Fortunately Intel's IA-64 was designed properly, with coarse
   multithreading, and explicit simultaneous instruction dispatch.
   There is also enough demand for IA-64 for them to keep releasing new
   versions.
   It is also fortunate that Project Zero doesn't release exploits for the
   predominant processor within thirty days, the same goes for the
   predominant web browser.
   I wonder what went wrong with processor design. Pentium 4 had about a
   hundred million transistors. Now they have roughly thirty times that.
   Given the amount of time spent on switching contexts and loading data,
   instead of improving artificial benchmarks, reducing idle time would've
   been better by allowing each core to hold thirty threads.
   Although carryless multiplication would have been useful right when AES
   was adopted....
   Extrastatecraft is a good read.