5 Attachment(s)
Intel Launches Raptor Lake at AMD: 24-core i9-13900K Arrives Oct 20 for $589
Intel shared specs, pricing, benchmarks, and architectural details for its 13th-Gen Raptor Lake lineup today, touting its flagship Core i9-13900K as the 'world's fastest desktop processor.' The lineup includes three new processors with frequencies that jump up to 5.8 GHz, a new record for the mainstream desktop, and more e-cores, giving Intel the core count lead on the mainstream desktop for the first time since AMD's Ryzen arrived back in 2017. Intel also announced that the 6 GHz version of the Raptor Lake is a limited edition model that will come to market later this year. This will be the first 6 GHz desktop PC chip.
Paired with a series of other advances, including bumping up the power consumption on lower-end models and the debut of the new Raptor Cove CPU cores, Intel claims the 13th-Gen Raptor Lake processors deliver up to 15% higher single-threaded performance and 41% more multi-threaded performance than their Alder Lake predecessors. The company also says that gaming performance has improved up to 25% on some titles.
If accurate, that means Raptor Lake will either match or exceed AMD's Ryzen 7000 processors that launched yesterday, with the $699 Ryzen 9 7950X and $299 Ryzen 5 7600X both stealing the performance lead from Intel at their respective price points.
The Raptor Lake K-Series Core i9, i7, and i5 chips come to market on October 20, spanning from the flagship $589 Core-13900K to the $319 Core i5-13600K, but they're just the leading wave: Intel will eventually bring 22 Raptor Lake chips to market for desktop PCs, along with a separate lineup for laptops.
Intel has already announced that Raptor Lake will reach up to 6GHz, but it didn't launch that chip today. Intel confirms that this will be a pricey KS model, but we'll have to wait for more info. Intel's Raptor Lake focuses heavily on performance and now supports faster DDR5-5600 memory while maintaining support for value-option DDR4. Intel also employed a significantly improved 7nm process node and made several architectural advances, including fatter, faster, and smarter caches that significantly improve performance. Additionally, a new thread director baked directly into the new Windows 11 2H22 update will improve thread targeting, and thus improve performance with Raptor's hybrid architecture.
Intel recently announced that it would raise the pricing for its chips due to economic factors. While the company left its Core i9 and i7 pricing the same as we saw with the previous gen, it raised pricing for the Core K-series by $30. Let's take a closer look at the details.
Intel 13th-Gen Raptor Lake Specifications and Pricing
As before, Intel's Raptor Lake comes with a mix of both big performance cores (p-cores) for latency-sensitive work and small efficiency cores (e-cores) for threaded and background applications. The e-cores stick with the same Gracemont architecture as before, but the p-cores move from the Golden Cove design to Raptor Cove. As before, you can also opt for graphics-less F-series models to save some cash on the three Raptor Lake chips.
The 24-core Core i9-13900K has the highest clocks of the new chips, with a 600 MHz improvement bringing it to a 5.8 GHz peak. Intel says it's bringing its Thermal Velocity Boost and Adaptive Boost Technology to the Raptor Lake Core i9 models, so the 13900K will have to meet certain temperature and power conditions to reach 5.8 GHz.
While Intel has raised the peak frequencies for both p-cores and e-cores on all Raptor Lake chips, we see a 200 MHz decline in the base clocks. That shouldn't mean much in practical application and is probably done to manage the TDP rating. Speaking of which, the 13900K has the same 125W Base Turbo Power (BTP) as before, but the Maximum Turbo Power (MTP) has increased by 12W and is now 253W.
The Core i9-13900K also has eight more e-cores than its predecessor, delivering much more threaded horsepower. The e-cores also come with a 400 MHz increase to the peak boost, bringing them to 4.3 GHz. Intel says that even though the e-cores use the same Gracemont architecture, changes to the caching policy and a few other fine-grained optimizations (more below) now yield the same IPC and frequency as the original 14nm Skylake cores, but at lower power. Notably, the e-cores give Intel the core count lead on the mainstream desktop PC, as AMD's Ryzen tops out at 16 cores with the 7950X.
The 24-core $589 Core i9-13900K will square off with AMD's 16-core $699 Ryzen 9 7950X, so Intel could have a pricing advantage if it can match or beat it.
The $409 Core i7-13700K also gets a 400 MHz p-core frequency bump to 5.4 GHz, four more e-cores for a total of eight, and a 400 MHz e-core boost to 4.2 GHz.
Surprisingly, Intel has increased the MTP for this chip to 253W, a 63W increase over the prior gen. The 13700K contends with the $399 Ryzen 7 7700X.
Meanwhile, the $319 Core i3-13600K is the lone chip to receive a price increase; both the full-fledged model and the graphics-less 13600KF are $30 more expensive than the prior gen. This is by far the highest-volume SKU and slots into a pricing tier where AMD isn't as competitive, so Intel's increase here will generate the most cash while not threatening its strategy of undercutting AMD on the high end. However, AMD's $299 Ryzen 5 7600X is a potent gaming chip, so we'll have to see how the performance story plays out when Raptor Lake hits our bench table.
The 13600K receives four more e-cores than the prior-gen, 200 MHz higher p-core clocks to reach 5.1 GHz, and a 300 MHz e-core improvement yields a 3.9 GHz boost.
The 13600K also has a 31W MTP increase to 181W, so Intel is raising power limits here, too.
Some of the improvements apply to all three chips. For example, Intel increased the L2 cache from 1.25MB to 2MB for each p-core, and doubled the amount of L2 cache for each cluster of e-cores to 4MB. The amount of L3 cache has also increased due to adding more e-core clusters, each of which has an adjacent L3 cache slice as part of the design. That leads to cache capacity increases for all K-series Raptor Lake chips.
