Intel 13th Gen Raptor Lake CPU Series Detailed

The 13th-generation Intel Raptor Lake CPUs will have a high boost frequency of 6.0 GHz, in addition to more cores, improved connectivity, a redesigned core architecture, support for PCIe 5.0 SSDs, and more. Intel says that compared to Alder Lake, Raptor Lake will have a ‘40% performance scaling,’ an increase of 15% in single-threaded performance, an increase of 41% in multi-threaded performance, and an overall ‘40% performance scaling. On October 20, these chips will launch to compete with AMD’s Zen 4 Ryzen 7000 CPUs, ushering in a new phase of competition between Intel and AMD for desktop PC dominance and, more specifically, the title of best CPU for gaming.

Intel’s Alder Lake helped them regain the lead it had been slowly losing to AMD’s rapidly improving Ryzen processors in our CPU benchmark rankings. Intel’s fall from favor following persistent delays moving to its oft-delayed and apparently fatal 10nm production node was capped in 2020 by AMD’s Ryzen 5000 CPUs, which excelled Intel’s chips in every performance, price, and power parameter that mattered at the time.

When we reached Alder Lake, everything turned around. These processors were the first to take use of Intel’s reworked 10nm technology, officially known as “Intel 7,” which allowed for increased clock speeds and reduced power consumption, paving the way for Raptor Lake. Intel’s Raptor Lake CPUs will be etched on an improved version of the same process node, and will include the company’s rethought x86 hybrid architecture. This is a design that incorporates both big and tiny cores, the former for increased speed and the latter for greater efficiency.

CPU	Price	Cores / Threads (P+E)	P-Core Base / Boost Clock (GHz)	E-Core Base / Boost Clock (GHz)	Cache (L2/L3)	TDP / PBP / MTP	Memory
Core i9-13900K / KF	$589 (K) – $564 (KF)	24 / 32 (8+16)	3.0 / 5.8	2.2 / 4.3	68MB (32+36)	125W / 253W	DDR4-3200 / DDR5-5600
Core i7-13700K / KF	$409 (K) – $384 (KF)	16 / 24 (8+8)	3.4 / 5.4	2.5 / 4.2	54MB (24+30)	125W / 253W	DDR4-3200 / DDR5-5600
Core i5-13600K / KF	$319 (K) – $294 (KF)	14 / 20 (6+8)	3.5 / 5.1	2.6 / 3.9	44MB (20+24)	125W / 181W	DDR4-3200 / DDR5-5600

Raptor Lake, like its predecessor, will enable disruptive new features like as PCIe 5.0 and DDR5, but will continue to offer DDR4 for less costly build alternatives. Raptor Lake will also be compatible with current motherboards, providing an upgrade path for Alder Lake customers. However, new 700-series motherboards will be available at launch with enhanced connection options. Intel is also offering further CPU overclocking capabilities for Raptor Lake.

We were able to locate an unknown 34-core variant at an Intel trade exhibition. Intel has finally provided us a comprehensive breakdown of what to anticipate, including price and specifications. This page contains all the information available to us from official and unofficial sources. We will update the post if additional information becomes available, but here is what we know so far.

Series at a Glance

• Intel 7 process node has up to 24 cores and 32 threads (34-core spotted)
• Up to eight Raptor Cove Performance cores and sixteen Gracemont Efficiency cores (E-Cores)
• Raptor Lake-S (65W to 125W desktop) and Raptor Lake-P (15 to 45W mobile) confirmed
• Supposed 5.8 GHz increase
• Up to 36MB of L3 Cache (a 20% boost) and up to 32MB of L2 Cache (2.3x increase)
• Support for Dual-Channel DDR4-3200 and DDR5-5600 memory, x16 PCIe 5.0 and x4 PCIe 4.0 interfaces, Thunderbolt 4 / USB 4
• PCIe 5.0 M.2 SSDs and AI M.2 Modules are supported.
• +15% single-threaded performance, +41% multi-threaded performance, and ‘40% performance scaling’ overall.
• Socket LGA 1700, compatible with current cooling, is compatible with BGA mobile chips.
• 700-Series Chipset: Motherboards with Z790, H770, and B760
• Up to 20 PCH PCIe 4.0 and eight PCIe 3.0 slots.
• Enhanced CPU overclocking capabilities, such as per-core and Efficient Thermal Velocity Boost

Specifications and Features

Recently, Intel said that their high-end Raptor Lake CPUs would achieve speeds of up to 6 GHz at default settings, and that they had established a world record for overclocking at 8 GHz with liquid nitrogen. Allen ‘Splave’ Golibersuch recently overclocked a Core i9-13900K to 8.2 GHz using liquid nitrogen (LN2), surpassing the maximum overclocking frequency of AMD’s Ryzen 7000 processors.

