Fox5 said:Thye also had a race between a man running and a car, and the man one, despite having a fraction of the horsepower of the car. Which ties in nicely with the whole idea that it's not just how big the numbers are, but where and how they are applied. Comparing component metrics really ain't that useful
Wait, the man won? Over what distance? Did they not push the gas pedal on the car?[/quote]Over a 26 mile Marathon (London Marathon route) at rush-hour. There were no limits other than legal ones (30 mph speed limit on some roads).
The man beat the car by 11 minutes (journey time about 2 hours, 30 minutes for 26 miles). The moral of the story is that the biggest engine in the world is sod-all good to you if the bottlenecks choke it. Only when you see the components in context of their application can you make fair predications. eg. If with the above news I were to say a man and car were to race on the motorway a distance of 2 miles, we can be pretty safe betting on the car! But just saying a man and car are racing isn't enough info. No moreso is 'CPU x has n megahertz while CPU y has m megahertz'. And that's why a comparison between alpha code running on alpha hardware is incomparable to final code running on final hardware. Even if metrics are only 2x bigger, final hardware might be 5-10x faster, or notm depending on so many factors.
