Intel: benchmarks ‘not good enough’ for Rambus
Ah, that explains it then
In a remarkably audacious explanation of why Rambust continually fails to impress when compared with the much cheaper alternatives, an Intel spin paramedic today explained why we've all got it terribly, terribly wrong.
A couple of weeks back, Intel yet again demonstrated its dysfunctional left hand, right hand communication process by simultaneously publishing benchmarks for Rambus and SDRAM memory which showed the cheap stuff was miles better. Check out Intel 815e platform thrashes 820.
The tests on the PC-133 i815e chipset came out up to five per cent faster than the Rambus-based i820. On average, the 815 performs two percent better than the more expensive Rambus solution.
But is Intel downhearted? The Hell it is.
"The benchmarks used don't reflect a real-world situation - say where you'd have Word and Excel running concurrently," commented the Intel spinmeister with a commendably straight face.
"Today's benchmarks run their tests consecutively, which doesn't give Rambus a chance to show how well it performs when there's a lot of heavy memory usage.
"We're working with the major benchmarking organisations to develop more realistic test methodologies which will show the kind of performance of which Rambus is capable."
At the time of writing, the leading benchmarking organisations we asked for comment hadn't replied.
Maybe we were asking the wrong sort of questions. ®