were pretty much giant resistors, not CPUs - so they work quite differently, even if both convert electric energy to heat.

We have found clever ways to utilize the electrical work done (i.e. pushing electrons and holes around) by electrical energy to do some neat calculations for us. It is a direct result, not a byproduct.

That is literally the same thing. A byproduct is a direct result of whatever process creates it.

Due to material imperfections, some of this very same energy is converted to heat.

... What? Are you saying that with a theoretical perfect material we would have CPUs that use zero energy?

Some of it is converted to EM radiation by variations in the electric field (electric and magnetic fields are coupled). Variations, i.e. distruptions by particles carrying electrical charge.

... and how much is lost this way, exactly? If your CPU was outputting several watts of EMI, things like WiFi and other wireless communications would be impossible, as interference would overwhelm any nearby signal.

 

[THU31]

I don't understand how you expect similar efficiency when adding 8 cores but w/e, we will see. Im willing to bet a paycheck or two

You should not be surprised at what can be achieved with a mature process node, they have a lot of experience in that area.
This will be the fourth generation on 10 nm (I am counting the mobile chips).

I expect they managed to lower the voltages at the highest boost clocks. AMD on the other hand are running crazy voltages to push the clocks as high as possible. You can cut power by half and only lose 5% performance, which is insane.

I also expect that Zen 4 might be more efficient when just comparing the P-cores, especially at low TDPs.
But for productivity, E-cores are exceptionally efficient, so Raptor Lake should be on a similar level.

 

[Mussels]

If only a certained banned TPU member who kept making new accounts to argue with me was still around, he loved to say how 12th gen was more power efficient than zen 3 in gaming

AMD have finally released a CPU that's equally as power hungry!
(and at stock settings vs stock settings, intel still takes more power)



5700g is a surprise winner for efficiency here, WTAF?

 

[Valantar]

If only a certained banned TPU member who kept making new accounts to argue with me was still around, he loved to say how 12th gen was more power efficient than zen 3 in gaming
View attachment 263417AMD have finally released a CPU that's equally as power hungry!
(and at stock settings vs stock settings, intel still takes more power)



5700g is a surprise winner for efficiency here, WTAF?
View attachment 263418

While I'm not directly disagreeing with you, efficiency is more than "power draw while gaming" - performance also needs to be factored in. Divide that average power draw with the average fps of each cpu across that test suite and you'll have gaming efficiency.

 

[Mussels]

The energy used needs to GO somewhere. And where is that if not heat?

It goes back to the other pin in your wall outlet. It's like two hoses with water, pushing a waterwheel and going back out - the axle of the waterwheel heating up is independent of the water, but that resistance slows the flow down and the reduced kinetic energy is converted to heat.

The heat is from friction, that's it. It's physical movement causing friction which causes heat, which is why higher clock speeds produce more heat even at the same voltage - they're flicking on and off more often, producing more heat.

 

[W1zzard]

5700g is a surprise winner for efficiency here, WTAF?

Not a measurement error btw, it's a laptop CPU with all its power optimizations brought to desktop

 

[Mussels]

While I'm not directly disagreeing with you, efficiency is more than "power draw while gaming" - performance also needs to be factored in. Divide that average power draw with the average fps of each cpu across that test suite and you'll have gaming efficiency.

I mean, one of the charts literally is the results of gaming and gaming only.
Performance is something seperate, as that ongoing argument was 100% entirely about efficiency.
In neither single threaded, multi threaded, nor gaming only results does intel's 12th gen come ahead of zen 3.

"performance per watt" is an entirely different metric, and they don't win there either.
And again since i'm referring to a discussion in other threads: This was purely about zen 3 vs 12th gen and efficiency with gaming, if you mix more generations or tests it gets derailed into something else entirely

Not a measurement error btw, it's a laptop CPU with all its power optimizations brought to desktop

I can believe it, but it's still shockingly far ahead of everything else.

