Question Anandtech.com article on 13900k and 7950x power scaling

[adamge]

A Lighter Touch: Exploring CPU Power Scaling On Core i9-13900K and Ryzen 9 7950X

www.anandtech.com

It's an interesting article, but IMO severely crippled by the reliance on "configured" power limit rather than measuring power usage. It's even more frustrating when one of the datapoints is measured power usage, so the author has access to this data.

In this scenario where configured power limits are basically hand waving guidelines, measured power usage (specifically, average power usage by the CPU during the benchmark run) is the only useful, meaningful, and concrete way to analyze the performance results relative to the power usage.

 

[Hulk]

As I wrote in the comments section for the article I'm grateful for any new content on Anandtech. But with that being said...

1. More power usage data points needed to draw conclusions. 125W and then a jump to 230W. Too large. There is a lot going on in that gap that needs to be investigated.
2. Benchmarks aren't the be all end all application for power usage. Outside of benchmarks not a lot of applications use all of the cores or even use all of the resources in a core. So while 5.5GHz all cores loaded under Cinebench might result in a ridiculous power usage for the 13900K, in a real world application such as an NLE those frequencies are reached an maintained with a package power of 160W in many cases. This can be seen in the gaming benches, which are real world and things even out quite a bit.
3. Bursty loads aren't taken into account either and it should be noted that cranking power up to 250W for 3 seconds every now and then while working so you aren't waiting on the computer might be a worthwhile tradeoff in terms of keeping a creative flow going.

 

[Markfw]

@Hulk While you may have a point or two, I would have to say that I agree with the article overall. The 7950x is more efficient from almost any angle. I run my 3 at 142 watt, since that is Co -25, and I set temps to 85. It absolutely kills my 5950x's in anything, and I have used the avx-512 at times.

 

[Mopetar]

As I wrote in the comments section for the article I'm grateful for any new content on Anandtech. But with that being said...

1. More power usage data points needed to draw conclusions. 125W and then a jump to 230W. Too large. There is a lot going on in that gap that needs to be investigated.

It's not ideal to have a gap that large, but from the data that's available it's pretty easy to conclude that AMD pretty much hits a wall at 105W and that Intel can at least scale past that point, even if it's not particularly efficient.

At least there are results in the other direction and performance is measured below 65W. Most people aren't likely to tune to those settings, so it's more academic than anything, but I still found it interesting.

 

[nicalandia]

Whoever argues in favor of the 13900K Being more efficient than the 7950X is delusional. No amount of tweaking will make the 13900K be more efficient than a similar tweaked 7950X

 

[lopri]

I find it peculiar what he describes as "scaling better" is the one that does not respond to increased voltage. His test shows 13900K scales well with power while 7950X does not. What that tells me is that AMD should have released 7950X as a 105W part, not 170W (230W actual) part.

And his 13900K running y-cruncher is only at 86C.. hmm I remember my 13700K getting close to 110C. Well, with an air cooler, so there's that. He's using an AIO 360mm.

I do not think the testing methodology is clearly explained in the article. The author says he is reading AIDA64 which in my experience corresponds to HWiNFO64's reading. It seems to me this chart shows how AMD's Eco Mode operates. (e.g. Eco Mode 65W in BIOS -> 90W PPT in Windows) I thought setting 90W PPT in BIOS resulted in 90W in Windows as well according to HWiNFO64. I might be misremembering it.

 

[Hulk]

@Hulk While you may have a point or two, I would have to say that I agree with the article overall. The 7950x is more efficient from almost any angle. I run my 3 at 142 watt, since that is Co -25, and I set temps to 85. It absolutely kills my 5950x's in anything, and I have used the avx-512 at times.

That's a fair comment. I just wanted more power points tested and more real world applications.
Most of the games tested showed the same result so instead of beating a dead horse why not test some other actual applications?

 

[TESKATLIPOKA]

I see a big problem with R9 7950X or rather R9 7945HX's temperature.

In laptops you won't have such a powerful cooling solution.
So you are most likely limited to 65W(88W PPT).
But Intel can be set to even 125W.

Performance will be in favor of Intel in CB R23.