Notably, the maximum speed of 6 GHz is 300 MHz higher than the maximum speed of 5.7 GHz for AMD’s Ryzen 7000 CPUs, although Intel has not yet revealed which product will reach this maximum speed. We are also uncertain as to whether a 6GHz processor would come with the first batch of chips or be a limited edition ‘KS’ variant.

Intel’s published specifications indicate the Core i9-13900K as having up to 24 cores and 32 threads, the i7-13700K as having 16 cores and 24 threads, and the Core i5-13600K as having 14 cores and 20 threads.

The Raptor Lake chips will feature a 15% increase in single-threaded performance and a 41% increase in multi-threaded performance over Alder Lake, as well as a ‘40% performance scaling’ overall.

Raptor Lake also has better overclocking capabilities, support for an AI M.2 module, and Alder Lake system compatibility.

The 13th-generation Raptor Lake processors will have Performance Cores (P-cores) with a new microarchitecture called Raptor Cove. These cores are optimized for single-threaded and light-threaded applications, such as gaming and productivity. The Efficiency Cores (E-cores) also exhibit hints of a redesigned microarchitecture, such as a doubled L2 cache, however these cores retain the Gracemont architecture. These cores are used for heavy workloads, background processes, and multitasking.

CPU	Price	Cores / Threads (P+E)	P-Core Base / Boost Clock (GHz)	E-Core Base / Boost Clock (GHz)	Cache (L2/L3)	TDP / PBP / MTP	Memory
Core i9-13900K / KF	$589 (K) – $564 (KF)	24 / 32 (8+16)	3.0 / 5.8	2.2 / 4.3	68MB (32+36)	125W / 253W	DDR4-3200 / DDR5-5600
Ryzen 9 7950X	$699	16 / 32	4.5 / 5.7	–	80MB (16+64)	170W / 230W	DDR5-5200
Core i9-12900K / KF	$589 (K) – $564 (KF)	16 / 24 (8+8)	3.2 / 5.2	2.4 / 3.9	44MB (14+30)	125W / 241W	DDR4-3200 / DDR5-4800
Ryzen 9 7900X	$549	12 / 24	4.7 / 5.6	–	76MB (12+64)	170W / 230W	DDR5-5200
Core i7-13700K / KF	$409 (K) – $384 (KF)	16 / 24 (8+8)	3.4 / 5.4	2.5 / 4.2	54MB (24+30)	125W / 253W	DDR4-3200 / DDR5-5600
Core i7-12700K / KF	$409 (K) – $384 (KF)	12 / 20 (8+4)	3.6 / 5.0	2.7 / 3.8	37MB (12+25)	125W / 190W	DDR4-3200 / DDR5-4800
Ryzen 7 7700X	$399	8 / 16	4.5 / 5.4	–	40MB (8+32)	105W / 142W	DDR5-5200
Ryzen 5 7600X	$299	6 / 12	4.7 / 5.3	–	38MB (6+32)	105W / 142W	DDR5-5200
Core i5-13600K / KF	$319 (K) – $294 (KF)	14 / 20 (6+8)	3.5 / 5.1	2.6 / 3.9	44MB (20+24)	125W / 181W	DDR4-3200 / DDR5-5600
Core i5-12600K / KF	$289 (K) – $264 (KF)	10 / 16 (6+4)	3.7 / 4.9	2.8 / 3.6	29.5MB (9.5+20)	125W / 150W	DDR4-3200 / DDR5-4800
Core i5-13400 / F	?	10 / 16 (6+4)	3.4 / ?	?	24MB	65W / ?	DDR4-3200 / DDR5-5600
Core i5-12400 / F	$199 – $167 (F)	6 / 12 (4+0)	4.4 / 2.5	–	25.5MB (7.5+18)	65W / 117W	DDR4-3200 / DDR5-4800

The Core i9, i7, and i5 flagships from Intel’s previous generation are shown above, along with what we know about the forthcoming Raptor Lake processors. Intel will first release three K-series machines alongside their graphics-less KF counterparts, with a larger rollout scheduled for early next year.

The Core i9-13900K will have a total of 24 cores, comprised of eight P-cores and sixteen E-cores. This is eight more E-cores than the previous generation’s flagship (but the same number of P-cores). These extra E-cores are derived from a new, bigger 8+16 die (8 P-core + 16 E-core) that Intel will employ exclusively for its Core i9, i7, and K-series i5 processors. This bigger chip has extra cache capacity for the cores (more on this in the section on architecture), although Core i3 and below processors will have the same amount of cache as the current Alder Lake versions.

