We recently reported that Intel is preparing special laptop CPUs with BLLC, its equivalent of AMD’s 3D V-Cache. CPUs with BLLC will feature up to 144 MB of L3 cache promising significant boost in games. BLLC will come to laptops in the Razor Lake lineup, but that will not be the only interesting thing about them. A leaker has shared further details, suggesting that this could be a surprisingly powerful generation.
Intel is now moving toward the release of the Nova Lake processor CPU generation, using TSMC’s 2nm process and new P-Core and E-Core architectures. It should be released late this year or early next year as the fourth-generation Core Ultra—either Core Ultra 400 or possibly 4000 (some leaks are now saying that Intel could start using four-digit model numbers again—for example, the 28-core model could reportedly be called Core Ultra 9 4950K). However, the new information we will discuss here concerns the generation that comes after that—codenamed Razor Lake, which should therefore be introduced as fifth-generation Core Ultra (Core Ultra 500 or 5000). We have already had some leaked details about it here:
The leaker using the nickname Jaykihn has now stated that the Razor Lake processor generation will continue to use TSMC’s 2nm process for most of its variants (although Intel may base some models on its own manufacturing processes—see the link above). Meanwhile, articles have appeared claiming that Razor Lake could even use the high-performance N2X process aimed at high clock speeds—but beware, this was the result of a mistake. According to Jaykihn, it will actually use the second improved version of the more conventional N2P process (N2P V2, which could involve the use of UHP libraries). This technology should still be somewhat better than the process that will be used for Nova Lake, but the difference will be smaller than in the case of N2X.
New P-Cores
The most important thing, however, is that according to Jaykihn, Razor Lake will not merely be a minor evolutionary improvement over Nova Lake, as we had previously thought—even though Nova Lake itself will already be a major architectural upgrade. Normally, we would therefore expect an “intermediate generation” to follow, featuring more or less the same cores or only slightly modified versions—such intermediate generations included processors such as Tiger Lake, Meteor Lake, or most recently Panther Lake. However, Razor Lake is reportedly not such a case. The Griffin Cove big cores (P-Cores) that these processors will contain come with, according to Jaykihn, a new microarchitecture, corresponding to a “tock” in the older Intel terminology.
The Griffin Cove core will therefore not be the same as Coyote Cove in Nova Lake, meaning it should involve extensive changes. This could significantly improve performance per 1 MHz (the so-called “IPC”) and, consequently, raise single-threaded performance. These processors could therefore represent a more significant innovation and a more important step-up in terms of Intel’s competitiveness than previously expected. While the P-Core is supposed to be new, the E-Core is not, and will apparently retain the Arctic Wolf architecture from Nova Lake.
A New Architecture Every Generation?
Intel could potentially move to a much more aggressive CPU inovation pace in the future. Until now, it has essentially been the case that a new P-Core appeared only in every second generation (10th- and 12th-generation Core, Core Ultra 200 and 400, while the generations in between more or less reused the cores from the previous generation). However, Nova Lake (4th-generation Core Ultra) will have new P-Core and E-Core architectures, then Razor Lake (5th generation) will have a new P-Core, and the following Titan Lake processors (6th-generation Core Ultra?) apparently introduces new cores as well—the Copper Shark architecture, which is reportedly the first core from the new Unified Core line. Even the next generation after that, called Hammer Lake (perhaps 7th-generation Core Ultra), is supposed to have another new core, the unified Thunder Hawk architecture, which is reportedly also supposed to bring back Hyper-Threading.
If it has long seemed that Intel had fallen asleep at the wheel and fallen behind in processor development, this roadmap of the next four generations could potentially turn the tables again. If each of these generations significantly increased performance, Intel could suddenly put AMD under serious pressure while also competing more convincingly with Arm processors from companies such as Apple. AMD is now planning to release new cores at two-year intervals (Zen 7 in 2028, Zen 8 in 2030), so if Intel started ti move faster, it could leapfrog AMD, leaving the competitor in its proverbial rear-view mirror.
The question, however, is whether Intel will manage to maintain such a rapid development cycle in reality. The effort to introduce a new core in every generation could, for example, result in longer intervals between the releases of these generations (Nova Lake, Razor Lake, and so on) than the standard one year. In that case, the effect could be less impressive. Who knows, perhaps we will once again see Intel fail to step-up the pace after all and once again having to insert “filler” refresh chip families such as Raptor Lake between those new generations. Plans and promises are one thing; only time will tell whether the new CEO will manage to execute such plans to catch up in practice. A similar plan by his predecessor Patrick Gelsinger, who was supposed to deliver five new manufacturing processes in four years and catch up with TSMC by 2025, ultimately did not work out particularly well.
English translation and edit by Jozef Dudáš
⠀
