As part of today’s announcements, during Intel’s question and answer session after prepared comments, CEO Pat Gelsinger explained how Intel will revive its fortunes when it comes to its cutting-edge computing products. One of Gelsinger’s mantras seems to be that unquestionable leadership products bring unquestionable leadership margins to these products, and for Intel to execute, it needs to return to its old days.
In the past, during the 1990s, 2000 and 2010, Intel’s manufacturing philosophy was known as ‘Tick-Tock’. This means that, for each product generation, the leading computing hardware was a Tick (process node enhancement) or a Tock (microarchitecture enhancement). Each generation would alternate between the two, allowing Intel to take advantage of a familiar design on a new process node or using a mature node to enable a new performance-focused design. This policy was destroyed when Intel’s 10nm delays forced Intel to adopt a model more like the Tick-Tock-Optimization-Optimization-Optimization model.
Today, CEO Pat Gelsinger said that at Intel’s core it must re-establish the Tick-Tock model that has enabled repeated leadership in the CPU ecosystem, driven by a healthy CPU roadmap. Part of this is restoring discipline within Intel’s ranks to continuously provide microarchitecture updates and process node updates at an expected regular rate. Pat said as part of the call that Intel will seek a confirmed annual improvement of the process node and, as a result, there may be many Ticks in the future, with an impetus for more Tocks as well.
In addition to this comment, Pat Gelsinger also stated that Intel’s CPU roadmaps are ready by 2021, 2022 and 2023. The company is therefore looking to 2024/2025 for “unquestionable CPU leadership performance”, which traditionally means the faster processor for single-thread and multi-thread workloads. This is certainly a laudable goal, however, Intel will also have to adapt to a changing scenario of chip processor designs (coming in 2023), improving on-die accelerators (GNA already present) and also what it means having a leadership performance – in the modern era, leadership performance does not mean much if you are also investing in many Watts. Intel said its 7nm process is now comfortably on track to deliver Meteor Lake, a client CPU using tiles / chips, in 2023, however, we are probably looking for a 7nm variant or even external processes for a 2024 product / 2025. Intel also said that it is looking to consider the core of its cutting edge computing in external casting processes, although one might argue that it does not explicitly say ‘CPU’.
It is also important to note that Intel / Gelsinger is not calling its disaggregated silicon ‘chips’ and prefers to use the term ’tiles’. This is because Intel’s tiles are equivalent to long cables in 3D packaging technologies like EMIB and Foveros, compared to the encapsulation-based multi-die interconnection that requires buffers as well as control fabric. Tiles, by this definition, are more expensive to implement than chips and have additional thermal considerations because they have high-power silicon nearby, so it will be interesting to see how Intel balances these new packaging technologies with the most cost-sensitive elements in its portfolio. , as client processors.
It is known that Intel’s microarchitecture teams have not been idle waiting for 10nm to pass through the pipe, with a series of designs ready and waiting to arrive when process node technology matures. With any luck, if Intel gets a headwind with 7nm, when it gets to 2024, everything can come dense and fast.