Intel's first dual-core Itanium claimed to give comparable RISC systems a run for their moneyIntel is claiming that their forth-coming dual-core Itanium, code-named "Montecito", will give a 4-way RISC system a run for its money when used in a dual-CPU configuration (that would be four cores). The benchmark used was LINPACK, and the Intel configuration achieved 45 GFLOPS.
The previous record, said Intel, was 27.5 GFLOPS for an IBM pSeries RS/6000 system shown in a PDF, here.
Phil Brace, general manager of the Intel server group, said: "We are approaching the ability to reach a teraflop in as few as a 20 server system cluster."
All Itanium jokes aside, I've always had some respect for the architecture. Apart from the fact that the physical CPU looks insanely awesome, the Itanium has always had mind-blowing FPU performance. What snagged it was poor x86 emulation performance, and cost.
I think that if Intel had a fully-performing x86 unit on the Itanium core, and priced the processors affordably (at the very least, high-end affordable), then they might have had a winning architecture. Low-level programmers everywhere refer to the x86 architecture as an abortion. EPIC sounds like a very interesting and well-designed architecture, and with the recent push for parallel computing, it would have made sense to adopt it. AMD seems to have filled that gap with AMD64, though.
Article Link: Intel claims Itanium Montecito fastest FP on block