Intel may have a price edge if it can equal or surpass AMD’s Ryzen 9 7950X, which has 16 cores and costs $699. The Core i9-13900K has 24 cores and costs $589, while the Ryzen 9 7950X has 16 cores and costs $699.

The Core i7-13700K costs $409 and has a 400 MHz p-core frequency increase to 5.4 GHz, four more e-cores for a total of eight, and a 400 MHz e-core frequency increase to 4.2 GHz.

Intel has boosted the MTP for this processor to 253W, an increase of 63W over the previous generation. The 13700K rivals with the $399 Ryzen 7 7700X.

The 13600K has four more e-cores than its predecessor, 200 MHz faster p-core speeds that reach 5.1 GHz, and 300 MHz faster e-core clocks that offer a 3.9 GHz increase. The 13600K has a 31W MTP increase to 181W, indicating that Intel is likewise increasing power limitations in this regard.

While Intel has boosted the peak speeds for both p-cores and e-cores on all Raptor Lake CPUs, we notice a 200 MHz drop in the base clocks. Probably done to regulate the TDP rating, this should have little practical significance. The 13900K has the same 125W Base Turbo Power (BTP) as previously, but the Maximum Turbo Power (MTP) has risen by 12W to 253W.

The Core i3-13600K is the only processor to rise in price; both the full-featured variant and the graphics-less 13600KF are $30 more costly than the previous generation. This is by far the highest-volume SKU and is in a pricing category where AMD is not as competitive, so Intel’s price hike here will produce the greatest revenue without jeopardizing its goal of undercutting AMD at the high end. However, AMD’s $299 Ryzen 5 7600X is a powerful gaming CPU, so it will be interesting to see how Raptor Lake performs when it arrives on our test bench. The 13600K has four more e-cores than its predecessor, 200 MHz faster p-core speeds that reach 5.1 GHz, and 300 MHz faster e-core clocks that offer a 3.9 GHz increase. The 13600K has a 31W MTP increase to 181W, indicating that Intel is likewise increasing power limitations in this regard.

Certain enhancements apply to all three chips. Intel, for instance, raised the L2 cache from 1.25MB to 2MB for each p-core and from 2MB to 4MB for each cluster of e-cores. The addition of additional e-core clusters, each of which includes an adjacent L3 cache slice as part of its architecture, has also boosted the quantity of L3 cache. This results in increased cache capacities for all K-series Raptor Lake processors.

Using one DIMM per channel (1DPC), Intel’s DDR5 memory capability has been raised to 5600 MT/s, a significant improvement above the 4800 MT/s supported by Alder Lake. Intel boosted 2DPC speeds to 4400 MT/s, an improvement over the previous generation’s 3600 MT/s. Intel will also continue to support DDR4 memory, which it estimates will coexist alongside DDR5 on the market until 2024. This strategy provides a value alternative for Intel systems, in contrast to AMD’s all-in DDR5 strategy.

Intel has verified that the Alder Lake processors will be compatible with the LGA 1700 socket, making them backward compatible with the current 600-series chipsets and forward compatible with the upcoming Raptor Lake motherboards. In addition, the 16 PCIe 5.0 PCIe lanes originating from the CPU may now be partitioned into twin x8 configurations, allowing compatibility for PCIe 5.0 M.2 SSDs. This subject will be covered in more depth in the motherboard section.

Raptor Lake iGPU reportedly utilizes the same Xe-LP Gen 12.2 architecture as Alder Lake. A recent OpenGL test revealed that the Raptor Lake Core i9-iGPU 13900’s has 32 EUs operating at 1.65 GHz. This indicates that the iGPU has the same amount of cores as its predecessor, but operates 100 MHz (6.5%) quicker. Aside from modest clock speed enhancements, we do not anticipate any significant modifications to the iGPU’s architecture. Intel recently added support for Raptor Lake-P (mobile) and Raptor Lake-S (desktop) CPUs to their media driver, indicating that mobile versions of Raptor Lake are in development.

Intel has also hinted at a new M.2 slot-compatible AI accelerator. It is difficult to determine what practical function this would serve for the majority of use cases; nevertheless, certain fringe use cases may benefit. Intel has not disclosed any other information on this product, and there has been no additional information, so we must wait to discover more.

E-core E-volution

Intel’s designs for the Core i5 E-core processor have evolved. The Core i5-12600K, the most recent member of Intel’s K-series processors, features four E-cores. In comparison, the Core i5-13600K has eight E-cores. Core i5 versions that aren’t K, such as the Core i5-12400, often lack E-cores. Intel’s non-K Core i5 Raptor Lake processors are believed to be gaining E-cores, which might boost the performance of the mid-range Core i5. Because of its e-cores, Intel has the most cores for normal desktop PCs, whereas AMD’s Ryzen 7950X has a maximum of 16 cores. Despite the fact that the e-cores employ the same Gracemont architecture as the original 14nm Skylake cores, Intel claims that tweaks to the caching technique and a few other minor enhancements result in the same IPC and frequency while consuming less power.

