Conclusion

So after lots of testing at various power levels on the Intel Core i9-13900K and the AMD Ryzen 9 7950X processors, we saw some very interesting results. While it's clearly established that both brands' current generation flagships are high-performance desktop goliaths in their own right, the performance of these processors doesn't scale down with power consumption in quite the same way. And that gives us some interesting results to dissect.

For our analysis, we've split our conclusion into two parts with a final summary; this allows us to make it more distinct in terms of benefits (if any), performance, and of course, if it's simply worth reducing the power on either the Core i9-13900K or Ryzen 9 7950X.

Core i9-13900K Power Scaling Summary

Starting with Intel's Core i9-13900K, we did experience some interesting results across the multitude of benchmarks and both games we tested. While the Core i9-13900K and Ryzen 9 7950X trade blows competitively when left at stock settings, Intel's offering didn't seem quite as performance competitive (and thus overall efficiency) at lower power envelopes. 

Despite being the more power-hungry processor overall, albeit, with better heat transfer than AMD's offering as evidenced by CPU core temperatures, the Core i9-13900K didn't scale quite as effectively as we would have expected. Or, at least, it didn't retain its high performance quite as well at lower TDPs when placed opposite AMD's best chip.

Using our CineBench R23 Multi-Thread test results as our baseline for compute performance, we can see how well the Core i9-13900K scales when tasked with rendering. Using the stock settings figure as our baseline at 100% performance, we can see that dropping down to 125 W; we saw performance sit at around 78%. As we dialed down the power to 105 W, which isn't a massive shift in power consumed, we got about 72% of the performance compared to stock.

Going down to the fan-favorite 65 W point, we saw performance levels drop to 56.6% of stock performance. Which, keeping in mind this is just 26% of the rated power consumption (and less still if the yCruncher power numbers fully translate here). So on a normalized basis, the energy efficiency of the i9-13900K has more than doubled (2.2x) when dialing it down to 65 Watts – it gives up a lot of performance, but it saves even more on power consumption. And if we do rope in those yCruncher measured power figures, then we're looking at something closer to 2.62x the power efficiency versus stock, primarily on account of the i9-13900K's incredibly high stock power consumption.

Overall, at the 65 W mark we saw a peak load power of 71.4 W, and a peak core temperature of 39°C. 

In our gaming tests, the Core i9-13900K performed equally well at all of the power settings we tested, which shows that despite curtailing the power limits massively, the performance for gamers is still there; not all games are created equally, and I'm sure there will be games that may see some drops in performance. The only pitfall we saw with the Core i9-13900K in our game testing was some slightly lower 5% lows in frame rate performance, which has the potential to be concerning, but it's going to depend on the game and the minimum framerates desired (e.g. does it drop below the 60fps mark).

Ryzen 9 7950X Power Scaling Summary

Things get very interesting moving onto the AMD Ryzen 9 7950X processor and the Zen 4 core based on TSMC's 5 nm node. At the top end of the power spectrum at stock settings, the Ryzen 9 7950X shows more efficiency in terms of performance per watt than the Core i9-13900K. Still, we observed some fascinating results as we restricted the power to the AM5 socket through AMD's Package Power Tracking (PPT) technology.

The first thing I want to point out is that, going by yCruncher, the Ryzen 9 7950X uses around 34% less power at stock settings than the Core i9-13900K to achieve similar performance levels. So AMD is already starting from a better place in terms of power efficiency.

Looking at our results in CineBench R23 Multi-Threaded, the Ryzen 9 7950X scaled exceptionally well from a PPT of 230 W down to 65 W. The most important results are obtained at 65 W, with CineBench R23 MT showing that the Ryzen 9 7950X kept 81.1% of its performance in that benchmark's rendering workload when compared directly to stock performance, all with a PPT setting just 28% of the stock limit. So to put things on a normalized basis once again, AMD's power efficiency has almost trebled (2.87x), keeping a lot of performance for a sizable drop in power consumption.

The catch here, however, is that the AMD platform as a whole was far more lax in sticking to its programmed PPT values, as evidenced by yCruncher power consumption. Despite setting the 7950X to 65 W, we still measured 90.3 W under that workload. So on the assumption that translates to CineBench, the 7950X's power efficiency gains aren't as impressive; we're looking at 42% of the power consumption for 81.1% of the performance, or a power efficiency of 1.93x over stock.

Looking at the rest of the metrics at 65 W, we saw a peak load power of 90.3 W, and a peak core temperature of 52°C. 

 

Finally, in the case of both processors, performance at 35 W wasn't precisely jaw-dropping, but it has to be noted that there are diminishing returns to any form of efficiency, and 35 W was, perhaps, a stretch too far for this much physical silicon. Despite this, the Ryzen 9 7590X held 49.3% of its performance in CineBench R23 MT, and the Core i9-13900K had just 30.6% of its performance at 35 W.

Reducing The 'TDP': Watts The Point

So when it comes to reducing TDP or restricting power to the CPU, some may ask, what's the point? Well, one example of where aiming for a higher-end processor and simply limiting power to it is in small form factor PCs. Smaller chassis typically have fewer cooling options available, and for the dinky systems, running a CPU at 253 W (13900K) or 230 W (7950X) can cause as many problems as it solves in such a tight space. Add any mid-range or high-level graphics card to the mix, and things will heat up quickly. 

It's worth noting that a typical 240 mm AIO premium CPU cooler, such as Cooler Master's MasterLiquid Pro 240, can accommodate between 210 to 230 W of thermal power generated, so anything smaller that's ample for SFF computing is going to be even less effective in shifting heat from the CPU to the exterior of the chassis. It's even recommended by many that users look for premium cooling, such as a 240 mm or, ideally, a 360 mm AIO CPU cooler when using one of these flagship processors.