 

[nguyen]

If only a certained banned TPU member who kept making new accounts to argue with me was still around, he loved to say how 12th gen was more power efficient than zen 3 in gaming
View attachment 263417AMD have finally released a CPU that's equally as power hungry!
(and at stock settings vs stock settings, intel still takes more power)

Wait wut

12900K use 1% more power while being 4% faster than 7950X, that means 12900K is more efficient in gaming

 

[Anoldnewb]

Wrong! This would actually break thermodynamics. X watts electric cannot be equal to Y watts thermal output, and at the same time you'd get tons of rendering (for example) be done on the CPU!

 

[Valantar]

I mean, one of the charts literally is the results of gaming and gaming only.
Performance is something seperate, as that ongoing argument was 100% entirely about efficiency.
In neither single threaded, multi threaded, nor gaming only results does intel's 12th gen come ahead of zen 3.

"performance per watt" is an entirely different metric, and they don't win there either.
And again since i'm referring to a discussion in other threads: This was purely about zen 3 vs 12th gen and efficiency with gaming, if you mix more generations or tests it gets derailed into something else entirely

Performance per watt is by far the most accepted metric of efficiency for computers, and anything electrically powered, really. Unless you are factoring in performance, you aren't measuring efficiency but power draw. Those two are not the same thing.
If two electric cars drain their equally sized batteries, but one drives further, or one drives the same distance more quickly, then that car is more efficient. If you ignore the distance, speed or time driven on that amount of power, you are rendering yourself unable to discuss efficiency by default. If two marathon runners each consume 5000Kcal during the run, but one finishes in three hours and one in five, then the three-hour run has used said energy more efficiently: producing more, or higher quality work for the same energy expenditure.

You can of course argue that the baseline for "performance" in an efficiency comparison is "being able to complete the task", but that also introduces such a massive range of variance into the results that comparison becomes essentially impossible. A more practically applicable understanding of this in terms of PC gaming would be allowing for +/- 10% performance as being treated as "the same", as that's roughly the limit for what is perceptibly faster/slower. But you would still need a graph of more than just power draw to talk about efficiency.

I can believe it, but it's still shockingly far ahead of everything else.

Mobile(-based) chips generally are - they don't deliver peak absolute performance, but they're tuned for efficiency - maximizing the amount of useful work done per unit energy - so they tend to vastly outperform anything else in that regard. Desktop chips haven't been focused on efficiency for ages, as their cooling and power limits are so much higher, you can always just add a bit more.

Edit: Also worth mentioning that the 5700G CB 1t efficiency score aptly illustrates the power needs of IF in AMD's MCM CPUs. Other than the 5700G clocking marginally lower than its non-G counterparts and having half the cache, it's the same thing after all, so the only major difference is monolithic (internal fabric) vs. MCM (through-package IF). IF consumes somewhere in the realm of 20W on a 2CCD MCM Ryzen, which is massive when considering that at least Zen3 cores peak at about 20-21W themselves. Which also explains why the MCM chips overtake the mobile chips in highly threaded loads, as the detrimental effect of IF power becomes less of a factor as total power draw increases. Of course, the 5700G's cores with their relatively low clocks are likely closer to 10W than the 20-21W of a 5950X too.

 

[Anoldnewb]

Check this out for cinbench comparison of 7950 at 170, 105 and 65 watts.

Ryzen 9 7950X vs Core i9: Can ECO Mode save you money?

AMD says the Eco Mode on its new top-of-the-line Ryzen processor can beat Intel, even running at full power. It's time to put it to the test.

www.pcworld.com

 

[JustBenching]

I mean, one of the charts literally is the results of gaming and gaming only.
Performance is something seperate, as that ongoing argument was 100% entirely about efficiency.
In neither single threaded, multi threaded, nor gaming only results does intel's 12th gen come ahead of zen 3.

"performance per watt" is an entirely different metric, and they don't win there either.
And again since i'm referring to a discussion in other threads: This was purely about zen 3 vs 12th gen and efficiency with gaming, if you mix more generations or tests it gets derailed into something else entirely


I can believe it, but it's still shockingly far ahead of everything else.

What? Efficiency is by definition power used divided by work produced. Power draw on its own doesn't tell you anything about efficiency. If it did, then the 7950X would be the most inefficient cpu on planet earth, since according to gnexus, it topped the power draw charts.

 

[Daven]

In the initial 5800X3D review, the CPU was identical to the 12900K at 720p. Now its over 10% slower!!! What happened?!?!?