The die modifications are very significant. Thirteen 13th-generation Intel CPUs support hyperthreading. Two of the chips will utilize Raptor Lake B0 die, while the other two will use Alder Lake C0 or B0 die. Two additional CPUs will utilize Alder Lake C0 or H0 die.

This indicates that the non-K Core i5 machines, like the i3 versions, have been upgraded with e-cores. Intel previously said that e-cores will be pushed to the bottom of the Core i5 stack, and this evidence backs up that assertion.

Model	Base Clock	L3 Cache	PBP	Silicon	Stepping
Core i9-13900K/KF	3.0 GHz	36 MB	125W	RPL B0	Raptor Lake
Core i9-13900/F	2.0 GHz	36 MB	65W	RPL B0	Raptor Lake
Core i9-13900T	1.10 GHz	36 MB	35W	RPL B0	Raptor Lake
Core i7-13700K/KF	3.40 GHz	30 MB	125W	RPL B0	Raptor Lake
Core i7-13700/F	2.10 GHz	30 MB	65W	RPL B0	Raptor Lake
Core i7-13700T	1.40 GHz	30 MB	35W	RPL B0	Raptor Lake
Core i5-13600K/KF	3.50 GHz	24 MB	125W	RPL B0	Raptor Lake
Core i5-13600	2.70 GHz	24 MB	65W	ADL C0	Alder Lake
Core i5-13600T	1.80 GHz	24 MB	35W	ADL C0	Alder Lake
Core i5-13500	2.50 GHz	24 MB	65W	ADL C0	Alder Lake
Core i5-13500T	1.60 GHz	24 MB	35W	ADL C0	Alder Lake
Core i5-13400/F	2.50 GHz	20 MB	65W	RPL B0 | ADL C0	Alder Lake | Raptor Lake
Core i5-13400T	1.30 GHz	20 MB	35W	ADL C0	Alder Lake
Core i3-13100/F	3.40 GHz	12 MB	60W/58W	ADL H0	Alder Lake
Core i3-13100T	2.50 GHz	12MB	35W	ADL H0	Alder Lake

Versus AMD Zen 4 Ryzen 7000

Here’s how the competition between Raptor Lake and Ryzen 7000 will look. Both companies are pushing the clock speeds of their modern chips to the highest levels we’ve seen. This will increase power use, especially when working on multiple tasks at once.

CPU	AMD Zen 4 Ryzen 7000	Intel Raptor Lake
Release Date	September 27th	October 20
Node / Design	TSMC 5nm Compute die, 6nm I/O DIe	Intel 7 – Monolithic Die
Cores / Threads	Up to 16 Cores / 32 Threads	Up to 8P + 16E | 24 Cores / 32 Threads
Peak Clocks	5.7 GHz	6.0 GHz
TDP / PBP / MTP	170W / 230W	125W / 253W
Memory	DDR5 Only (No DDR4 support)	DDR4-3200 / DDR5-5600
PCIe	PCIe 5.0 – 24 Lanes	PCIe 5.0 x16, PCIe 4.0 x4 (SSD)
Graphics	RDNA 2 iGPU	UHD Graphics 770

The Zen 4 Ryzen 7000 chips from AMD can only use DDR5 memory, but the Raptor Lake chips can use both DDR4 and DDR5 memory. That gives Intel an edge in the overall cost of the system, since DDR5 still costs more. But there are no longer any DDR5 shortages, and prices keep going down as more supply comes online and demand goes down. That means the price difference should get a lot smaller, but we expect DDR5 to always be more expensive.

Silicon Die Size

The size of the Raptor Lake die, which is 23.8 mm long and 10.8 mm wide. This means that the chips are a little bit longer and a little bit wider than the 12th-generation Alder Lake chips.

CPU	Die Area	Die Dimensions	Cores	Process
Raptor Lake Core i9-13900K	257 mm^2	23.8 x 10.8 mm	8 P-Cores | 16 E-Cores	Intel 7
Alder Lake Core i9-12900K	208 mm^2	20.4 x 10.2 mm	8 P-Cores | 8 E-Cores	Intel 7
Rocket Lake Core i9-11900K	281 mm^2	24 x 11.7 mm	8 P-Cores	14nm
Comet Lake Core i9-10900K	206 mm^2	9.2 x 22.4 mm	10 P-Cores	14nm

The 11900K, 12900K, and Raptor Lake Core i9-13900K are all shown here. Intel increased the number of cores in Raptor Lake from a peak of eight E-cores to a total of 16 by adding two more quad-core clusters. They also increased the L2 cache by 60% (doubling the e-cores to 4MB) and the L3 cache by the same amount.

