[QUOTE=trboyden;1201077]Application testing is a measure of how fast an application runs on THAT machine.[quote]
Yes, that's exactly what I've been saying. That's the point to the whole thing. You want to know how fast your apps will run on a particular machine.
From that information and based on the testing of the same application on another machine with the same OS and software configuration, you can surmise that the second machine's hardware causes the application to be slower or faster. You use a hardware test like SPEC to verify those assumptions and isolate which hardware component creates the bottleneck that slows that computer down or offers an advantage in speed.
The problem with SPEC as those on the web sites I provided said, SPEC DOEXN'T tell you how fast your apps will run.
SPEC doesn't test the machine, it only test part of what the cpu and, sometimes, what the memory bus can do.
SPEC doesn't test Altivec on the PPC, and it doesn't test the various SSE's on the x86, and that's just the beginning.
SPEC is like testing how much horsepower you have in your car, but it doesn't tell you how fast that car can go, because it doesn't test anything other than the engine.
Like I said SPEC is not all-inclusive, you will need to use other tests to get the big picture or to narrow down to a specific measurement to further enhance the analysis.
A great example of this is when the first tests of Windows were done on Intel Macs. Those first Intel Macs with their newer Intel hardware ran Windows better than other machines available during that period. Apple still uses that in their marketing today. SPEC tests showed that the Intel hardware in those Macs was capable of running Windows faster than other PCs available at that time. Did you need the SPEC test to make that conclusion? Maybe not, but your were able to verify your conclusion and then claim it as fact.
Well, we can agree on part of out argument then, that's progress.
But, you can see some of what we're saying with the Win-on-Intel tests. You don't need SPEC to measure what you've already measured. Spec may tell you something that agrees with your OS and software tests, but it may not.
If the OS and software makes heavy use of SSE, then SPEC tests won't help you very much. The same thing is true of the bus, HDD, and videocard, which Apple uses extensively for the OS these days (as do other software developers), as you know.
Also, consider the fact that Vista has been reliably reported across the Windows sites as being considerably slower when running programs than XP is. SP1 for Vista has also been shown to not give a noticeable speed increase, despite MS's statements. Some estimates are that Vista runs many programs at HALF the speed of XP on equivalent machines.
This is something else that SPEC won't tell you.