 

[Markfw]

I see a big problem with R9 7950X or rather R9 7945HX's temperature.

In laptops you won't have such a powerful cooling solution.
So you are most likely limited to 65W(88W PPT).
But Intel can be set to even 125W.

Performance will be in favor of Intel in CB R23.

I did not see any laptop benchmarks. I don't know how you can say what will happen. Not only is cooling different, the chips are different. The only article's I saw that had real chips, heavily favored AMD's 7000 series.

 

[scannall]

I see a big problem with R9 7950X or rather R9 7945HX's temperature.

In laptops you won't have such a powerful cooling solution.
So you are most likely limited to 65W(88W PPT).
But Intel can be set to even 125W.

Performance will be in favor of Intel in CB R23.

Not so sure about that. I have a Ryzen 5800, and a Ryzen 6900HX. The 5800 idles at 18-20 watts. The 6900HX idles at 3ish. And performs better than the 5800. https://browser.geekbench.com/v5/cpu/compare/17339395?baseline=19505769 Both systems are at stock settings. The 6900HX tops out at 60 watts, and the 5800 at 78.

 

[TESKATLIPOKA]

I did not see any laptop benchmarks. I don't know how you can say what will happen. Not only is cooling different, the chips are different. The only article's I saw that had real chips, heavily favored AMD's 7000 series.

I am talking about Dragon Range. That's Raphael or not?
In most laptops is an AIR cooling solution and It even has to cool the GPU. This simply can't compare to EKWB EK-AIO Elite 360 D-RGB 360mm which only cools the CPU.
These charts are telling us a lot.

 

[TESKATLIPOKA]

Not so sure about that. I have a Ryzen 5800, and a Ryzen 6900HX. The 5800 idles at 18-20 watts. The 6900HX idles at 3ish. And performs better than the 5800. https://browser.geekbench.com/v5/cpu/compare/17339395?baseline=19505769 Both systems are at stock settings. The 6900HX tops out at 60 watts, and the 5800 at 78.

Those are only 8C16T Zen3+ at 6nm process, heat is not that concentrated.
What I am talking about is Zen4 5nm 16C32T.
You have that chart about temperatures and that CPU should be the same as R9 7945HX with lower TDP.

 

[scannall]

Those are only 8C16T Zen3+ at 6nm process, heat is not that concentrated.
What I am talking about is Zen4 5nm 16C32T.
You have that chart about temperatures and that CPU should be the same as R9 7945HX with lower TDP.

What I am saying is you can't infer much from that. Until the mobile parts are in the wild and actually benchmarked then it's speculation.

There is a lot of cheer leading for peoples favorite team happening around here lately. How about be happy that there are great things happening. Both Intel and AMD have compelling offerings, so buy according to your needs, and it will be fine.

 

[lopri]

Trying to infer laptop thermals and power consumption from this is unwise in my view.

Going back to the y-cruncher peak power, I hope there is a clarification with regard to testing methodology. That chart can invalidate the rest of the charts.

Edit: It is peak power, so I guess it really doesn't matter either way.

 

[moinmoin]

I am talking about Dragon Range. That's Raphael or not?

That's the same dies, but very likely not the same behavior set in the firmware. E.g. STAPM (Skin Temperature Aware Power Management) is very likely a major limiting factor in Dragon Range that wasn't in Raphael.

 

[lopri]

System integrators can also shut off iGPUs which appear to be a power hog (even when not used) with new Ryzens.

 

[DrMrLordX]

Performance will be in favor of Intel in CB R23.

Nah, we have forum users hitting better scores than that 38k on their 7950X @ below 200W.

 

[Panino Manino]

I see a big problem with R9 7950X or rather R9 7945HX's temperature.

In laptops you won't have such a powerful cooling solution.
So you are most likely limited to 65W(88W PPT).
But Intel can be set to even 125W.

Performance will be in favor of Intel in CB R23.

These graphs don't show clocks.
I think Zen 4 uses much less power to hit a certain clock and performance, it's just that when a certain clock is reached the power consumption and temperature rises disproportionally. Lower the clock a bit and the temperature problem is solved.
Also, mobile APUs are monolithic.