All of these things made it necessary to make an 8+16 die that is 49mm2 bigger than the 8+8 die from the last generation. Each cluster of E-cores weighs about 8.65 mm2, which means that only 17.3 mm2 of the die size increase goes to the E-cores. The rest goes to the larger L2 and L3 caches. That’s a great example of how the e-cores are so good at using space. But making chips with bigger dies is more expensive because there are fewer dies on a wafer when the die is bigger.

On the 13900K, Intel has increased the speed of the e-cores to 4.3 GHz and raised the speed of all the cores by 600 MHz.

To take advantage of the faster memory speeds, the fabric needs to be faster. Intel has raised the ring bus frequency to 5 GHz and made it 900 MHz faster during all-core turbo. This fixes a big problem we saw with Alder Lake. Intel says that just the improvements to the ring bus can lead to a 5% increase in frame rates. We’ve also seen good gains when we overclocked this setting on Alder Lake.

Testing and IPC

Intel has provided a small number of its own benchmarks for the 13th-Gen CPUs, which we’ll talk about first. As always, you should take vendor-provided data with a grain of salt. We’ve also seen a lot of leaked benchmarks, which help us figure out what Raptor Lake will look like. But, as with all leaked test results, the benchmarks below should be taken with a grain of salt. They are often done with pre-production chips and platforms that haven’t been optimized, so the final results could be much better.

Intel says that the Core i9-13900K is up to 25% faster in some games and up to 5% faster in more than half of the 32 games it tested. Intel’s Core i9-12900K is about 5% faster at gaming than AMD’s new Ryzen 9 7950X, according to our tests. Intel says that most games will be at least 5% faster than the 12900K. This means that the race between chipmakers will be very close.

Intel also talked about what it thought would happen with AMD’s now-old Ryzen 5000 chips, but the company didn’t have the new Ryzen 7000 to test. The percentages at the top of the second slide are compared to AMD’s Ryzen 9 5950X. However, the Ryzen 7 5800X3D is only shown on the chart as a thin line above the 5950X bar.

With its 3D V-Cache technology, the expensive and rare 5800X3D is currently the fastest gaming chip on the market. Still, you should keep in mind that the 5800X3D doesn’t speed up all games and has much worse application performance than chips with similar prices. Raptor Lake looks like it will be faster than the 5800X3D in some games, but it won’t be able to beat the 5800X3D as the fastest gaming card overall. That’s not too surprising, since AMD’s new Ryzen 7000 chip can’t beat its old chip.

Intel is the first company to market with 99th percentile framerates. As a reminder, the 99th percentile framerate is often used in CPU and GPU reviews because it shows how smooth a game is. As you might expect, Intel is ahead of the Ryzen 9 5950X in this case. You will notice that the 5800X3D is not on this slide.

Intel also included a few benchmarks for creating content, which showed big speedups compared to its previous-generation chips and the Ryzen 9 5950X.

The new Ryzen 9 7950X is very good at creating content, but it looks like the 13900K will keep the race close.

Benchmark	Core i9-13900K	Core i9-12900KF	Percent Change
CPU-Z bench 1T	892.2	815.5	+9.4%
CPU-Z bench nT	16,606	11,348	+46.3%
Geekbench 1T	2,133	1,939	+10%
Geekbench nT	23,701	19,304	+22.8%
Cinebench R23 1T	2,206	1,940	+13.7%
Cinebench R23 nT	37,385	26,939	+38.8%
3D Mark Timespy CPU	23,839	20,121	+18.5%
PugetBench Premiere Pro	1,213	1,003	+20.1%

In a wide range of tests, the 13900K was, on average, 10% faster with a single thread and 35% faster with multiple threads than the 12900K. Given that almost all of it comes from the extra eight E-Cores and cache, that’s a pretty good multithreaded gain. Some of the single-threaded benchmarks, like CPU-z and Cinebench, showed even bigger gains, with a whopping 46% and 40% improvement, respectively. Later, when the correct power settings were made, the outlet added power measurements that showed a more accurate power draw of 253W. During a long test, the chip ran at 77C because of this.

Test	Core i9-13900K	Core i9-12900K	AMD 5950X
Cores / Threads	24C/32T	16C/24T	16C/32T
1T score	2,133	1,987	1,686
Multi-thread score	23,701	17,272	16,508

Above, we can see the results of a different Geekbench 5 test, which showed that the Core i9-13900K, which was paired with 32GB of DDR4-6400 memory, could boost to 5.7 GHz. The 13900K was 7% faster in single-thread tests and 37% faster in multi-thread tests than the previous-generation 12900K, according to this test. More importantly, it was 26% faster than the Ryzen 9 5950X in single-thread and 43% faster in multi-thread.

