A Lighter Touch: Exploring CPU Power Scaling On Core i9-13900K and Ryzen 9 7950X
by Gavin Bonshor on January 6, 2023 8:00 AM EST- Posted in
- CPUs
- AMD
- Intel
- 65W
- TDP
- Ryzen 9 7950X
- Core i9-13900K
- Power Scaling
- Efficiency
- 105W
- 125W
- 253W
- 230W
- PPT
Core i9-13900K & Ryzen 9 7950X Scaling Performance: Peak Power/Temps & Gaming
In our gaming performance testing, we're using two very different titles to measure any gains (if any) when dropping the power on the Core i9-13900K and Ryzen 9 7950X. The first title is Total War: Warhammer 3, which is notoriously hungry in terms of processor performance. We felt this would be a good measuring stick to see how performance stands as we restrict power to the processor. The second title is Borderlands 3, which albeit more graphically intensive than TW: Warhammer 3, still benefits from processing power as many other titles do.
We'll first go over the peak package power and core temperatures.
Peak Power and Core Temperature: Feat yCruncher and AIDA64
One of the main benefits of reducing power consumption and electrical load on a component is temperature; less power means less heat. As we reduce the TDP and test with power restrictions, we should also see a noticeable reduction in heat and CPU core temperatures. To measure the peak package power load from the CPU and to determine the peak core temperatures, we are using AIDA64 to record both variables. Putting full load on the CPU is yCruncher, which we use to measure peak processor package power in our CPU reviews.
Starting with the peak power figures, it's worth noting that AMD's figures can be wide off the mark even when restricting the Package Power Tracking (PPT) in the firmware. For example, restricting the socket and 7950X to 125 W yielded a measured power consumption that was still a whopping 33% higher. By comparison, the 13900K exceeded its set limits by around 14% under full load. In all cases though, this is still a significant power reduction versus their stock settings, especially in the case of the power-hungry i9-13900K.
Following on from the temperatures, despite pulling a figure of 330.3 W under full load, the peak core temperature of the i9-1300K was 8°C lower than the Ryzen 9 7950X, which hit 94°C under full load. Given that the power figures given aligned more with the settings on the 13900K than they did on the 7950X, the drop in temperatures on the Intel processor was much better received, with 53°C at 125 W and just 39°C at 65 W.
There's certainly more performance at 65 W from our compute testing on the Ryzen 9 7950X, but it's drawing more power than it should be. It's also running hotter despite using a premium 360mm AIO CPU cooler, which is more than enough even at full load. As a reference, the room that all the testing was done at ranged between 22 and 24°C, so this shouldn't impact any of our thermal results too much.
Total War: Warhammer 3: 1080p Ultra and 4K Medium Settings
In Total War: Warhammer 3, we saw something very interesting. Dropping the power on the AMD Ryzen 9 7950X, even down to just 35 W, didn't seem to impact performance at either 1080p Ultra or 4K Medium settings. In both cases we're performance-bound by other factors, be it single-threaded performance or GPU performance. This is a good precedent being set here by AMD, as even at such a low power, it's not enough to warrant noticeable drops in framerates, which is partly down to utilization with our AMD Radeon RX 6960 XT graphics card.
Touching on the Intel Core i9-13900K, although average frame rates and 5% lows at 1080p remained stable and similar, the 5% lows were much less desirable when testing at 4K. Average frame rates seem stable, but having much lower 95th percentile frames could become troublesome depending on the title, visual settings, and utilization of both CPU and GPU at lower power. Though, like many benchmarks relating to games, this phenomena is going to vary on a game-by-game basis.
Borderlands 3: 1080p and 4K Ultra Settings
Looking at performance in Borderlands 3, we can see that both the Core i9-13900K and Ryzen 9 7950X perform well even at just 35 W. This shows how powerful both these chips are for gaming, with high core counts which, despite operating at much lower power than stock, that it doesn't affect performance too much.
The only real notable result in Borderlands 3 was the Intel Core i9-13900K at 1080p, with lower 5% low framerates than the Ryzen 9 7950X. Although this could be an anomaly, we tested this three times, and all the results were similar. Even so, at just 35 W, the performance in gaming was typically unaffected, which shows that the game is more often waiting on our Radeon RX 6950XT video card.
Of course, gaming performance isn't going to be too much of a war of attrition to seek benefits even at lower power envelopes on the processor at resolutions such as 1440p and 4K, where performance is primarily GPU-limited. Even at 1080p, where there's a cross-over between CPU and GPU utilization, performance is still good.
There will certainly be a big difference at lower resolutions, such as 720p and lower. Still, users looking at a $500-600 processor, a $300-500 motherboard, and $150+ for memory are highly unlikely to be gaming at these resolutions, so we focused more on the realistic scenarios in gaming as opposed to purely synthetic ones.
61 Comments
View All Comments
GeoffreyA - Saturday, January 7, 2023 - link
Thanks, Gavin, for the interesting, insightful article.sonny73n - Saturday, January 7, 2023 - link
How do you limit CPU power draw in BIOS and still getting Turbo Boost for Ryzen?t.s - Tuesday, January 17, 2023 - link
I choose to disable turbo boost. I set my Ryzen 5 5600G to 3.4GHz max, GPU to 1600. It consume about ~50 watt when playing dota 2.erotomania - Wednesday, January 25, 2023 - link
I overclock my 5600G to 4.55 GHz, iGPU also, and it consumes about 60W when gaming. Maybe 65-70 if CPU and iGPU are hammered.razvan.a - Saturday, January 7, 2023 - link
Very interesting article, thank you. If it is possible, please add to it, or in a new article, two tests for which I searched over the net, but I don’t seem to find the answers anywhere:1. Power draw in idle state in different eco modes. This is very important because in many scenarios most of the time the CPU is in idle state. For example a programmer 90% of the time only enters code, or a creator only adjusts some pixels. In the default mode it seems that Intel is more efficient, with about 10W idle, and AMD some times bigger, but I wonder if in eco modes this changes.
2. If the article is about power efficiency, can you add some graphs to show exactly this, in terms of total power consumed to finish a job (best if the total consumed power is measured at the outlet)? For example, a CPU/entire system needed 58.3Wh to render an image.
wr3zzz - Saturday, January 7, 2023 - link
I would love similar analysis on future laptop CPU. It's been several generations since the last fanless Intel y-series CPU was released and I guess fanless TDP is no longer in the roadmaps. I need a notebook CPU that can run the most tasks without needing to turn on the fans.Vitor - Saturday, January 7, 2023 - link
That 330w peak power by Intel is horrifying.Samus - Saturday, January 7, 2023 - link
This is an all-around great article that will be references for years to come. It explains a lot to amateur PC builders and gamers alike that, for one, you can scale back one component at little performance compromise, making headroom for another. Additionally, if you are hitting the thermal limits of your cooler or chassis, undervolting a bit with an offset of like -0.100 will have virtually no impact on performance but tremendous impact on overall heat production.Sunrise089 - Saturday, January 7, 2023 - link
Great article, thank you for the content Gavin and Ryan!zodiacfml - Sunday, January 8, 2023 - link
Thanks team. No one does this except Anandtech.