Replies to post #59384 on Intel Corp (INTC)

[mmoy]

I've always preferred OOO to HT as it's what I've gotten used to in programming and what the compilers are good at it. It may be that removing the assumption of OOO in compilers will result in better Silverthorne performance.

[chipguy]

HT probably dissipates less power than the OOO circuity, and HT had a better "ROI".

Hardware multithreading is at least an order of magnitude
simpler than OOO execution and is a good match for a
light weight, relatively simple, relatively high frequency in-
order CPU core, RISC or x86. For many integer workloads
an Atom core could be stalled for half of all CPU cycles
so a second thread that itself can exploit half of the first
thread's stalled cycles will increase throughtput by 50%
in theory. In reality it will be less because two threads will
knock some of each other's code and data out of cache
thus increasing miss rates.