Screenshot from Intel's Xtreme Tuning Utility (XTU)

It's not just SFF enthusiasts who can benefit from lowering overall power consumption. Still, those who are conscious of how much energy they are using aren't going to look favorably on 300 W of power being spat out from the CPU on its own, especially if there's no alternative to dropping a small percentile of performance for a fraction of efficiency. Less heat + less power = less problems.

Final Words: Zen 4 is Very Efficient, But Everyone Improves With Lower TDPs

Despite more affordable options capable of delivering solid gaming performance (check out our latest Best CPUs for Gaming guide for more details), in regards to taking x86 computing to new heights, both Intel and AMD can be proud of their achievements. Our testing and our independent reviews of the AMD Ryzen 9 7950X and Intel Core i9-13900K processors are clear that both are the cream of the crop regarding high-performance desktop computing. In fact, at the top end of the spectrum, both processors trade blows in compute elements such as rendering, encoding, simulation, and gaming.

The biggest winner in our testing is AMD with the Ryzen 9 7950X. The biggest takeaway from our analysis is that despite curtailing the max power consumption on the Ryzen 9 7950X by over 50%, plenty of performance remains on the table to be tapped into. Seeing the chip retain 80%+ of its stock performance even when the peak power consumption is just 42% is a very encouraging outcome. This makes SFF computing more appealing, even at the higher end.

Intel's i9-13900K, in comparison, doesn't fare quite as well. Though as it's clear both platforms are running well outside their "sweet spots" on the v/f curve for these flagship SKUs, it still benefits as well: a 65 W setting still gives us around 60% of the performance for 21%(!) the measured peak power consumption. The end result is that the i9-13900K sheds more power consumption on both an absolute and relative basis than AMD's chip, giving up more performance in the process.

Just keep in mind that TDPs (and PPTs) are better thought of as an "aspirational" value than a hard limit. That goes for both stock settings, and when power limiting these chips. In the end, neither platform quite adhered to our desired power limits, with AMD's being the worse of the two. Meaning that outside of stock settings, the 7950X ended up consuming more power at any given limit. But in keeping with our desire to focus on real-world performance here, this still offers a very valid take on what to expect from dialing either processor down to "65" Watts, as I doubt very many tweakers are targeting a specific wattage, versus just dialing down their power limits to save power and boost efficiency.

Ultimately, if you're short of space right now, you can still benefit from lower temperatures and reduced power consumption with either platform. Though when it comes to overall (i.e. best) performance, the edge goes to the Ryzen 9 7950X. With a nearly 2x increase in energy efficiency shifting down to a 65 Watt(ish) PPT setting, it's clear that the chip was pushed well past its sweet spot in order to achieve the best possible absolute performance. All of which means there's a lot of energy efficiency to gain from pulling it back to that spot. No wonder AMD decided now was the time to bring one of its chiplet-based desktop CPUs to the mobile arena with the new Ryzen Mobile 7045 HX-series processors – as audacious as a 16 core CPU is in a laptop, AMD's Zen 4 silicon is up to the task.

Core i9-13900K & Ryzen 9 7950X Scaling Performance: Peak Power/Temps & Gaming
Comments Locked

61 Comments

View All Comments

  • Kangal - Saturday, January 7, 2023 - link

    Came here to say the same thing.
    Can you potentially use a Power-Meter at the wall-plug to obtain real-world (W) consumption. Run your benchmark repeatedly back-to-back in a 30min timed period and plot the average Watts used. You can also measure them at Idle, which we can subtract from the Test Figures, to account for the Energy used from the other components (Mobo, RAM, SSD, PSU). You can even unplug the dGPU for the compute tests to get even more precise readings.
  • Jase76 - Wednesday, January 11, 2023 - link

    Still no update, come on Gav, make it happen
    Power should scale fairly linear with power so I'm guessing either the AMD chips are using a lot more power than advertised, or Intel's using a lot less. Has to be worth testing! Hell, is 65W even the same on the AMD X chips as the non-X equivalents
  • Gavin Bonshor - Friday, January 6, 2023 - link

    Your point is duly noted
  • ydeer - Tuesday, January 10, 2023 - link

    Wall-power consumption normalized testing would indeed be much appreciated, because that is what ultimately counts (for me and probably many other efficiency-minded people).

    I would also be very interested to see a series like this for GPUs across different architectures and chip sizes, again perhaps with power consumption at the wall in mind. That would be a great series.
  • Thunder 57 - Friday, January 6, 2023 - link

    I would like to see this too. As for Y-cruncher, there may be a bug where when using AVX-512 PPT isn't accurately enforced. Or PPT is just broken in some cases. Either way, it's something worth looking in to.
  • Pneumothorax - Friday, January 6, 2023 - link

    Impressive performance from the 7950x at laptop power levels. Going to be exciting to see how AMD laptops do with very little power and extended battery life.
  • brucethemoose - Friday, January 6, 2023 - link

    I am salty about AMD's 230W stock limit now.

    A 30%+ increase in power for an (at best) 5% increase in performance? Thats just wasteful. If this were an older processor, I'd call that a mediocre overclock, but AMD is *shipping* them that way :/.
  • Pneumothorax - Friday, January 6, 2023 - link

    That power spike for so little gain is almost criminal. Red is taking too many lessons from Blue.
  • III-V - Friday, January 6, 2023 - link

    Don't forget Bulldozer was a thing. It's not AMD's (or Intel's) first time making a processor that runs hot
  • Sunrise089 - Saturday, January 7, 2023 - link

    If they didn’t eke out every percentage point of performance sites would post context-less charts showing Intel winning all benchmarks (versus just a majority) and sales would suffer. I wish that wasn’t the state of affairs but sadly it is.

Log in

Don't have an account? Sign up now