Reading time: 1 minute
Math games in benchmarketing
BMSeers points to an rewritten press release of IBM that shows how silly the “Power6 30x Sun comparision” is and how IBM benchmarketing works.. At least IBM clarified the way how those numbers were crafted. My favourite one:
IBM then assumed a IBM utilisation of 60% versus a Sun utilisation of 20%, which nicely gave them another 3x advantage
No problem with the consolidation of unused resources, but when you want to make a point about the speed of your product, you don´t do this with some math games giving you an mathematical, but not technical advantage. At least you should make clear that you assumed different levels of utilisation. The IBM press release tries to generate the impression that the P6 systems are much faster than they really are. They take an actual system, but use processors introduces 2 years ago. They calculate with weird performance ratios. I assume, that they wanted give the impression that their p570 are much faster than the actual SF890 by “forgetting” to write down most of the math, assumptions and conditions. And dear IBM, nobody really uses SF890 as an appserver. SF490 or T2000 would be much better systems for such tasks. But well, you used the bulkiest system in your portfolio to “substantiate” your silly claims. This press release was a really bad foul. PS: This pile of bullshit rises the question, what´s the problem with Power6, when they need to play such games with consolidation and meaningless TPC-C benchmarks to state the superiority of their technology..