Test	Core i5-13600K	Core i5-12600K	Core i9-12900K	Ryzen 9 5950X
CPU-Z 1T	830	753	803	684
CPU-Z nT	10,032	6,692	10,921	12,078
CB23	1,387	1,886	1,965	1,652
CB23 nT	24,420	17,161	27,287	26,271

There have been tests of the Core i5-13600K. These results show that the Core i5 with four extra e-cores is much faster than the Core i5 from the last generation. In the CPU-z single-threaded and multi-threaded tests, the Core i5-13600K is 8% and 44% faster than the i5-12600K.

Also, the chip beat the Ryzen 9 5950X in both the single-threaded and multi-threaded CPU-z tests, and it was only 7% slower in the threaded Cinebench test. Another listing in the GeekBench database showed similar gains and was just as impressive. Considering that the Core i5 will sell for around $300 and the 5950X will sell for more than $500, these results are very good. In these tests, the Core i5-13600K is about as fast as the current Core i7.

We’ve also recently seen a direct matchup between the Core i9-13900K and the Ryzen 9 7950X. The 7950X won by 9%.

Architecture

Raptor Lake’s p-cores are called “Raptor Cove,” while the E-cores are still called “Gracemont.” Both cores do have a lot more L2 cache, which suggests that the underlying designs need to be changed. The L3 cache on the Raptor Lake chip can now share up to 36 MB, which is 6 MB more than the last generation. But this only means that two more 3MB L3 cache clusters have been added. It does not mean that the per-core cache capacity has grown.

The hybrid architecture puts a 3MB piece of L3 cache (labeled LLC) next to each “block” of cores. The P-cores, which are dark blue, are close to an L3 slice, and the E-cores, which are light blue, come in groups of four and are close to a 3MB slice of L3 cache. All of the cores share these pieces of cache.

The new Raptor Lake 8+ 16 die, which we talked about in the last section, has 16 E-Cores. Those extra E-Cores would mean that the Alder Lake die above would have to be stretched to fit two more light blue clusters of four E-Cores. Two more 3MB L3 cache slices would be added to these two clusters, bringing the total to 36MB, which is the same as the Core i9-13900K.

In an interesting twist, Alder Lake’s cache scheme gave Intel a new way to turn off cache for SKUs with fewer cores. In the past, when Intel wanted to make lower-end SKUs with fewer cores, it turned off whole slices of L3 cache (shown in the image as LLC blocks) along with any cores it turned off. With Alder Lake, Intel stopped turning off an entire slice of L3 and instead turned off some banks in each slice. This reduced the amount of memory each slice could hold (from 3MB to 2MB, for instance). This makes a lot of sense because it keeps stops on the ring bus active and lets us make SKUs with more cache space than we would normally expect based on the number of disabled cores.

Intel has increased the amount of L2 cache that each P-Core and E-Core can hold. Each E-Core now has 2 MB of private L2 cache, which is 60% more than what Alder Lake had (1.25 MB per core). Intel also increased the amount of L2 cache shared by each group of four E-Cores to 4 MB, which is twice as much as Alder Lake’s 2 MB. This means that the L2 cache could be as big as 32MB. As we’ve seen with Intel’s previous chips that got a big increase in L2 capacity, this kind of improvement tends to lead to better performance in multi-threaded workloads, but it is possible that it could lead to higher IPC in some workloads by keeping the cores fed with more data. It should also help clear the ring of some of the traffic that would be there if L3 data were being moved around. This will make it easier to add more users.

Intel’s larger 4MB L2 cache on the e-core clusters uses a new dynamic prefetching algorithm that uses real-time telemetry data and machine learning to adjust the prefetch algorithm based on the type of workload it is running. Intel says that this helps improve performance by up to 16% in an unnamed application and by up to 2% in Adobe Photoshop, After Effects, and Lightroom.

Intel’s L3 cache also uses a different dynamic “Inclusive/non-inclusive” (INI) method, but this one changes on the fly between inclusive and non-inclusive caching schemes. Most caches are hard-coded to work with one of these two policies and can’t be changed. As a reminder, inclusive caching keeps a copy of L2 data in the L3 cache. This usually improves performance in single-threaded work by increasing the hit rate for serial workloads, while non-inclusive caching only keeps part of the L2 cache in the L3 cache to improve performance in threaded work. Intel’s INI can change this policy on the fly by using machine learning and real-time telemetry. This gives you the best of both worlds.

