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.

 

[Carfax83]

There they are measuring total system power which makes it very difficult to compare CPU efficiency as the choice of motherboard can easily swing power usage up or down tens of watts.

I agree with this criticism. I also think they should have gone into more depth on the performance per watt aspect for other applications other than CBR23.

 

[Abwx]

Im comparing the equivalent power limits. At eco 65w the 7950x is 7% more efficient than the 13900k at 65w, is it not? Thats the exact same number pc world measured in their review.

Your comparison is flawed cause you are pushing 60% extra wattage to the 13900k. Of course in that scenario its less efficient and it shows your bias. Cause let me flip the script, i bet you a paycheck the 13900k at 90w is going to curbstomp the 7950x at 145w in efficiency as well 😄

When set to 65W the 13900K actually use 71W while the 7950X use 90W, hence the comparison is not valid because the CPU using less power will be inherently more efficent due to physics laws, each time you reduce perf by 10% you increase efficency by roughly the same amount.

If you decrease the 7950X power down to 71W its efficency will increase by a factor of roughly 40% and we ll get back to my number.

Comparisons should be done at same perfs or eventually at same wattage, i made a comparison at same perfs but there s also a comparison at same power that can be done from the graph.

 

[Just Benching]

When set to 65W the 13900K actually use 71W while the 7950X use 90W, hence the comparison is not valid because the CPU using less power will be inherently more efficent due to physics laws, each time you reduce perf by 10% you increase efficency by roughly the same amount.

Oh really? That didnt seem to stop you though, in the last post you compared amd at 90w with intel at 150w. You even declared a winner. Now that I flipped the script its the laws of physics and how is poor amd supposed to overcome them . Gotcha 😁

 

[Just Benching]

Anyways, back to anandtechs data

7950x at 105w = 35975 score @ 145.6w =
247.08pts per watt

13900k at 105w = 29372 scors @ 118.6w =
247.65pts per watt



Clearly saving hundreds of watts x 3 daily. No questions about it.

 

[Carfax83]

It sure will be nice for the X3D parts to take the performance crown while being more power efficient *AND* not requiring $350 premium DDR5 memory to get that last 1% of gains which all two of the RTX 4090 enjoyers who game at CPU-limited resolutions swear by. Then whichever flavor of CPU you prefer, they're all bound to get cheaper, right? Win/win.

This may surprise you, but because of ray tracing, being CPU limited is much more prevalent now than it has ever been. That HWUB graph has already been heavily critiqued because they never tested any CPU bound scenarios, which is why the graph looks as bunched up as it does.

If you're GPU limited, then the CPU hardly matters at all so why they even bothered with all the testing if it's meaningless in the end? To do proper CPU testing, reviewers need to find the most CPU limited areas in the games and also enable settings like ray tracing which increase CPU dependency. Compare the HWUB graph to something like what PCGH.de does with their testing.

This is a worst case scenario for Zen 4, because The Witcher 3 Next Gen has terrible CPU optimization which makes it hammer 2 cores (due to non native DX12), and when you factor in the BVH calculations from ray tracing and the increased crowd density and draw distances from the next gen enhancements, then it's literally the perfect storm for CPU testing.

Here the 13900K is a good 40% faster than the 7950x at 720p to maximize CPU dependency. Currently, no CPU can maintain 60 FPS in this title at all times and even at 4K you can be CPU bottlenecked.

 

[Just Benching]

This may surprise you, but because of ray tracing, being CPU limited is much more prevalent now than it has ever been. That HWUB graph has already been heavily critiqued because they never tested any CPU bound scenarios, which is why the graph looks as bunched up as it does.

If you're GPU limited, then the CPU hardly matters at all. To do proper CPU testing, reviewers need to find the most CPU limited areas in the games and also enable settings like ray tracing which increase CPU dependency. Compare the HWUB graph to something like what PCGH.de does with their testing.

This is a worst case scenario for Zen 4, because The Witcher 3 Next Gen has terrible CPU optimization which makes it hammer 2 cores (due to non native DX12), and when you factor in the BVH calculations from ray tracing and the increased crowd density and draw distances from the next gen enhancements, then it's literally the perfect storm for CPU testing.

Here the 13900K is a good 40% faster than the 7950x at 720p to maximize CPU dependency. Currently, no CPU can maintain 60 FPS in this title at all times and even at 4K you can be CPU bottlenecked.