Intel has increased its DDR5 memory support up to 5600 MT/s if you use one DIMM per channel (1DPC), a big increase over the prior 4800 MT/s speed with Alder Lake. Just as importantly, Intel increased 2DPC speeds up to 4400 MT/s, an improvement over the previous-gen 3600 MT/s. Intel will also continue to support DDR4 memory, which it predicts will coexist in the market with DDR5 until the end of 2024. This approach ensures a value option for Intel platforms, unlike AMD's all-in approach with DDR5.
Raptor Lake and Alder Lake will be compatible with new 700-series and existing 600-series motherboards. As usual, Intel will bring its Z-series (Z790, in this case) motherboards to market first, with the value-centric B- and H-series coming when it launches the rest of the Raptor Lake lineup.
The 700-series motherboards bring minor improvements over the 600-series, but Intel has increased the number of PCIe 4.0 lanes hanging off the chipset to 20, an increase of eight additional lanes. Intel also added support for one more USB 3.2 Gen 2x2 200Gps port, bringing the total to five. As before, the Raptor Lake chip itself supports 16 lanes of PCIe 5.0 and four lanes of PCIe 4.0 for a storage device.
Intel shared its own performance benchmarks, but as with all vendor-provided benchmarks, take them with some salt and wait for reviews. We've included the test notes at the end of the article.
Intel claims that the Core i9-13900K is 5% faster than the 12900K in more than half of the 32 games it tested, and up to 25% faster in some. In our testing, AMD's recently released Ryzen 9 7950X is roughly 5% faster in gaming than Intel's Core i9-12900K. Given that Intel claims most titles will land at 5% faster than the 12900K or above, we will surely see a very close race between the chipmakers.
Intel also shared projections against AMD's now previous-gen Ryzen 5000 chips, but the company didn't have access to the new Ryzen 7000 for testing. The percentages at the top of the second slide are versus AMD's Ryzen 9 5950X, but you'll notice that the Ryzen 7 5800X3D is only represented as a thin line above the 5950X bar on the chart. The exotic and pricey 5800X3D is currently the fastest gaming chip on the market, courtesy of its 3D V-Cache tech. Still, you do have to remember that the 5800X3D doesn't accelerate all games and comes with severe tradeoffs in application performance compared to similarly-priced chips.
It appears that Raptor Lake will beat the 5800X3D in some titles, but it won't be able to unseat the 5800X3D from the title of being the fastest for gaming overall. That's not totally unexpected -- even AMD's new Ryzen 7000 can't beat its previous-gen chip.
Intel is also now adding 99th percentile framerates to its marketing repertoire in the 'leadership gaming frame consistency' slide, a first. As a reminder, 99th percentile framerates are popular in CPU and GPU reviews, like ours, because they are a good indicator of the smoothness of a game. So naturally, Intel claims a lead here over the Ryzen 9 5950X. You'll notice the company didn't include the 5800X3D in this slide.
Intel also included a few content creation benchmarks, showing big speedups relative to its previous-gen chips and the Ryzen 9 5950X. The new Ryzen 9 7950X is incredibly powerful in content creation apps, but it looks like we'll see another tight race when the 13900K arrives.
Intel 13th-Gen Raptor Lake Architecture
Intel significantly improved its 'Intel 7' process node by employing faster third-gen SuperFin transistors and increasing the channel mobility (Intel declined to provide more specifics on the latter). Intel claims this yielded what would typically be considered a full node's worth of performance improvement. As you can see above, this allowed Intel to 'shift and extend' the voltage/frequency curve to deliver better performance at both high and low voltages (>50mV reduction at the same frequency, or a 200 MHz increase at the same voltage).
We'll soon see a 6 GHz Raptor Lake model because the improved node characteristics paired with improved speed paths within the design (likely using leakier, faster transistors in key paths) allowed Intel to dial up the peak clock rates by 600 MHz. Intel says that Raptor Cove is the fastest core it's ever built, but that the Raptor Lake mobile chips will be tuned more for the lower-power end of the V/F curve to improve efficiency.
Intel's larger 4MB L2 cache on the e-core clusters also employs a new dynamic prefetching algorithm that uses real-time telemetry data and machine learning to dynamically adjust the prefetch algorithm based on the type of workload it is executing. Intel says this helps improve performance in some types of applications by ~2%, like Adobe Photoshop, After Effects, and Lightroom, and in an unspecified application by up to 16%.
Intel also employs a different dynamic 'Inclusive/non-inclusive' (INI) approach for its L3 cache, but this adjusts between inclusive and non-inclusive caching schemes on the fly. Caches are typically hard-coded to run with either of these policies and can't be changed.
As a reminder, inclusive caching keeps a copy of L2 data in the L3 cache and typically improves performance in single-threaded work by providing a faster hit rate for serial workloads, while non-inclusive caching only keeps part of the L2 cache in L3 to improve performance in threaded workloads. Intel's INI uses machine learning and real-time telemetry to make changes to this policy on the fly, thus providing the best of both worlds.
We spoke with Intel Fellow and Chief CPU Architect, Adi Yoaz, at Intel's Israel event about the new scheme, and he tells us that this approach is the result of the microcontroller that Intel originally designed into each Alder Lake core, but hasn't yet fully used. New firmware updates to this microcontroller employ machine learning algorithms that better analyze telemetry data to make adjustments to the caching scheme at a 200-microsecond granularity. We'll share more details about this tech in our review.
Speaking of telemetry, Intel's Thread Director, which steers threads to the right type of core (p- or e-core) based on the type of workload, gets a significant overhaul in the Windows 11 22H2 update. Intel has updated the workload classification engine with machine learning algorithms and also improved the handling of foreground and background tasks.
