Innovation
Friday
Friday 30/08/02There are three things that will never go away: death, taxes and arguments over benchmarks. This time, it's AMD claiming that the Sysmark software from BAPCo has been revised to make AMD look bad and Intel look good.
Friday 30/08/02
There are three things that will never go away: death, taxes and arguments over benchmarks. This time, it's AMD claiming that the Sysmark software from BAPCo has been revised to make AMD look bad and Intel look good. Not by deliberately changing the code to run secret Pentium 4 instructions, but by changing the mix of applications run to favour those that the Pentium 4 just happens to run better. Well, perhaps so. Perhaps not. Does it matter? Benchmarks that run different mixes of application software in an attempt to synthesise a realistic working environment are always on a hiding to nothing anyway -- an accurate simulation would just have the processor wasting a few million instructions every second between keystrokes. If you're genuinely running some mega processing-intensive piece of software where every ounce of efficiency counts, then you're going to go and do the tests yourself. And if you can't be bothered, then it doesn't matter that much. Truth is, benchmarks don't matter, any more than the top speed of your car matters. When was the last time you bought a television based on the maximum brightness of the picture tube, or the sensitivity or signal-to-noise ratio of the front end RF amplifier? Put it another way, I bet that if you took the top executives and engineers from both AMD and Intel, and sat them down at a PC where care had been taken to hide the name and speed of the processor running within, none of them would be able to work out whose chip it was or whether it was 1GHz or 3 GHz. The Pepsi Taste Challenge would be a piece of coke, er, cake by comparison. But the chip companies, locked in a vicious struggle for sales, can't possibly say that. So they'll continue to synthesise outrage together with test results, and we'll continue to report on it all, and everyone will pretend that it's very important. Just do your part and agree, won't you? To have your say online click on TalkBack and go to the ZDNet UK forums.
There are three things that will never go away: death, taxes and arguments over benchmarks. This time, it's AMD claiming that the Sysmark software from BAPCo has been revised to make AMD look bad and Intel look good. Not by deliberately changing the code to run secret Pentium 4 instructions, but by changing the mix of applications run to favour those that the Pentium 4 just happens to run better. Well, perhaps so. Perhaps not. Does it matter? Benchmarks that run different mixes of application software in an attempt to synthesise a realistic working environment are always on a hiding to nothing anyway -- an accurate simulation would just have the processor wasting a few million instructions every second between keystrokes. If you're genuinely running some mega processing-intensive piece of software where every ounce of efficiency counts, then you're going to go and do the tests yourself. And if you can't be bothered, then it doesn't matter that much. Truth is, benchmarks don't matter, any more than the top speed of your car matters. When was the last time you bought a television based on the maximum brightness of the picture tube, or the sensitivity or signal-to-noise ratio of the front end RF amplifier? Put it another way, I bet that if you took the top executives and engineers from both AMD and Intel, and sat them down at a PC where care had been taken to hide the name and speed of the processor running within, none of them would be able to work out whose chip it was or whether it was 1GHz or 3 GHz. The Pepsi Taste Challenge would be a piece of coke, er, cake by comparison. But the chip companies, locked in a vicious struggle for sales, can't possibly say that. So they'll continue to synthesise outrage together with test results, and we'll continue to report on it all, and everyone will pretend that it's very important. Just do your part and agree, won't you? To have your say online click on TalkBack and go to the ZDNet UK forums.