Feature Articles

How much can Intel squeeze out of their 10nm Intel 7 process? Turns out plenty. A look into 13th Gen Core 'Raptor Lake'

By Zachary Chan - 15 Oct 2022

Intel Raptor Lake

Note: This feature was first published on 28 September 2022.

Intel 13th Gen Core ‘Raptor Lake’ series in a nutshell

In the blue corner, Intel’s direct response to AMD’s Ryzen 7000 series is its own 13th Gen Core series of processors, codenamed Raptor Lake. However, unlike AMD’s brand new Zen 4 microarchitecture that’s now based on a 5nm process, Intel’s Raptor Lake is “merely” another refinement of their 10nm process node, also known as Intel 7, and the performance hybrid architecture that debuted with last generation’s Alder Lake chips. Raptor Lake has more of everything compared to Alder Lake - faster P-cores, more E-cores, more L2 and L3 cache, and faster DDR5 support. However, the underlying architecture and technology remains the same.

Now, I used air quotes for merely in the last paragraph because as easy as it is to write off Raptor Lake as Alder Lake+, Intel managed to squeeze so much gains from these tweaks that they claim Raptor Lake’s performance matches or even exceeds the expectations of a microarchitecture generation gap. 

What’s new for Raptor Lake and performance hybrid architecture?

Let’s first take a look at the top level changes that will come to the desktop K-class CPUs before diving into some technical stuff. If there is any thing you take away from this article, it should be these points. Raptor Lake CPUs would feature: 

  • Same number of Performance Cores (P-core) as its predecessors, but with faster boost clocks, up to 5.8GHz for the i9-13900K.
  • Double the Efficiency Cores (E-core) of its Alder Lake counterparts with the i9-13900K having 16 E-cores, for a total processing capability of 24 cores (and 32 threads). Intel will also extend the performance hybrid architecture to the entire i5 lineup for Raptor lake. Expect to see 6P+8E for the i5-13600K and 6P+4E for the rest of the i5s.
  • Expanded 2MB L2 cache per P-core, and 4MB L2 cache per E-core cluster, more than doubling the total L2 cache of its predecessors. The i9-13900K will feature a whopping 32MB of L2 (up from 14MB on the i9-12900K). 
  • Larger L3 smart cache, up to 36MB for the i9-13900K (up from 30MB in the i9-12900K)
  • Faster DDR5-5600 support (up from DDR5-4800 of Alder Lake)

The initial three desktop K SKU 13th Gen Core processors launched.

Here are the full specs of the three K-SKU Raptor Lake chips launched today (well six, if you count KF without integrated graphics).

So, are there anything that’s actually new with the Raptor Lake cores? Yes. Raptor Lake’s P-cores are codenamed Raptor Cove, and are designed with what Intel is calling their 3rd generation SuperFIN transistors, improved speed paths and a new dynamic prefetcher algorithm. This allows the Raptor Cove P-cores to basically be more stable at scaling with voltage, unlocking up to 600MHz additional boost speed over the last generation.

Daniel Rogers, Senior Director, Mobile Product Marketing, presenting some of the Raptor Cove P-core improvements in Raptor Lake. Intel Tech Tour 2022, Israel.

Similarly, the E-cores have gotten higher max boost speeds and smarter prefetcher algorithms as well. The i9-13900K E-cores have a max boost of 4.3GHz (over 3.9GHz on the i9-12900K). However, I do note that the base E-core frequencies are lower on the 13th Gen CPUs as their 12th Gen counterparts. For example, the i9-13900K has a 2.0GHz base compared to 2.2GHz on the i9-12900K. This could very well be a balancing feature due to the additional E-cores on these CPUs.

Lastly, Raptor Lake CPUs will have a faster compute fabric as well (900MHz more than Alder Lake), with a larger L3 cache and again, better dynamic L2/L3 prefetching algorithms. 

On the software side, Intel Thread Director has also been improved to better handle and route processes to the various cores, offering more performance or more efficiency when needed. This comes in addition to better multi-threaded support on Windows 11 as well. 

Intel is also pushing Raptor Lake as a great overclocking platform, and to that, updated the Extreme Tuning Utility for both experienced overclockers with per-core OC, as well as a 1-button Speed Optimizer for those new to overclocking.

