As part of today’s announcements, during Intel’s questions and answers after the prepared comments, CEO Pat Gelsinger explained how Intel will revive its fortune in terms of its leading computing products. One of Gelsinger’s mantras seems to be that undisputed management products provide undisputed management margins for these products, and for Intel to be able to perform, it must go back to its old days.
Previously, through the 1990s, 2000s and into the 2010s, Intel’s production philosophy was known as ‘Tick-Tock’. This means that the foremost computing hardware for each product generation was either a Tick (process node enhancement) or a Tock (microarchitecture enhancement). Each generation would switch between the two so that Intel could take advantage of a familiar design on a new process node, or use a mature node to enable a new performance-focused design. This policy was canceled when delays on Intel̵
Today, CEO Pat Gelsinger stated that the core of Intel must re-establish the Tick-Tock model that enabled repeated leadership in the CPU ecosystem, encouraged by a healthy CPU roadmap. Part of this is to restore discipline in Intel’s ranks to continuously provide both microarchitecture updates and process node updates at a normal expected cadence. Pat stated as part of the conversation that Intel will look towards a confirmed annual improvement of the process nodes, and as a result there may be many ticks in the future, with a push for more tocks as well.
On top of this comment, Pat Gelsinger also stated that Intel’s CPU roadmap is already embedded through 2021, 2022 and 2023. The company is thus looking forward to 2024/2025 for “undisputed CPU management performance”, which traditionally means the fastest single-threaded processor and workload with several threads. This is certainly a commendable goal, but Intel will also have to adapt to a changing landscape of chiplet processor design (coming in 2023), improve accelerators on death (GNA already present), and also what it means to have management performance – in modern time does not mean much performance if you also push a lot of watts. Intel stated that the 7nm process is now comfortably on track to deliver Meteor Lake, a client CPU using tiles / chips, in 2023, but we are probably looking at a 7nm variant or even external processes for a 2024 / 2025 product. Intel has also stated that they want to assess the core of their foremost calculation on external foundry processes, although it can be argued that this does not explicitly say ‘CPU’.
It is also worth noting that Intel / Gelsinger does not call its disaggregated silicon as ‘chiplets’, and prefers to use the term ’tiles’. This is because Intel’s tiles form long wires across 3D packaging technologies such as EMIB and Foveros, compared to package-based multidys interconnection that requires buffers as well as control material. Tiles by this definition are more expensive to implement than chips, and have extra thermal considerations by having highly driven silicon close together, so it will be interesting to see how Intel balances these new packaging technologies with the more cost-sensitive elements in the portfolio, such as client processors.
It has been known that Intel’s microarchitecture team has not been inactive and waited for 10 nm to come through the tube, with a number of designs ready and waiting to follow when the process node technology matures. Luckily, if Intel can get headwinds with 7 nm, when 2024 rolls around, it can all come thick and fast.