We've been hearing about Raptor Lake for over a year and how it will address the shortcomings of Alder Lake. Today we can tell you everything about Intel's newest beast. Conceptually, Raptor Lake is very similar to Alder Lake. You get a single, big, monolithic die of silicon—unlike AMD's chiplet approach on Ryzen, which combines multiple smaller, specialized silicon pieces. Intel is also continuing to split their processor config into P-Cores ("Performance") and E-Cores ("Efficient"). The idea here is to push workloads onto cores that are optimized to provide maximum performance for foreground, interactive, applications, or higher efficiency, for background tasks.

If we take a closer look at the architectural details, we can see that Intel has improved the P-Core's L2 cache size considerably, to 2 MB per-core, up from 1.25 MB on Alder Lake. The E-Core L2 cache has been doubled to 4 MB per four-core cluster, and there's various small architectural improvements. Intel also increased the L3 cache size and bumped the clock speeds by quite a bit. Contrary to earlier rumors, Raptor Lake retains the ability to run with either DDR5 or the cheaper DDR4 memory, when paired with the right motherboard. All 13th gen processors are compatible with older motherboards using the Z690/H670/B660/H610 chipsets.

The Core i9-13900K is Intel's flagship, and comes with a big core-count increase over 12900K. While the 12900K was a 8+8 configuration with 24 threads total, the 13900K is 8+16, which ups the thread count to 32, reaching numeric parity with AMD's Ryzen 9 7950X. Averaged over our application benchmarks the new Intel CPU can indeed match the 7950X, even though it comes with a mix of P-Cores and E-Cores, while AMD offers 16 high-performance cores in their competitor. If we take a closer look at individual applications we find the 13900K well ahead of the 7950X in most workloads with the exception of highly compute intense tasks such as rendering, that fully load all available cores, all the time. The vast majority of applications today will not load all cores to the max, which is why we're trying to have a mix of many different scenarios in our test suite. Intel's strong single-threaded performance, paired with ultra-high boost clocks of up to 5.8 GHz is how the 13900K achieves such impressive scores in lower-threaded workloads, and the large number of E-Cores available helps with multi-threaded performance. I have to admit I was sceptical of the hybrid core approach first, but it seems that the E-Cores are a great innovation that help Intel stay competitive.

In gaming, the 13900K does very well, too, which is somewhat expected, given its Alder Lake heritage. Across all our games at 1080p resolution, the Core i9-13900K ends up 4% faster than the 12900K, which makes it 8% faster than AMD's Zen 4 Ryzen 7000 offerings. While this gap is significant, it very much depends on the game. If you take a closer look at our gaming results, it is clear that there are titles that run better on AMD, also some really like the big cache of the Ryzen 7 5800X3D, others simply run best on Intel, and there's also a bunch that show barely any differences. As resolution is increased, the bottleneck shifts more and more to the GPU, which reduces the deltas between various processors in our test group. At the 4K Ultra HD resolution, the separation between all the top CPUs are just a few percent. These are results with an RTX 3080, RTX 4090 results will come soon in a separate article, it will be interesting to see if a much more powerful GPU can affect this conclusion in any noteworthy way.

Intel is using an improved "Intel 7" process for the 13900K, which is 10 nanometer technically. With Zen 4, AMD has advanced to a 5 nanometer node, so they definitely have a process advantage. In our power consumption testing, the Core i9-13900K doesn't do so well. Averaged over all our application tests we saw a power draw of 170 W for the CPU alone, with peaks up to 285 W. AMD's offerings are much more gentle in their power requirements: 7950X 125 W avg, 235 W max, 5950X 87 W avg, 118 W max. The 12900K uses 133 W on average and 257 W maximum, so Intel's power draw is even higher than that. The 13900K does offer much better performance of course, so the energy efficiency, or "performance per watt", is also worth considering. Energy efficiency in applications looks much better for Raptor Lake, the 13900K roughly matches the 12900K efficiency, but falls behind AMD Zen 4 by a fairly big margin. Gaming efficiency is the lowest of all modern CPUs, 13900K offers 2.1 frames per Watt, 12900K and 7950X offer 2.45 frames per Watt—a 17% difference.

