Home
Creators
Search Recent Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

Itanium: Intel’s Great Successor (Patreon)

Published:
2026-03-01 22:00:07
Imported:
2026-03

Downloads

Content

Files

Itanium: Intel’s Great Successor

Links: - Patreon (Support the channel directly!): https://www.patreon.com/Asianometry - X: https://twitter.com/asianometry - Newsletter & Podcast (available through Stratechery Plus): https://asianometry.passport.online/ - LinkedIn (feel free to connect): www.linkedin.com/in/jon-y-asianometry

Comments

Tanj

The VW with wider tires and better engine was very successful. Indeed, Porsche eventually took control of VW - ironically that snark also paralleled exactly the AMD64 controlling Intel.

wleslie

Your Magnum Opus!

Joe Marler

As always, a great video on a deep topic. Thanks! When I was at Microsoft in the late 1990s, we were told: "Merced is THE way forward, at least on servers." But my friends on performance teams kept talking about the ever-improving Xeon capability. The sensitivity of VLIW to certain workloads was well known at Multiflow, and several key people from Multiflow went to HP and Intel for Itanium. So they knew of the workflow-dependent issues. Multiflow engineer Geoffrey Lowney wrote a paper about those issues in 1992, "The Multiflow Trace Scheduling Compiler." On Itanium, there was supposedly an architectural plan to mitigate VLIW weaknesses, and significant simulation studies suggested it was achievable. However, when much of the early VLIW and Itanium plans and modeling were done, that was before the major successes of wide, out-of-order superscalar CPUs. Also, Intel was a CISC CPU company, and those early 1990s OoO successes were RISC-based, so there was little data on the feasibility of CISC OoO superscalar designs. In The Pentium Chronicles, the senior architect of the P6 mentioned the lack of data on CISC OoO superscalar methods in the early 1990s. Also, the technique of "cracking" CISC instructions to micro-ops for parallel scheduling (in the NexGen Nx586 and P6) was not proven until Itanium was well underway. If you told an Intel engineer in 1990 that within his lifetime there would be ubiquitous desktop CPUs that were 10-wide out-of-order superscalar with 500+ instructions in flight, he probably wouldn't have believed it. So to some degree it appears that Itanium was done in by the unexpected success of wide-issue, out-of-order superscalar CPUs.