Most surprising is the 12900k result. 4400 ddr5 ram (lol), still delivers.

 

[Hitman928]

Anyways, back to anandtechs data

7950x at 105w = 35975 score @ 145.6w =
247.08pts per watt

13900k at 105w = 29372 scors @ 118.6w =
247.65pts per watt



Clearly saving hundreds of watts x 3 daily. No questions about it.

But in your case, the 7950x would complete its tasks 22.5% faster which means either you can be 22.5% more productive in a day, or the system gets to go to sleep sooner while the Intel system is still computing away so the total energy used will be significantly in favor of the AMD system.

Realistically, equalizing for efficiency is kinda silly because there is no inherit limit or motivation to do so. You can equalize for power use if you are power/cooling limited or, if there is a specific performance level you need to reach, you can equalize for performance and then see which CPU uses the least amount of power to reach that performance. If you equalize for power, then AMD comes out ~15% ahead in efficiency (this is for their current desktop chips in MT situations, I expect laptop and server comparisons to go much more in AMD's favor). If you equalize for performance, AMD comes out significantly more ahead due to the fact that both CPUs are pushed well beyond their efficiency points to the point where AMD at 90W can match Intel at 140W. Most people will want to equalize for power so it makes more sense to use that method. As such, AMD has an efficiency lead, but it's not all that significant for the desktop SKUs.

 

[Abwx]

Oh really? That didnt seem to stop you though, in the last post you compared amd at 90w with intel at 150w. You even declared a winner. Now that I flipped the script its the laws of physics and how is poor amd supposed to overcome them . Gotcha 😁

You got it right, the 13900K need 150W to perform like a 7950X at 90W or so.

 

[Just Benching]

But in your case, the 7950x would complete it's tasks 22.5% faster which means either you can be 22.5% more productive in a day, or the system gets to go to sleep sooner while the Intel system is still computing away so the total energy used will be significantly in favor of the AMD system.

Realistically, equalizing for efficiency is kinda silly because there is no inherit limit or motivation to do so. You can equalize for power use if you are power/cooling limited or, if there is a specific performance level you need to reach, you can equalize for performance and then see which CPU uses the least amount of power to reach that performance. If you equalize for power, then AMD comes out ~15% ahead in efficiency (this is for their current desktop chips in MT situations, I expect laptop and server comparisons to go much more in AMD's favor). If you equalize for performance, AMD comes out significantly more ahead due to the fact that both CPUs are pushed well beyond their efficiency points to the point where AMD at 90W can match Intel at 140W. Most people will want to equalize for power so it makes more sense to use that method. As such, AMD has an efficiency lead, but it's not all that significant for the desktop SKUs.

I probably dont disagree with anything you said. Not sure the difference in iso wattage is 15% but ill take it, its close to my figures so whatever (from what I've observed its between 7 and 12% depending on wattage used but sure, ill go with 15).

Thats not a big gap that saves you 300 watts per day like some... uhm... brand agnostic users are claiming. Its a gap, its respectable, but sad it only applies to heavy mt.

Basically the 7950x has all the pros and cons of a server cpu. Good at parallel and mt workloads, but pretty terrible at simple day to day stuff that home users do daily. The 13900k is a more balanced approach when it comes to average users. I mean that 40 extra wattage for browsing stigmatised me, i knew there was a difference, didnt expect anything near that. Its absurd.

 

[Just Benching]

You got it right, the 13900K need 150W to perform like a 7950X at 90W or so.

The 13900k at 90w is more efficient than the 7950x at 150 watts.

 

[Markfw]

Anyways, back to anandtechs data

7950x at 105w = 35975 score @ 145.6w =
247.08pts per watt

13900k at 105w = 29372 scors @ 118.6w =
247.65pts per watt



Clearly saving hundreds of watts x 3 daily. No questions about it.

That is no where in that article.

 

[Rigg]

Yeah I don't disagree with this at all, although it's a bit more nuanced. Both CPUs have their strengths and weaknesses. If someone was doing lots of rendering or folding, I would tell them to get a 7950x without hesitation.

But there is so much room for tweaking and improving with Raptor Lake when it comes to power consumption and thermals. For example, here is a CB23 run at 5.3ghz 235w.

Spoiler: CB23 @ 5.3ghz 235w air cooled

What do the score/temps look like when sustained for 10-30 minutes? Single run CB23 scores can be pretty far from actual sustained performance. Especially if near the thermal limit. The render doesn't take long enough to heat soak the cooler. Nice score regardless.