I'm not sure if power consumption is a deal breaker for 13900K—seems we've all got used to power consumption increases. I've hardly seen anyone in the last months who complained about the power draw of their 12900K or Zen 4 processor. NVIDIA released GeForce RTX 4090 just a few weeks ago, and people still want the card, even though it consumes 400-500 W during gaming. That still doesn't mean we should blindly accept ever-increasing power numbers. Besides the obvious "power bill" argument, there is also the problem of cooling. Keeping 13900K at good temperatures requires an excellent cooling solution. Even at stock, with a good CPU cooler, our processor reached 100°C during heavy workloads and started to throttle a little bit. Unlike on AMD, Intel does allow you to increase the temperature limit, up to 115°C, which helps avoid throttling. High temperatures seem to be the new normal anyway. AMD's Zen 4 is even designed to operate at 95°C, and when cooler, the CPU will aggressively boost the clocks to maximize performance until it reaches that temperature. Pairing the 13900K with watercooling is always an option, but it will not magically solve all temperature problems, rather it will give you a few degrees better temperatures to help avoid that throttling point.

Thanks to an unlocked multiplier, manual overclocking is easy on the 13900K. I've reached 5.6 GHz all-core stable, the E-Cores went up to 4.4 GHz. While these numbers definitely sound impressive, they are still a lot lower than the 5.8 GHz max boost of the 13900K, and this will cost you quite a bit of performance in low-threaded workloads and games, during which the CPU will regularly boost that high. With a manual all-core OC you're losing these boosts. A manual OC can still make sense for workloads like rendering or encoding, where all cores will be fully loaded all the time. When overclocked, the 13900K is able to beat the 7950X in these scenarios, too. Still, and considering the issues with power and heat, I feel like for the vast majority of users, manual overclocking makes little sense. Processor manufacturers have become really good at eking the last bits of performance out of their processors, at default settings.

We tested how well the 13900K holds its "up to 5.8 GHz" promise, and I have to say I'm a bit disappointed. While we can confirm that 5.8 GHz is reached, it is reached only on two of eight processor cores, even when only that just one core is loaded. As soon as more than two cores are active, the CPU goes to 5.1 GHz, and holds that even when all eight P-Cores are fully loaded (good). I still would have like to see a more fine-grained clock frequency scaling like the one offered by AMD on Zen 4. The other six cores can definitely run at higher speeds than 5.1 GHz, or we wouldn't have reached 5.6 GHz on all-cores with manual OC. Intel's per core overclocking can make sense here, keeping the 5.8 GHz two core boost, while increasing the speeds of heavier core loads, for example setting all eight cores to boost to 5.6 GHz.

According to Intel, the MSRP for the 13900K is $590, we're seeing higher prices on some stores right now, who are probably just wanting to cash in on the pre-orders. At around $600, the 13900K is a formidable processor at a price point that's priced aggressively compared to AMD's Ryzen 9 7950X ($700). Against the Ryzen 9 7900X ($550), the 13900K can also score, because it offers better performance across the board, at only slightly higher pricing. AMD recently launched their Zen 4 platform and motherboards are super expensive. Even the "cheapest" B650 chipset motherboard costs well over $200, while Intel motherboards can be found for around $100. Sure, these might not be the latest and greatest Z790, but there won't be any big differences when opting for a cheaper B660 board for example. The biggest selling point of Z790 is support for Gen 5 M.2 NVMe SSDs—an ability that AMD natively offers on their Zen 4 platform. Z790 also adds an extra USB 3.2 20 Gbps port, and trades eight PCIe Gen 3 downstream lanes for Gen 4. On Raptor Lake, adding Gen 5 M.2 means stealing some PCIe lanes from the graphics card, so when installing an SSD in the M.2 Gen 5 slot of your motherboard, your graphics card will run at PCIe x8 instead of PCIe x16 (without SSD in that slot, the GPU will run at full x16). This makes little real-life difference, but it feels like Intel didn't properly budget their PCIe 5.0 lanes in the CPU, unlike AMD who offers this capability and still gives you x16 on the graphics card.