Discussion about this post

User's avatar
Peter W's avatar

First, TSMC's pricing depends on the node; anything "N3" is amongst their currently most expensive. In contrast, AMD has their chips fabbed on a slightly older, but also significantly cheaper node (N4/5). Second, TSMC got into a huff over some of Pat Gelsinger's statements about whose upcoming nodes is better and refused to give Intel their customary discount of about 30%. Third, the most expensive fabbing node ( apart from one that basically isn't working at all) is one that sits idle. Intel decided to forgo "Intel 20" and put most of their efforts into Intel 18. That was a gamble, we'll see if it pays off soon.

Lastly, do you have a source for that statement that Intel lost a lot of money on fabbing the Meteor Lake CPU tile in Intel 4? Not doubting you, would just like to read it for myself - Thanks!

Expand full comment
mzocyteae's avatar

So ARL uses d5-8000 to marginally win over RPL d4-3600...

This is almost certainly a regression in spec cpu.

Expand full comment
7 more comments...

No posts