 

[Hitman928]

I probably dont disagree with anything you said. Not sure the difference in iso wattage is 15% but ill take it, its close to my figures so whatever (from what I've observed its between 7 and 12% depending on wattage used but sure, ill go with 15).

Thats not a big gap that saves you 300 watts per day like some... uhm... brand agnostic users are claiming. Its a gap, its respectable, but sad it only applies to heavy mt.

Basically the 7950x has all the pros and cons of a server cpu. Good at parallel and mt workloads, but pretty terrible at simple day to day stuff that home users do daily. The 13900k is a more balanced approach when it comes to average users. I mean that 40 extra wattage for browsing stigmatised me, i knew there was a difference, didnt expect anything near that. Its absurd.

That 40W from the video is total system power which is not an accurate way to compare CPU power draw or efficiency. Differences in the motherboard could easily account for most or practically all of that difference. If you look at CPU only power between a 13900k and 7950x for single threaded loads, the difference is in the single digits, i.e., not that significant and no one cares except Intel fanatics who want to find any corner case they can to try and make a case for why Intel CPUs are superior.

 

[Carfax83]

What do the score/temps look like when sustained for 10-30 minutes? Single run CB23 scores can be pretty far from actual sustained performance. Especially if near the thermal limit. The render doesn't take long enough to heat soak the cooler. Nice score regardless.

Good question. I'll do a 10 minute run and report back. BTW, check the Raptor Lake official thread. Some good info in the last few pages about tweaking DDR5. My memory performance is significantly better now:

 

[Carfax83]

Most surprising is the 12900k result. 4400 ddr5 ram (lol), still delivers.

Yeah, Alder Lake is still formidable in CPU bound scenarios even when it's crippled by relatively slow DDR5. Strap some DDR5 6400 to it and it will blast off.

 

[Rigg]

Good question. I'll do a 10 minute run and report back. BTW, check the Raptor Lake official thread. Some good info in the last few pages about tweaking DDR5. My memory performance is significantly better now:

Yep I saw that. Looking good. I was going to tell you in the build thread pretty much exactly what @Herald told you in the official thread. I saw that he already gave you some memory tuning tips that were getting good results so I didn't take the time to reply.

 

[Abwx]

The 13900k at 90w is more efficient than the 7950x at 150 watts.

That s one of the most stupid thing i ever had to read.

At 90W the 13900K is more efficent than at 150W and it works the same way for the 7950X and any other CPU since efficency decrease with increasing power, so you had to ressort to a total non sense to believe having a point.

Fact is that at same perfs the 7950X is more than 50% more efficient, this is the only valuable comparison for whom understand something about perf/watt.

 

[DrMrLordX]

Anyways, back to anandtechs data

Again you don't understand TDP limits on the different platforms. You can't take a 7950X with a TDP of 105W and compare it to a 13900k with a PL1=PL2=105W and then do efficiency comparisons between the two because they'll use different amounts of power. The 7950X doesn't gain much performance at higher power levels in MT scenarios, making it a huge winner at lower power draw vs. 13900k. The only way you can make the 13900k look efficient is to put the 7950X further out of its ideal range and then make perf/watt comparisons, which is what you're using the AT data to to (disengenuously). Also the amount of power a CPU on wall power uses in "lightly threaded" scenarios is basically irrelevant.

You're contributing nothing and talking in circles.

So the ppt limited zen 4 didnt consume vastly more than the limit it was set, it was all my misrepresentation.

I mean come on.....

The AT article isn't using a PPT limit, it's using a TDP limit.

 

[Carfax83]

Yep I saw that. Looking good. I was going to tell you in the build thread pretty much exactly what @Herald told you in the official thread. I saw that he already gave you some memory tuning tips that were getting good results so I didn't take the time to reply.

10 minute run generated a score of 39,709. I opened the window and let some cool air in though while the test was running. I can't say enough good things about the Fractal Design Torrent and the Noctua NH-U12A. An exceptional combo for air cooled setups.

Spoiler: CBR23 10 minute run

 

[IEC]

Utilizing an environmental temperature of 51F/10.5°C to augment your cooling and then claiming "air cooling" is a bit disingenuous.

