Intel Core i9-13900K Processor Review – Intel Back On Top
Competition is Good – 13th Gen Performance is Great
Intel’s 13th Gen Core processors are here, and we will take a look at the new flagship part with this review of the Core i9-13900K. This product formerly known as Raptor Lake builds on its Alder Lake predecessor by doubling the E-core count, while retaining the eight performance cores of the 12th Gen Core i9 CPUs. We are now looking at a 24-core, 32-thread part comprised of 8 Performance-cores and 16 Efficient-cores.
Clock speeds are another area that gets a boost with the Core i9-13900K, with a Max Turbo Frequency listed at 5.80 GHz. The P-cores can boost up to 5.40 GHz with the 13900K – up from 5.10 GHz with the 12900K, and the E-cores can boost up to 4.30 GHz – up from 3.90 GHz with the Core i9-12900K. In fact, these i9-13900K frequencies offer a notable increase from even the limited-edition Core i9-12900KS.
The 13th Gen Intel Core unlocked desktop CPU lineup
Intel didn’t just increase Performance-core clocks and Efficient-core counts, however, as the new Raptor Lake architecture offers significantly faster Efficient-cores, as well – now clocked up to 600 MHz faster than we saw with Alder Lake’s E-cores, and with a “significantly optimized prefetcher algorithm” among the improvements, according to Intel. All of this adds up to the E-core getting a bigger role in overall performance this generation.
So, we know there are higher clocks, and more (and faster) Efficient-cores with this new 13th Gen Core i9-13900K, compared to the flagship Alder Lake Core i9 processors. Will this result in enough of a performance boost to rival AMD’s Ryzen 7000 Series parts? Well, from the title you probably already know that. But is Intel on top in every test? Let’s find out.
We received both the Core i9-13900K and Core i5-13600K processors, and will have a look at the 13th Gen Core i5 in an upcoming review
Some Performance Benchmarks
First, a note about the test results to follow – which serves as a cautionary tale about benchmarking pre-release hardware. Our Core i9-13900K sample was tested using an MSI MAG Z690 TOMAHAWK WIFI motherboard, updated to the latest available BIOS at the time, version 7D32vH8. This was dated 9/14/22, and updated CPU microcode. It was the latest available revision after the first “ready for next-gen CPU” BIOS version, 7D32vH7.
Naturally, I discovered while writing this that MSI had provided an update on 10/17/22 (as in, two days ago) to version 7D32vH91. This happened after I had completed all testing. Thus, all of the results here could be invalid, as the update description reads, simply: “Update CPU Micro code”. Fantastic! We will probably have to re-visit performance later.
|PC Perspective Test Platform|
|Motherboard||MSI MAG Z690 TOMAHAWK WIFI (DDR5)
BIOS 7D32vH8 (09/14/22)
|Memory||G.Skill Trident Z5 Neo 32GB (2x16GB) DDR5-6000 CL30|
|GPU||NVIDIA GeForce RTX 4090 FE
Game Ready Driver 522.25
|Power Supply||be quiet! Dark Power 12 1500W|
|Operating System||Windows 11 Pro (Build 22000.978)|
We will first look at the venerable Cinebench R20 benchmark, which still works perfectly well to demonstrate relative performance between processors.
Do you remember when Intel’s position was that Cinebench R20 didn’t represent “real-world” performance (see 2019 IFA presentation – PDF)? Well, I wonder how they feel now that they are back on top. Yes, Intel is at the top of the chart for both single-core and multi-core performance in Cinebench. (Side note, Intel actually has a page devoted to CPU benchmarking, wherein 7-zip, Blender, and Handbrake are classified as “real-world” benchmarks.)
Cinebench R23 tells the same story, but unfortunately this is a far less comprehensive chart due to a lack of certain important processors on hand. Yes, I am making excuses, but we don’t have all of the processors from the first chart on hand to re-test everything on R23. We borrowed systems with the Ryzen 9 5950X and Ryzen 7 5800X3D, and have long since returned them. I just need to buy some CPUs!
Anyhow, here’s an incomplete Cinebench R23 chart, which at least compares the Core i9-13900K to some AMD Ryzen 7000 and 12th Gen Intel processors:
Another Cinebench version, and another easy victory in both single-core and all-core performance. And this is over the mighty 16-core, 32-thread AMD Ryzen 9 7950X processor! Clearly, the clock speed boosts, along with any architectural improvements, have paid off handsomely for Intel. AMD seems to have a processor core advantage, but doubling the E-cores with the 13900K has apparently more than made up for the eight P-core limitation, at least with this test.
How about another CPU benchmark result? This time it’s Blender:
So close, but in this Blender CPU cycles workload AMD’s Ryzen 9 7950X was … 3 seconds faster. All results are the average of three runs, but this was a consistent – albeit small – victory for AMD.
Next up, 7-Zip. AMD typically wins handily in this benchmark – at least with decompression.