Extreme Tuning Utility gets both more extreme, and incredibly easy at the same time.

What can you actually expect in terms of performance?

Intel is claiming up to 15% single-threaded and a massive 41% multi-threaded performance gains for Raptor Lake. Expect up to 24% improvements in gaming and 34% improvement in creator workflow applications.

Marcus Kennedy, GM, Gaming, Creator, and Esports, showing off the gains from a 13th Gen Core i9-13900K at the Intel Tech Tour 2022, Israel.

An interesting dynamic performance-per-watt chart shows that the i9-13900K can match the 19-12900K running at 241W, at just 65W.

In this chart is to be believed, the 13th Gen Core i9-13900K will deliver similar MT performance to the 12th Gen i9-12900K at a fraction of the power.

Like NVIDIA, Intel is starting to use 99th percentile numbers to prove that min-max fps rates aren’t the only thing that matter in performance benchmarking, and that a more consistently high (99%) fps rate will provide a clearer indication of performance and gaming experience across the board.

More people are getting in on 99th percentile numbers to show higher consistent fps.

The most impressive demo that was shown to us was a 13th Gen Core  i9-13900K vs AMD Ryzen 5800X3D in a simultaneous gaming (Rift Breaker) and live-streaming workload, where the Intel machine proved to have higher fps and higher 99th percentile than the AMD-based system. At the end of the benchmark, the surprise reveal was that the i9-13900K was additionally rendering a 4K HDR video in the background, while the AMD system did not. 

Instead of just watching on the big screen. I got to look at the benchmark setup up close.

Across the board, the i9-13900K had higher average fps and 99th percentile fps than the Ryzen 7 5800X3D.

Surprise, surprise. As the benchmark ended, the Intel Raptor Lake system was shown to be running an extra 4K encode in the background.

4K encoding was running throughout the benchmark, a big flex by Intel showing just how much improved the 13th Gen Core's Thread Director is at handling CPU-intensive simultaneous multi-tasking. Of course, having double the E-cores on the i9-13900K would have helped as well.

You could say Intel has had a good comeback run with Alder Lake that looks like it will continue with Raptor Lake. The benchmarks and scenarios played out to us in the charts and demos above look very promising, but of course, these are always one-sided and against the competition’s past-generation best. Until we get our hands on both Intel 13th Gen Core and AMD Ryzen 7000 processors for a head-to-head, we won’t know who’s got the upper hand as yet.

A Blender demo showing how much faster the 13th Gen Core i9-13900K (right) is compared to last year's 12th Gen Core i9-12900K (left).

What’s new on the LGA platform?

There will be a new 700 series chipset that will be launched with Raptor Lake, but from everything we’ve been shown, Raptor Lake will continue to work with existing 600 series motherboards as well, so raw performance of the 13th Gen processors should not be impacted if you use existing Z690 motherboards. The only additional features of the 700 series chipset seem to be additional chipset-based PCIe Gen 4 lanes & USB 3.2 Gen 2×2 ports.

The range of new Intel 700-series chipset motherboards will launch together with 13th Gen Core processors.

Closeup look at the connectivity backplate of an Intel Z790 motherboard.

CPU Roadmap Sneak Peek

With Alder Lake and Raptor Lake bringing back Intel’s Mojo, they’re looking at breaking the last eight year’s cycle of stagnation with process node improvements. Intel was essentially “stuck” at 14nm from 2015 to 2018, and now they’ve been “stuck” again at 10nm for another four years since 2019. While Intel has shown that size doesn’t always matter at squeezing out meaningful performance improvements on an aging process, they’ve committed to deliver 5 node bumps in the next 4 years, moving from Intel 7 to Intel 4 (7nm) with full EUV Lithography for next year’s Meteor Lake, then Intel 3 by H2 2023, followed by Intel 20A (2nm) and brand new RibbonFET transistor technology in 2024 and Intel 18A in 2025.  

Isic Silas, Intel Corporate VP, Client Computing Group presenting at Intel Tech Tour 2022, Israel

Shlomit Weiss, Senior VP and GM, Design Engineering Group presenting Intel's node roadmap plan at the Intel Tech Tour 2022, Israel.

 

Join HWZ's Telegram channel here and catch all the latest tech news!
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.