Didn't AMD made exceptional claims about the battery autonomy?

 

[Thunder 57]

These graphs don't show clocks.
I think Zen 4 uses much less power to hit a certain clock and performance, it's just that when a certain clock is reached the power consumption and temperature rises disproportionally. Lower the clock a bit and the temperature problem is solved.
Also, mobile APUs are monolithic.

Didn't AMD made exceptional claims about the battery autonomy?

The 7945 won't be. It is very similar to a 7950X except it will be in a laptop. The new APU is the 7940 which is monolithic, limited to eight cores, and includes the new XDNA AI part.

 

[IEC]

If you're willing to go beyond 180W, Raptor Lake can scale a bit further than Zen 4 with the extra power limit.

Whereas you're likely hitting either temp limit or v/f cliff with Zen 4. You're getting a few dozen MHz for the last 50%+ extra power, i.e. not worth it. If you cap it to reasonable power limits you'll use fewer joules to do the same task and the efficiency really shines.

AnandTech article just highlights what those of us who have played with Zen parts know for quite some time: best thing to do for efficiency is to set a power limit much more sane than the stock limits, e.g. 90-125W.

It's why I'm excited to see what the X3D parts will do at their 120W TDP, as that should be significantly closer to the sweet spot based on my testing.

 

[poke01]

Whoever argues in favor of the 13900K Being more efficient than the 7950X is delusional. No amount of tweaking will make the 13900K be more efficient than a similar tweaked 7950X

One main reason is node. Intel 7 is horrible for efficiency.

 

[adamge]

2. Benchmarks aren't the be all end all application for power usage. Outside of benchmarks not a lot of applications use all of the cores or even use all of the resources in a core. So while 5.5GHz all cores loaded under Cinebench might result in a ridiculous power usage for the 13900K, in a real world application such as an NLE those frequencies are reached an maintained with a package power of 160W in many cases. This can be seen in the gaming benches, which are real world and things even out quite a bit.
3. Bursty loads aren't taken into account either and it should be noted that cranking power up to 250W for 3 seconds every now and then while working so you aren't waiting on the computer might be a worthwhile tradeoff in terms of keeping a creative flow going.

I think the ideal methodology was nailed down back in the days when Anandtech.com had robust smartphone testing. Was it Brian that used to do that? They had benchmarks with a set amount of work to compute, and they measured total energy used, power throughout, and time elapsed to complete. Ultimately if we want to purchase the most efficient CPU for our tasks, we need this information to evaluate the optimal choice.

@Hulk While you may have a point or two, I would have to say that I agree with the article overall. The 7950x is more efficient from almost any angle. I run my 3 at 142 watt, since that is Co -25, and I set temps to 85. It absolutely kills my 5950x's in anything, and I have used the avx-512 at times.

I don't think you can make that conclusion from this article. The article clearly shows that a Ryzen 7000 set to X power usage is perhaps using something unknown but perhaps 1.8X power, so the datapoint showing it is faster at that setting is basically meaningless.

 

[lopri]

I think peak power is not very meaningful in this context. What is strange is that on my board 7700X follows the power limit I set in BIOS to the T. But there are several parameters to choose from, and some are more direct and some are indirect.

I can imagine certain scenarios where Raptor Lake's efficiency shines, owing to the E cores. It's just that its P cores are very power hungry. For mixed use scenarios, Raphael and Raptor Lake are a wash for both performance and power use. (imo of course)

 

[dttprofessor]

It's unfair,power limit is unblieveable.

 

[Markfw]

I think the ideal methodology was nailed down back in the days when Anandtech.com had robust smartphone testing. Was it Brian that used to do that? They had benchmarks with a set amount of work to compute, and they measured total energy used, power throughout, and time elapsed to complete. Ultimately if we want to purchase the most efficient CPU for our tasks, we need this information to evaluate the optimal choice.



I don't think you can make that conclusion from this article. The article clearly shows that a Ryzen 7000 set to X power usage is perhaps using something unknown but perhaps 1.8X power, so the datapoint showing it is faster at that setting is basically meaningless.

What ? I made the conclusion long ago from the FACTS that I have observed from my own systems. All this article did was confirm what I know.