You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
AMDs biggest pro is binning and cost and IP Reuse
Intel's mostly is able to mix Match different IP so more combination is possible but not cost more chiplets cost more more so on different node's. Intel kind of botched their chiplet implementation as well
Which is sad given you can argue the Pentium Pro in 1995 was a chiplet (Compute + Cache), or Westmere in 2009 had a separate iGPU on package. Other products inbetween too.
Arrow Lake's chiplet issues are proof of how broken Intel's engineering is right now. Intel's 3rd gen of modern chiplet CPU designs -- with Meteor Lake before it, and Lakefield before that.. With a ~ full node advantage over AMD they weren't clearly ahead of Zen 5%.
Which is sad given you can argue the Pentium Pro in 1995 was a chiplet (Compute + Cache), or Westmere in 2009 had a separate iGPU on package. Other products inbetween too.
Arrow Lake's chiplet issues are proof of how broken Intel's engineering is right now. Intel's 3rd gen of modern chiplet CPU designs -- with Meteor Lake before it, and Lakefield before that.. With a ~ full node advantage over AMD they weren't clearly ahead of Zen 5%.
That being said, it's still in the middle, it should be good from now on.
There is only an improvement
It's not that AMD's chiplet method was excellent from the beginning, and we've made improvements.