One particular of the greatest issues facing Intel as we move into 2021 is the degree to which the corporation will depend on 3rd-occasion foundries for its top-edge solutions. More than the earlier yr or so, Intel has declared that it would tap foundry partners for a wider array of solutions, but still left open up which businesses experienced won which small business.
In accordance to Reuters, Intel intends to manufacture DG2 — its forthcoming consumer graphics card — at TSMC, on an improved 7nm node that hasn’t even been named still. TSMC has offered 7nm in 3 flavors — N7, N7P, and N7+. N7P was the authentic N7 node with extra functionality enhancements, when N7+ introduced EUV lithography. The introduction of EUV was a big action all on its personal. TSMC may possibly have constructed a new 7nm node for its personal reasons the corporation once in a while introduces new variants of mature nodes. 7nm is nevertheless near plenty of to the top edge to plausibly be refined in some manner.
Alternately, Intel may possibly have paid out TSMC to employ a specific model of the node that satisfies its personal ambitions. Reuters claims the node will be much more highly developed than Samsung’s 8N, which Nvidia tapped for Ampere.
Intel CEO Bob Swan gave interviews forward of CES 2021, in which he seems to be charting a center floor among Intel yanking all of its current 3rd-occasion semiconductor production back into its personal fabs and going purely fabless. In accordance to Swan, Intel’s target is to maintain maximum overall flexibility in its methods:
[W]e may possibly outsource much more it indicates we may possibly use much more available 3rd-occasion IP, it indicates we may possibly make things for some others, not just, i.e., be a foundry ourselves. And is there a state of affairs exactly where we could be utilizing somebody else’s system engineering in our fabs? That is attainable. The critical is, as the marketplace evolves, how do we leverage the innovation? Not just inside our four partitions, but the innovation happening in the marketplace as a complete, and be quite versatile and adaptable to choose edge of individuals [innovations] together the way.
These reviews are not far too distinct from what Swan has reported in the earlier, and they indicate that Intel will announce a mixed system in which it retains its personal foundries, but clarifies which solutions will be constructed at other businesses on January 21. It does not really make any difference if Intel builds DG2 at TSMC, just like it does not really make any difference that Mobileye proceeds to use TSMC for its personal solutions. When persons assume about Intel, they don’t assume about GPUs or automotive computing. What is really going to push headlines and viewpoints about whatever system Intel announces on January 21 is not exactly where the corporation builds its GPUs, IoT, automotive, or storage solutions. The perceived impression will flip on exactly where future CPUs will be constructed and what foundry shoppers (if any) gain which solutions.
In accordance to 2020 details from IC Insights, TSMC was the second-largest foundry, with the equal of 2.5 million 200mm-equal wafer starts for each month. Intel was calculated at 817K wafer starts for each month. Though this leaves TSMC definitely dwarfing Intel, not all of TSMC’s foundries are even able of building the type of hardware Intel needs. Any attempt by Intel to change its small business to TSMC wholesale would also need a spectacular potential expansion on TSMC’s element.
Glance for much more specifics on Intel’s future chip system to arrive on January 21.