 

[Valantar]

In the initial 5800X3D review, the CPU was identical to the 12900K at 720p. Now its over 10% slower!!! What happened?!?!?

System updates, game updates, BIOS, microcode and AGESA updates. Things change over time. Might also be a slightly altered benchmark selection?

 

[Count von Schwalbe]

In the initial 5800X3D review, the CPU was identical to the 12900K at 720p. Now its over 10% slower!!! What happened?!?!?

Looks like the test setup is very similar in both reviews. I think the X3D moved down (not the 129K up), considering the relative positioning of the 5950X.

Probably the game selection happened to not favor a large L3 cache.

 

[Valantar]

It goes back to the other pin in your wall outlet. It's like two hoses with water, pushing a waterwheel and going back out - the axle of the waterwheel heating up is independent of the water, but that resistance slows the flow down and the reduced kinetic energy is converted to heat.

You're mixing up voltage and current. What happens to the water in a waterwheel? It is slowed down to a near stop, losing its kinetic energy as this is transferred into the wheel. Only the momentum of the wheel and the force applied by the new water coming in makes it go anywhere. Voltage is the "pressure" from oncoming water; current is the energy actually extracted from this process, including a loss of energy at "the other end". Of course, AC power oscillates back and forth 50/60 times a second, so a water wheel is a pretty bad analogy. There is no energy "going back to the other pin of your wall outlet" - if that was the case, we couldn't have fuses, as literally every wire would constantly be transmitting infinite current, and everything would be on fire. An electrical circuit will only draw the energy required/allowed by the parts in that circuit. If your circuit has a 100W motor, or light bulb, or whatever, then 100W passes through the circuit - and is converted to kinetic energy, or light, or whatever. (The vast majority of which will soon thereafter be converted to heat.) There will be no more energy passing through that circuit than what it consumes.

 

[TheoneandonlyMrK]

You're mixing up voltage and current. What happens to the water in a waterwheel? It is slowed down to a near stop, losing its kinetic energy as this is transferred into the wheel. Only the momentum of the wheel and the force applied by the new water coming in makes it go anywhere. Voltage is the "pressure" from oncoming water; current is the energy actually extracted from this process, including a loss of energy at "the other end". Of course, AC power oscillates back and forth 50/60 times a second, so a water wheel is a pretty bad analogy. There is no energy "going back to the other pin of your wall outlet" - if that was the case, we couldn't have fuses, as literally every wire would constantly be transmitting infinite current, and everything would be on fire. An electrical circuit will only draw the energy required/allowed by the parts in that circuit. If your circuit has a 100W motor, or light bulb, or whatever, then 100W passes through the circuit - and is converted to kinetic energy, or light, or whatever. (The vast majority of which will soon thereafter be converted to heat.) There will be no more energy passing through that circuit than what it consumes.

There's a curve ball in my mind, have at it.

The power doesn't travel down the wire's anyway it moves through the air around the circuit not through the circuit.

 

[Count von Schwalbe]

There's a curve ball in my mind, have at it.

The power doesn't travel down the wire's anyway it moves through the air around the circuit not through the circuit.

Not in DC. Very high frequency AC (or technically pulsed DC) will travel along the skin of the wire, but either of those will cause a CPU to let the magic smoke out.

 

[TheoneandonlyMrK]

Not in DC. Very high frequency AC (or technically pulsed DC) will travel along the skin of the wire, but either of those will cause a CPU to let the magic smoke out.

Not the skin, the rather air, but yes I agree.

 

[Valantar]

There's a curve ball in my mind, have at it.

The power doesn't travel down the wire's anyway it moves through the air around the circuit not through the circuit.

Yep, I know. And that's pretty much when I realize there's a reason why I never went into engineering

 

[W1zzard]

Probably the game selection happened to not favor a large L3 cache.

Nice observation, Far Cry 5 has been replaced with Far Cry 6, big difference in gains for the 5800X3D vs everything else

https://tpucdn.com/review/amd-ryzen-7-5800x3d/images/far-cry-5-1920-1080.png

https://tpucdn.com/review/amd-ryzen-9-7950x/images/far-cry-6-1920-1080.png