This microcontroller’s new firmware updates use machine learning algorithms to better analyze telemetry data and make changes to the caching scheme every 200 microseconds. In our review, we’ll talk about this tech in more depth.

In the Windows 11 22H2 update, Intel’s Thread Director, which sends threads to the right type of core (p-core or e-core) based on the type of work being done, gets a major makeover. Intel has added machine learning algorithms to the workload classification engine and made it better at handling tasks in the foreground and in the background.

Even though AMD’s Ryzen 7000 supports the extensions, Raptor Lake’s E-Cores still don’t support AVX-512, so Intel will keep the feature turned off. As before, the E-Cores still have AVX2 and VNNI turned on.

Intel’s “Intel 7” process node was made much better by using faster third-generation SuperFin transistors and making the channel mobility higher. Intel says that this gave a performance boost equal to what would normally be called a full node. As you can see above, this allowed Intel to “shift and extend” the voltage/frequency curve to improve performance at both high and low voltages (>50mV less at the same frequency or 200 MHz more at the same voltage).

Soon, there will be a Raptor Lake model that runs at 6 GHz. This is because Intel was able to dial up the peak clock rates by 600 MHz by improving the node characteristics and speed paths in the design. Intel says that Raptor Cove is the fastest core it has ever made, but the Raptor Lake mobile chips will be tuned more for the lower-power end of the V/F curve to make them more efficient.

To take advantage of the faster memory speeds, the fabric needs to be faster. Intel has raised the ring bus frequency to 5 GHz and made it 900 MHz faster during all-core turbo. This fixes a big problem we saw with Alder Lake. Intel says that just the improvements to the ring bus can lead to a 5% increase in frame rates. We’ve also seen good gains when we overclocked this setting on Alder Lake.

Motherboards, Z790, H770, B760, H610

The Raptor Lake chips will utilize the same LGA1700 socket as Alder Lake, which means they have the same socket and pinout. Both Raptor and Alder will be compatible with 600- and 700-series motherboards, allowing for a great deal of versatility for both generations. However, if you install a Raptor Lake processor on a 600-series motherboard, you will lose the PCH PCIe lane configuration enhancements that we’ll describe below. We’ve already seen a profusion of listings for Z790 motherboards, including DDR4-compatible Z790 motherboards from the likes of Asus and MSI, but exact model information is still scarce.

Raptor Lake is compatible with all LGA1700 coolers, so you won’t need a new CPU cooler. Existing 600-series motherboards will need a BIOS update in order to enable Raptor Lake, and some motherboard makers, like Asus and ASRock, have already published the upgrades.

Raptor Lake represents a number of key connection advances. The previous-generation Alder Lake processors enable 16 PCIe 5.0 lanes for a discrete GPU and four PCIe 4.0 lanes for an M.2 SSD from the CPU. On Raptor Lake, the same lanes are still available, but a new connecting mechanism enables enhanced usefulness.

For Raptor Lake, motherboard manufacturers may now divide the CPU’s 16 PCIe 5.0 PCIe lanes into two x8 configurations, providing compatibility for PCIe 5.0 M.2 SSDs. This does imply that the connection to the discrete GPU will be separated into an x8 connection (a switch might be employed here), but the current PCIe 4.0 link from the CPU for an M.2 SSD will remain active, enabling a total of three M.2 SSD ports that connect directly from the CPU.

PCH Connectivity (up to)	700-series (Raptor Lake)	600-series (Alder Lake)
PCIe 4.0	20	12
PCIe 3.0	8	16
USB 3.2 Gen 2×2 (20G)	5	4

The link between the CPU and the chipset (PCH) through x8 DMI 4.0 remains, but the chipset itself has been enhanced. Previously, the PCH supported up to 16 PCIe 3.0 lanes and up to 12 PCIe 4.0 lanes, however Intel has raised the number of PCIe 4.0 lanes to 20 while reducing the number of PCIe 3.0 lanes to eight, hence increasing connectivity. Additionally, Intel raised the maximum number of USB 3.2 Gen 2×2 (20G) connections from four to five.

All additional PCH connection options present on Alder Lake motherboards are unaffected by the Raptor Lake chipset. However, due to the additional features, the Z790, H770, and B760 motherboards will have substantially different allocations. We independently corroborated Intel’s mistakenly disclosed breakdown, which you can see above. The highlighted locations provide a list of the modifications (Z690 = Z790, H670 = H770, B660 = B760).

