At Intel’s Architecture Day in 2018, the company talked about the design changes it would make to future products. Going further, we were told, Intel would design CPUs in a way that would allow them to be deployed flexibly, rather than strictly restricting a particular CPU design to a process node. Although the company did not say so publicly, the idea that Intel could take a 10 nm CPU and move back to 14 nm was already being launched 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. Intel’s reason for targeting the Ice Lake Sunny Cove processor as opposed to the Willow chip The most advanced cove within Tiger Lake is that the design of the TGL has not yet been finalized. Rocket Lake was launched at the end of the first quarter of 2021, indicating that it took Intel about two years to build the new floor plan and backport the core.
This timeline provides a window into how chip design cycles work. In the first quarter of 2019, Intel’s top-of-the-line chip was the 9900K and the company still had its own solid stack of performance wins at the top of the CPU market, especially in games. Intel gave the green light to Cypress Cove and started working to bring Ice Lake back to 14 nm, while simultaneously preparing the Core i9-9900KS (released in October 2019) and the 10th generation Core i9-10900K (launched in May 2020 ).
This is Intel’s roadmap since its previous architecture day in 2020. Rocket Lake is not even shown in this document, which focuses on advances in Intel’s core technology.
This is how semiconductor manufacturing tends to work. AMD is currently launching Zen 3, finishing Zen 4 and working on Zen 5. Intel has just launched Rocket Lake with Alder Lake arriving in late 2021 and a 7 nm successor supposedly planned for 2023. This means that Intel and AMD are choosing design features and targets based on what they think will be the competitive situation 1-2 years later.
AMA claims that an eight-core array represented the largest UHD array + graphics that Intel could manufacture without clarifying whether this is related to the design of the LGA1200 socket or some other product limitation. The years since Ryzen’s launch, we think, have illustrated factual differences in how AMD approaches product design compared to Intel.
While not every AM4 motherboard supports all Ryzen CPUs, AMD has demonstrated that it designs its platforms for a longer lifecycle and better upgrade paths than Intel, despite the huge differences in their respective revenues and resources. If AMD were able to design the AM4 to support a shift from monolithic arrays to chips with a central I / O array and a doubling of the CPU core count, so there’s no reason for Intel to perpetually find out that its own products run out of space so easily – unless it’s planned. We continue to suspect that the limiting factors in Lake Rocket were thermal and related to energy. Intel is absolutely capable of making chips larger than an integrated eight-core RKL + Xe GPU. Claiming that you can’t, without reference to what the limiting factor is, sends an erroneous message about Intel’s manufacturing capability – unless you think the same company that can build a 28-core monolithic Xeon cannot manage a Xeon from 10 desktop CPU cores with an IGP.
Pat Gelsinger has pledged to bring Alder Lake to the desktop before the end of 2021. If Intel meets this schedule, it will be phasing out the fastest desktop platform in history. Rocket Lake compares best at the bottom of the stack, but the top chip is competitively fast and not competitively thirsty when it comes to power.
Part of what we see here is an interesting example of how a company can make a seemingly reasonable set of decisions in 2019 and find itself under pressure with the final product two years later. But the fact that Alder Lake is arriving so quickly after Rocket Lake also shows the size and strength of being Intel. Intel has supported several engineering teams working on multiple chips to the point that they could (perhaps) launch a new replacement desktop CPU less than a year after the first was launched. This kind of turnaround represents your own insurance for a bad launch, and it is not something that we see happening very often.
The closest analogy I can imagine would be in June and August 2002. In June, AMD launched a 130nm Thoroughbred core Athlon XP, the so-called “T-bred A”. The chip heated up and did not increase AMD’s clock much. Two months later, AMD unveiled the Thoroughbred B, a new chip spin with a ninth layer of metal and significantly higher frequencies. T-Bred A CPUs peaked at 1.8 GHz, but Thoroughbred B could reach 2.25 GHz, and placed AMD in a stronger position against Intel in the second half of the year. In any case, Rocket Lake occupies a unique place in Intel’s pantheon of desktop CPUs and, perhaps, a bit of a practical lesson in the difficulties of adapting a CPU design to a process node for which it was not built.
Now read: