This isn't the Intel Core i9 11900K we were supposed to have. The very heart of this processor is fundamentally different from the one that was meant to have beat out a rhythm for Intel's 11th Gen Core CPUs, and that, in the end, is why I doubt anyone will feel comfortable recommending this nominally $539 CPU. I certainly don't.
The Rocket Lake range of CPUs was in trouble from its inception, and that rings especially true for the Core i9 11900K as the flagship chip of this new generation of desktop processors. From the very first decision to create this set of CPUs, Rocket Lake was always on the back foot, because Intel's failure is the only reason for its existence.
Tiger Lake. That's what we should be writing about when it comes to this 11th Gen series. That's what we're doing on the mobile side of Intel's business, even the high-performance, gaming-focused chips that are coming out later in the year. But, instead of that continuation of the 10nm production process into desktop form, Rocket Lake is another 14nm chip, but one that is trying to take the previous 10nm core design and fit it into 14nm silicon.
It's not quite a square peg in a round LGA socket, but it's the reason why Rocket Lake is such a jarring CPU generation at the top end.
What's different about the Intel Core i9 11900K
The first thing to note is that Rocket Lake isn't Tiger Lake's 10nm core back-ported into a 14nm CPU package. The Cypress Cove core architecture that powers the Core i9 11900K is essentially the Sunny Cove design squeezed into a chip that can fit in the same socket as the old Comet Lake Core i9 10900K. That was the 10nm design used in the Ice Lake generation of laptop CPUs, while Tiger Lake uses a further refined Willow Cove core, one with fresh transistors optimisations, a redesigned cache system, and higher IPC.
The second thing to note is why a back-port was needed in the first place, and this is something for which I still haven't got a satisfactory answer out of Intel.
The basic situation is that Intel's shift to broader 10nm manufacturing has been significantly impacted by production issues, first around the initial Cannon Lake chips—which barely made it out of the fabs—and constrained by further yield problems with the subsequent Ice Lake design. Neither CPU architecture was then capable of pushing the high frequencies required of modern desktop processors and so those remained stuck on the 14nm node, and on modest iterations of the original Skylake core design from 2015.
But Intel couldn't spin out another version of the 14nm Skylake architecture for its early 2021 chip release if it wanted to catch up to AMD's excellent Ryzen 5000-series CPUs. The latest Zen 3 CPUs have either caught up to, or surpassed, Intel's comparative processors when it comes to gaming performance, and there's really nothing more Intel can squeeze out of the geriatric Skylake core to push ahead.
So the decision was made to bring the Sunny Cove design into one final 14nm CPU range, thus Rocket Lake was born. This would have likely been made while the Willow Cove core for Tiger Lake was being finalised and so the latest 10nm design was probably never really considered for the 11th Gen desktop architecture.
This wouldn't have been a decision Intel took lightly; there will always be sacrifices engineers have to make in order to retool such a core for 14nm production. The most obvious of which concerns space. The problem is that, while a 10nm core can be smaller than a 14nm design if it's fundamentally the same, just on a smaller node, Sunny Cove, and by extension Cypress Cove, are different. Intel originally used the