As usual, AMD is on top here, though the Ryzen 9 7950X’s margin of victory over the Core i9-13900K is small on the more demanding compression component of the benchmark. We saw the same thing with the Core i9-12900KS vs. the Ryzen 9 5950X, but there’s no denying the advantage in the less CPU-intensive decompression operation, which always seems to scale with core count.
Finally, a look at the x264 HD benchmark:
Intel was faster in pass 1, but AMD came back to win pass 2, which is the more challenging part of the benchmark. However, AMD’s margin of victory in pass 2 is only a few FPS, and the two are very evenly matched when you consider the total encode time. For Intel it was 123.75 seconds with the Core i9-13900K, and for AMD it was 121.70 seconds with the Ryzen 9 7950X.
A Quick Look at Gaming Performance
We looked at a few CPU benchmarks, and so far performance from Intel’s latest flagship is very impressive, trading blows with AMD’s flagship at the top of the charts. But what about gaming? Did I take it upon myself to install the world’s fastest graphics card in each system, and then run game benchmarks at ridiculously low resolutions and quality settings? Yes. Well, with one game. But first, the mandatory synthetic results.
Before addressing the results, let me just say that I was running the latest available BIOS for each system when I performed these 3DMark benchmarks, including AGESA 188.8.131.52 Patch A on our MSI X670E ACE motherboard (BIOS v1.25). The systems were outfitted with the exact same DDR5-6000 CL30 memory, running the same build of Windows 11 Pro, and had the same NVIDIA drivers installed (the latest, GRD 522.25).
Yes, the margins are pretty slim. The results above are each the result of three runs, averaged – except for Speed Way, which was so close I ended up running it five times on each system, averaging the highest three scores from each. Speed Way is still within the margin of error, and can be considered a tie – as it should, considering it will be completely GPU bound.
When not GPU bound, it seems that Intel’s Core i9-13900K certainly has the potential to provide better performance in games than the Ryzen 9 7950X. Time Spy is the closest approximation of a more CPU-bound scenario in the above chart, but what about a real game?
Having dropped Cyberpunk 2077 from the benchmark suite after its buggy performance for myself and other reviewers at the RTX 4090 launch, I decided to re-visit a reliable benchmark with the RTX 4090 FE, and Metro Exodus is as reliable as it gets (there’s a standalone benchmark app and you don’t even have to be online for it to work!).
As with the 3DMark tests above, the Core i9-13900K battles the Ryzen 9 7950X using our GeForce RTX 4090 FE card and the latest driver:
I think it’s fair to say that these two CPUs are really, really evenly matched in gaming workloads with a high-end GPU. At 720/low the results are effectively tied, and at 1080/normal the margin of victory for Intel is barely beyond the margin of error.
I know this is just one game, but I’m running short on time. I think a more notoriously CPU-intensive title like Microsoft Flight Simulator would make a good test, but I need to re-test on new BIOS versions anyway, so that can be a followup article.
Processor Voltage, Frequency, and Power Usage
While at first glace it may appear that Intel is again using the Intel 7 process from the 12th Gen Core parts, the Core i9-13900K is based on “Intel 7 Ultra”. This refined process helps explain how the Core i9-13900K is able to reach higher clock speeds than its predecessor, but it is not the full story as we will see.
“…the new process is a full PDK update over the one used by Alder Lake, their 3rd generation SuperFin Transistor architecture. Intel says this process brings transistors with significantly better channel mobility. At the very high end of the V-F curve, the company says peak frequency is nearly 1 GHz higher now. The curve itself has been improved, shifting prior-generation frequencies by around 200 MHz at ISO-voltage, or alternatively, reducing the voltage by over 50 mV at ISO-frequency.”
Now, had Intel been content with around a 200 MHz increase at the same voltage, we would not have a part with the level of performance we are seeing from the Core i9-13900K. This part does not appear to be limited to the voltage levels of its predecessor, and in fact feels like a pre-overclocked part.
I find it increasingly hard to rely on HWINFO64 logging these days (perhaps this is a personal problem), but if it can be trusted here is a comparison of 12th Gen Core and 13th Gen Core i9 processor Vcore during a Blender workload:
As you can probably make out, the Core i9-13900K is using quite a bit more voltage than the Core i9-12900K, though it did complete the test in considerably less time (the chart is truncated). It appears, according to the Vcore column in the HWINFO64 log file, that the 12th Gen part draws 1.25 V under an all-core load (Blender, Classroom CPU cycles render). The Core i9-13900K, on the other hand, is consistently over 1.3 V, and shows pretty large spikes (to 1.37 V) at each end of the test run.
These spikes are actually single-core voltage, recorded when the test first launched and again when it stopped. For a better look at how voltage draw differs between single-core and all-core workloads, the two images below show CPU-Z screenshots captured first during a Cinebench R23 single-core benchmark (left), and then during a the all-core R23 benchmark (right):
These numbers seem a bit more reasonable, but we are still looking at 1.364 V during a single-core workload, and 1.296 V during an all-core workload. You may also have noticed that the clock speeds appear to be identical in the above screenshots, but that is just a look at the frequency of Core #0. Clocks actually hit nearly 5.8 GHz at times during testing, on a pair of cores.
