At Intel’s Architecture Day back in 2018, the company talked about design changes it made to future products. Going forward, we were told that Intel would design CPUs in a way that allowed them to be deployed flexibly, rather than strictly locking a given CPU design into a process node. While the company did not say it publicly, the idea that Intel could take a 10nm CPU and reset it to 14nm was already floating as a solution to its problems.
During an AMA for the Rocket Lake platform, Intel revealed that the CPU backport project that created Cypress Cove was completed in the first quarter of 2019. The reason why Intel targeted Ice Lake’s Sunny Cove processor as opposed to the more advanced Willow Cove chip inside Tiger Lake, was that TGL design was not finished yet. Rocket Lake was shipped at the end of Q1 2021, indicating that it took Intel about two years to build the new layout and backdoor of the core.
This timeline provides a window into how chip design cycles work. In Q1 2019, Intel’s top-end chip was the 9900K, and the company still had its own solid stack of performance gains at the top of the CPU market, especially in gaming. Intel greenlit Cypress Cove and began working to support Ice Lake to 14 nm at the same time as the Core i9-9900KS (launched October 2019) and the 10th generation Core i9-10900K (launched in May 2020).
This is how semiconductor production tends to work. AMD is currently shipping Zen 3, completing Zen 4 and working on Zen 5. Intel just launched Rocket Lake with Alder Lake coming later in 2021 and a 7nm sequel supposedly intended for 2023. This means that Intel and AMD both choose design features and goals based on what they think the competitive situation will be 1-2 years later.
The AMA states that an eight-core nozzle represented the largest die + UHD graphics that Intel could produce without clarifying whether this is related to the design of the LGA1200 connector or any other product limitation. The years since Ryzen’s launch have, we believe, illustrated actual differences in how AMD approaches product design versus Intel.
Although not all AM4 motherboards support all Ryzen processors, AMD has shown that they design their platforms for a longer life cycle and better upgrade paths than Intel does, despite the large differences in their respective revenues and resources. If AMD were able to design AM4 to support both a switch from monolithic dies to chiplets with a central I / O die and a doubling of the CPU core numbers, then there is no reason why Intel constantly discovers that its own products run out of ceiling height so easily – unless they are meant to do so. We continue to suspect that the limiting factors on Rocket Lake were thermal and current related. Intel is fully capable of producing larger chips than an eight-core RKL + integrated Xe GPU. To claim that it can not, without reference to what the limiting factor is, sends an error message about Intel’s production capability – unless you think the same company that manages to build a 28-core monolithic Xeon can not manage a 10-core desktop CPU with IGP.
Pat Gelsinger has promised to bring Alder Lake to the desktop only before the end of 2021. If Intel sticks to this schedule, it will be the fastest phasing out of the desktop platform in history. Rocket Lake compares better lower down the pile, but the top piece is competitively fast and non-competitively thirsty in terms of power.
Part of what we see here is an interesting example of how a company can make one set of reasonable decisions in 2019 and find itself hard pressed with the final product two years later. But the fact that Alder Lake is coming so fast after Rocket Lake also speaks to the size and strength of being Intel. Intel has maintained several engineering teams working on multiple chips to the point that it (perhaps) could launch a new CPU replacement CPU less than a year after the launch of the first. This kind of turnaround operation represents its own insurance for a bad launch, and it is not something we see happen very often.
The closest analogy I can think of would be in June and August 2002. In June, AMD launched a 130 nm full-blooded Athlon XP, the so-called “T-wide A.” The chip went hot and did not get AMD much of a clock lift. Two months later, AMD unveiled Thoroughbred B, a new spin of the chip with a 9th metal layer and significantly higher frequencies. T-Broad A processors peaked at 1.8 GHz, but full-blooded B could beat 2.25 GHz, and that put AMD in a stronger position against Intel on its way into the back half of the year. In any case, Rocket Lake occupies a strange place in Intel’s pantheon of desktop CPUs, and perhaps a bit of a lesson in the difficulty of adapting a CPU design to a process node it was not built for.