TSMC has had better low power performance but the difference isn’t massive. I suspect the main reason is a) 18A is not ready for mass production until too late and b) limited 20A capacity coming online next year is needed for other products (remember intel has to ship absolute shitload of chips). So why not do it on TSMC if that is possible and the alternative is to delay the products to wait for capacity?
Edit: i’m not sure if 20A was again a more limited early version designed primarily for intel desktop products and 18A was supposed to be the long term version of the node with wider array of capabilities?
20A is essentially an early, incomplete release of 18A that can only be used on for the compute tiles. LNL has iGPU, NPU, and x86 cores all on the same tile, so 20A can’t be used.
Intel 3 and 4 weren’t designed with GPUs in mind either.
TSMC has had better low power performance but the difference isn’t massive. I suspect the main reason is a) 18A is not ready for mass production until too late and b) limited 20A capacity coming online next year is needed for other products (remember intel has to ship absolute shitload of chips). So why not do it on TSMC if that is possible and the alternative is to delay the products to wait for capacity?
Edit: i’m not sure if 20A was again a more limited early version designed primarily for intel desktop products and 18A was supposed to be the long term version of the node with wider array of capabilities?
20A is essentially an early, incomplete release of 18A that can only be used on for the compute tiles. LNL has iGPU, NPU, and x86 cores all on the same tile, so 20A can’t be used.
Intel 3 and 4 weren’t designed with GPUs in mind either.