Hybrid Thread Director

Alder Lake ushered in the hybrid era for x86 desktop PCs, and Intel’s Thread Director, which places threads on the relevant cores, is the glue that holds the architecture together. Raptor Lake will have enhancements to the Thread Director technology, as a consequence of what Intel describes as a design that is upgradeable and customizable and will evolve over time. Windows 11 22H2 includes a big revamp of Intel’s Thread Director. Intel has upgraded the workload categorization engine using machine learning techniques and enhanced the management of foreground and background operations.

As was the case with Alder, Raptor Lake will use both faster and slower cores with varying voltage/frequency optimizations. As a result, in order to unlock the greatest speed and efficiency, the operating system and applications must be aware of the chip layout to guarantee that workloads (threads) fall in the appropriate core depending on the kind of application.

Intel’s Thread Director technology comes into play here. This hardware-based solution provides Windows 11 with additional telemetry data to ensure that threads are scheduled to the P or E cores in an intelligent and optimal manner, while remaining invisible to software.

This technology operates by supplying the Windows 11 operating system with low-level telemetry data obtained from inside the CPU, therefore notifying the scheduler about the status of the core, whether it be heat, power, or otherwise. Existing thread-scheduling approaches continue to operate with the CPUs, but not as effectively. Alder Lake processors are compatible with a regular Windows 10 operating system. While the chips will function, you will be unable to take use of the expanded capabilities of Thread Director, which will have different effects on performance and power consumption depending on instruction type and application usage patterns. In other words, individual results may vary.

Pricing

CPU	Price	Cores / Threads (P+E)	P-Core Base / Boost Clock (GHz)	E-Core Base / Boost Clock (GHz)
Core i9-13900K / KF	$589 (K) – $564 (KF)	24 / 32 (8+16)	3.0 / 5.8	2.2 / 4.3
Ryzen 9 7950X	$699	16 / 32	4.5 / 5.7	–
Core i9-12900K / KF	$589 (K) – $564 (KF)	16 / 24 (8+8)	3.2 / 5.2	2.4 / 3.9
Ryzen 9 7900X	$549	12 / 24	4.7 / 5.6	–
Core i7-13700K / KF	$409 (K) – $384 (KF)	16 / 24 (8+8)	3.4 / 5.4	2.5 / 4.2
Core i7-12700K / KF	$409 (K) – $384 (KF)	12 / 20 (8+4)	3.6 / 5.0	2.7 / 3.8
Ryzen 7 7700X	$399	8 / 16	4.5 / 5.4	–
Ryzen 5 7600X	$299	6 / 12	4.7 / 5.3	–
Core i5-13600K / KF	$319 (K) – $294 (KF)	14 / 20 (6+8)	3.5 / 5.1	2.6 / 3.9
Core i5-12600K / KF	$289 (K) – $264 (KF)	10 / 16 (6+4)	3.7 / 4.9	2.8 / 3.6

Intel has adopted a “no holds barred” pricing strategy for Alder Lake, Core i7, and Core i9 CPUs in an effort to reclaim market share from AMD. For instance, the 13900K is $100 less expensive than AMD’s Ryzen 9 7950X, but it should offer the same or greater performance.

However, Intel has said that costs would increase because to the growing cost of components, supply chain issues, and inflation. Core i5 costs reflect this reality. Core K-series pricing increased by $30, but Core i9 and Core i7 remained unchanged from the previous generation. Both the model with complete features and the model without graphics cost an additional $30 compared to the previous version. This is by far the most popular SKU, and AMD isn’t as competitive in this price bracket, so Intel can increase the price without sacrificing its objective of undercutting AMD at the high end.

However, AMD’s $299 Ryzen 5 7600X is a potent gaming CPU, so it will be intriguing to see how Raptor Lake performs once it comes on our test bench.

The costs for AMD’s Ryzen 7000 processors are comparable to those of the previous generation, with the exception of the Ryzen 9 7950X, which is $100 less expensive than previously. AMD’s latest CPU generation begins at $299, indicating that the firm would continue to prioritize selling its most costly and high-margin technology.

A significant portion of the price differential between Intel’s Raptor Lake and AMD’s Ryzen 7000 will ultimately be attributable to platform expenses. We may see trends similar to those observed with Alder Lake motherboards: DDR5-capable boards will cost more than DDR4-capable boards because the quicker interface demands more costly manufacturing processes and materials. But the availability of DDR4 motherboards will give it a significant edge over Ryzen 7000 CPUs, which only support DDR5 memory. We believe the price of DDR5 will drop significantly by the time both of these systems are available, but the verdict is still out.

AMD’s AM5 motherboard product stack will feature compatibility for PCIe 4.0-only motherboards, but Intel’s stack will not. There will be two pricing categories for AMD motherboards, with costs moving along a distinct axis. It will be interesting to see which technique yields the greatest price differentials between full-scale implementations.