These HWINFO logs have per-core CPU clock data, naturally, which I will present as two separate charts. The first attempts to display the P-core frequencies:
The CPU frequency spikes more or less correspond to the voltage chart earlier in this section, and we see that those momentary spikes in voltage were not wasted, as frequencies approached the 5.8 GHz limit of this processor with those lines at the top representing a recorded value of 5785.8 MHz.
The next chart rather needlessly charts E-core frequencies (spoiler, they are all the same, averaging just under 4.3 GHz):
As to power draw, this CPU, if HWINFO64 logs can be believed, was pulling as much as 325 watts total package power under load. This is with our MSI board configured with the “liquid cooler” setting, which allows the processor to operate without any kind of power limit (unless you consider 4096W for PL1 and PL2 a limit). Back when I first tested the 12th Gen Core i9-12900K, I noted that this processor pulled right up to the max 241W limit. Again, the Core i9-13900K feels like an overclocked part right out of the box – if you set up your power limits like I did (running without limits is the default for this board, by the way).
As a footnote to this power draw discussion, I provide the following total system power draw numbers, taken while benchmarking the RTX 4090 in the Intel and AMD systems for the 3DMark results:
Ryzen 9 7950X + GeForce RTX 4090:
- 3DMark Speed Way 596 W
- 3DMark Time Spy Extreme 654 W
Core i9-13900K + GeForce RTX 4090
- 3DMark Speed Way 560 W
- 3DMark Time Spy Extreme 607 W
These results were the highest reading from 3 runs of each test, using a watts up? PRO meter. How is it possible that the Intel system is drawing so much less power under these gaming loads, when the total draw was – apparently – up to 325W in an all-core scenario like Blender? I think the answer has to do with workload, as gaming just isn’t as intensive as an all-core render.
In any case, it seems like – at least with current microcode for both processors – the Core i9-13900K may be the more efficient choice for gaming, even when the performance is identical.
Today’s review feels incomplete, as there is so much more testing that can be done – particularly with gaming workloads. But in the days leading up to this review I was continually impressed by the outstanding performance of this new 13th Gen part, which really does make up the ground that AMD just gained last month with the Ryzen 7000 Series launch.
Intel’s Core i9-13900K trades blows with AMD’s fastest CPU in every test, and even wins Cinebench for the first time in years. In the gaming workloads we did test the Core i9-13900K came away with a slight edge over the Ryzen 9 7950X, and did so while drawing less power from the wall.
Not only is the Core i9-13900K capable of matching – or exceeding – the performance of AMD’s Ryzen flagship, depending on the workload, but does so at a lower price. As I write this the Intel Core i9-13900K is listed at $659.99 USD, while the AMD Ryzen 9 7950X is $699 USD. It’s also worth noting that Intel’s existing Z690 platform is fully compatible with the new CPU after a quick BIOS flash, and these boards are considerably more affordable than an X670E solution right now.
Who would have guessed a decade ago that Intel would end up being the better value for gamers? Or that Intel is actually less expensive at the high end, including platform board? It’s crazy, but that’s the world we are living in right now. I’m guessing AMD will respond, but as long as the consensus amongst influential reviewers is similarly positive, Intel’s 13th Gen flagship it may be considered the CPU to get this generation.
The Core i9-13900K is a very impressive processor, and Intel has clearly done a lot of work to get their current process to support a product like this. It feels so close to the edge that I wonder if there’s any overclocking headroom with this CPU. I mean, voltage and power seem to have been pushed to their limits – but it worked.
If you’re on the fence about your next CPU upgrade, I think Intel just made your decision a lot easier.
This is what we consider the responsible disclosure of our review policies and procedures.
How Product Was Obtained
The product is on loan from Intel for the purpose of this review.
What Happens To Product After Review
The product remains the property of Intel but is on extended loan for future testing and product comparisons.
Intel had no control over the content of the review and was not consulted prior to publication.
PC Perspective Compensation
Neither PC Perspective nor any of its staff were paid or compensated in any way by Intel for this review.
Intel has not purchased advertising at PC Perspective during the past twelve months.
If this article contains affiliate links to online retailers, PC Perspective may receive compensation for purchases through those links.
Raw performance is there but given the transistor count of these CPUs, having to disabling AVX512, and the power consumption really shows how hard Intel had to push things to be competitive. For that reason alone I like AMDs architecture and approach more than what Intel is doing with the 12th and 13 gen.
AMD is also pushing things quite far considering the voltage and power from their 5nm chips. Ryzen 9 7950X pulls 250W all-core, and Intel is getting slammed for hitting over 300W all-core … still on 10nm.
Another excellent review Sebastian, Intel is back baby, with very impressive results showing how close it really is at the top end of CPU performance.