P4 benchmarks – the real thing?

Could be - they're anodyne enough...

Yet more Pentium 4 benchmarks have popped their tiny little heads over the parapet, but this time they could be the real thing.

This claim, made on JC's pages which was sent them anonymously by the enigmatic 'One guy on the net', could well be true, because the numbers follow Intel's traditional iMark style.

iMark is an abomination introduced by Chipzilla which, while showing the relative performance of its processors, prohibits that performance from being compared to any other product, even from within Intel. You can compare PIII with PIII, or Celeron with Celeron, but not PIII with Celeron, or - God forbid - with Duron.

P4 continues this great tradition by being compared in relative terms with a 1GHz PIII:

3DMark 2000

P4-1.40: 1.08
P4-1.50: 1.13
Quake III
P4-1.40: 1.21
P4-1.50: 1.23
SysMark 2000
P4-1.40: 1.01
P4-1.50: 1.06

These numbers are allegedly Intel Confidential numbers for the processor, using final product, not preproduction parts. While the performance differentials are hardly huge, remember that PIII is banging its head on an absolute performance ceiling, while the P4 is a mere baby at these core speeds. Remember the narrow performance gain moving from a 400MHz PII to a 450MHz PIII?

There's a lot more to come from the architecture that should finally give Rambus a chance to shine. ®