I didn't see anyone mention this, but this is completely wrong: "This annoyed them and they decided to make a new CPU that no one would want to use." Facts: 1

I didn't see anyone mention this, but this is completely wrong: "This annoyed them and they decided to make a new CPU that no one would want to use." Facts: 1. Itanium grew out of internal HP efforts to define their next generation architecture, presumably in response to their difficulty scaling the Precision Architecture. 2. Intel only got involved part way through. 3. Nobody ever "decides to make a CPU that no one will use". Itanium for all its issues, was used in great number, but nowhere near what it was supposed to. The failure of Itanium was a bet of handling cache misses with the compiler (like Transmeta, my alma mater) and having a design by committee (lots of nice ideas in isolation, a disaster in aggregate). There are so many technical problems with the ISA, too many features, too much performance draining complexity. Something as critical as registers is burried behind (IIRC) three levels of indirection (rotating windows, modulo scheduled registers, etc). The big bet on predication didn't help code density nor cycle time. It's really a shame that Intel throughout history has botched ISA design after ISA design. That it can be done right is exemplified by Alpha, RISC-V, and ARM64. If the rumored acquisition of SiFive happens, then maybe they can finally get it. EDIT: Typos