 

[Rigg]

Utilizing an environmental temperature of 51F/10.5°C to augment your cooling and then claiming "air cooling" is a bit disingenuous.

I might have to see what CB23 score my air cooled 13700k can get if I set it up in my garage the next time we have a below zero day . My single run score is 1000-1500 points higher than my sustained scores depending on how I configure it. Might be because of more aggressive than stock thermal limits. I don't give a crap what AMD/Intel are saying about 95-100c throttling under load being okay now I'm not running my CPU that hot.

10 minute run generated a score of 39,709. I opened the window and let some cool air in though while the test was running. I can't say enough good things about the Fractal Design Torrent and the Noctua NH-U12A. An exceptional combo for air cooled setups.

Spoiler: CBR23 10 minute run

That's cheating!

I like really want to like the torrent. The cable management is absolute trash though. A couple more mm of space for cables in the back and it would probably be my favorite case. I really like everything else about it. Having tried lots and lots of cases in different builds in the past few years I prefer the OG meshify C or the Lian Li 011 series cases to just about anything else I've tried. Personally it think Thermalrite and Scythe make heatsinks every bit as good as (and in some ways superior to) Noctua, but their fans are pretty hard to beat.

Anyway, back to our regularly scheduled programing.

 

[Carfax83]

Utilizing an environmental temperature of 51F/10.5°C to augment your cooling and then claiming "air cooling" is a bit disingenuous.

Who said my environmental temperature was 51F? That's outside. My heat is set to 72c, but I always have the duct shut off in my office when it's cold outside. The actual temperature inside my office was roughly 67c, and that's with an 8 inch opening in the window. I'm not trying to have any insects fly through my window LOL!

That said, if this isn't air cooling, what is it?

 

[Carfax83]

I might have to see what CB23 score my air cooled 13700k can get if I set it up in my garage the next time we have a below zero day . My single run score is 1000-1500 points higher than my sustained scores depending on how I configure it. Might be because of more aggressive than stock thermal limits. I don't give a crap what AMD/Intel are saying about 95-100c throttling under load being okay now I'm not running my CPU that hot.

Having cool ambient temperatures definitely helps any air cooled setup. But honestly, I think the case is the biggest factor, followed by the fan and heatsink orientation. The high amount of air pressure and velocity generated by the intake fans (2x 180mm and 3x 140mm) gets pulled in by the heatsink and is then swiftly vacated out to the rear exhaust fan (1x 140mm) by the high speed Noctua fans. When I put my hand behind the exhaust fan, the air felt really warm during the 10 minute run. If you're sensitive to noise though, you probably would hate my rig. Personally, white noise doesn't bother me so I crank up the fans to 100% and leave them there.

That's cheating!

I prefer the term "advantageous behavior."

I like really want to like the torrent. The cable management is absolute trash though. A couple more mm of space for cables in the back and it would probably be my favorite case. I really like everything else about it. Having tried lots and lots of cases in different builds in the past few years I prefer the OG meshify C or the Lian Li 011 series cases to just about anything else I've tried. Personally it think Thermalrite and Scythe make heatsinks every bit as good as (and in some ways superior to) Noctua, but their fans are pretty hard to beat.

I've always been horrible at cable management anyway, so it works for me. One other thing I really like about the Fractal Torrent is that due to the air pressure and velocity, dust cannot settle anywhere. I built this machine last year in November, and I blow it out once a week with an electric air blower. Before I blow it out though I inspect the interior for dust, and I have never seen dust build up. Not even on the power cables connected to the motherboard and GPU. Dust just cannot find a foothold anywhere.

 

[VirtualLarry]

I prefer the term "advantageous behavior."

That's what She said!...

 

[cortexa99]

The zen4 idle power seems quietly got fixed by AGESA patch long before. I assume some amateur reviewers didn't update BIOS before testing.


Computerbase reported 13-20 watts idle power for Zen4 series while 12900KS has 11 watts idle, at the meanwhile CB R23 1T burns roughly the same watts for 7950X and 12900KS. Although CB 1T is not a good test and doesn't reflect realworld performance, IMO.

AMD Ryzen 7000 im Test: So schnell sind 7950X und 7700X: Leistungsaufnahme und Effizienz

Ryzen 9 7950X & Ryzen 7 7700X im Test: Leistungsaufnahme und Effizienz / TDP und PPT steigen deutlich

www.computerbase.de

I don't know why this become a debatable topic suddenly.