Intel reportedly planned to outsource production of its Meteor Lake tGPU chipset to TSMC, with mass production scheduled for the second half of 2022, before being delayed to the first half of 2023 due to design and verification issues. Now, Intel is said to have delayed mass production to the end of 2023, virtually cancelling the 3nm chip production capacity that it had booked with TSMC for most of next year.
Errr, All Intel's fault? TSMC isn't at high volume manufacturing (HVM) for N3 even if Intel wanted to be at mass productions. TSMC set a target of 2H22 for HVM , but they haven't hit it yet. Earlier they said they were going to recognize revenue from N3 until Q1 2023. That first shift to 1H 2023 probably has a TSMC component to it also.
Intel has design defect issues with the Sapphire Rapids so it is doubtful that Meteor Lake is bug free if there are any shared baseline features being shared across both ( e.g. , the tile-to-tile interconnect implementation , PCI-e/CXL features , etc. ). Or Intel has some major rework to do just on the graphics side.
This a bit of a the dual edged sword with tiles design across multiple process nodes. If one tile gets sidetracked than the whole package is delayed ( somewhat like mostly finished automobiles and trucks waiting on chips for a backup camera. )
The odd duck here is that why there isn't some discrete GPU part here to soak up at least some of that N3 capacity. If flip from making iGPU tiles to full dGPU would be an easy way to soak up wafers. Pointing at Meteor Lake and the CPU package may be off target. Intel original plan was to solely make iGPU tiles for all of 2023? Really? [ Without good drivers Intel may not want to make more expensive hardware faster. But that isn't a chip design and verification problem being pointed to by this article. ]
If Intel waits until end of 2023 they could possibly pick up TSMC N3E and just skip N3 all together. Which also might be a motivation. It is also odd they would be waiting untll the buzzsaw of close to max A17 production to jump into mass production.
It is not clear if the disruption at TSMC caused by Intel will impact Apple's 3nm chip production volume or schedule. The first 3nm chip from Apple is
rumored to be the M2 Pro, debuting in
the 14-inch MacBook Pro, 16-inch MacBook Pro, and a high-end Mac mini model.
If Intel has wafer starts they can't use they can certainly give/sell them to Apple. The questions more so is whether Apple would want to ramp any product faster. The M1 Pro/Max/Ultra Mac models all experienced launch bubble delays. If the M-series SoC played any role in that slowness ( although Apple always undercounts the launch bubble to create that 'scarcity' fanboy driver and minimize inventory costs. )
However, Apple also has its own multiple tile/chiplet dependency problem. IF they crank up N3 tiles production, but can't get more interposers, then the fancy packages can't be made any faster. If they can get all the parts this might bring the Mac Pro forward in time in terms of launching (e.g, earlier into the Spring). Apple couldn't soak up Intel entire wafer allotment, but some. There is also likely some other players waiting that Intel had outbid for wafer starts. It is just going to be harder for TSMC to count what the production will be because will have to aggregate more customers to find total demand.
Apple's AR/VR googles could be on N3 also. Could that ramp faster? Maybe.
I'm not sure why this is being partially spun as some "problem" for Apple. A huge block of wafers coming free means they can roll their stuff out faster if can keep up with the rest of the parts to the systems.
