Bench(mark) graph multipliers

[Cat]

It would be nice to add a performance multiplier to a product comparison. See http://www.anandtech.com/bench/Product/287?vs=53

I want to roughly scale the Q6600 up to 3.4 GHz by multiplying its benchmarks by 1.42. That way I can have a good idea of how much the architecture impacts performance. Each benchmark must have a lower-or-better-is-higher flag, but it seems trivial to do.

 

[ViRGE]

Maybe I'm missing something here, but this doesn't seem to make a lot of sense. Performance doesn't scale linearly with clockspeed in every application (though at times it can get close).

 

[Cat]

That's a very good point. The multiplier is just a faster way of very roughly estimating things without doing it by hand, but maybe it needs a disclaimer so that people don't misinterpret the data.

The benchmark data could also provide a best guess at what sort of speedup clockspeed provides for an architecture under a certain application. I assume there are 3 or 4 CPU speeds per arch run under the same application in AT's database. This is what I do by hand right now, and it's a pain.