I see people claiming the P4 w/ HT was a snappier desktop PC but I seem to recall Athlon XP and later Athlon 64 PC's performing on par with or better than same era Intel P4's in business benchmarks like Winstone and PCMark.
Problem is that you can't really measure 'snappiness'.
In the server world there's a similar scenario...
They are more interested in the 'average response time' than in the actual performance.
For example... say you run a website that serves maybe 1000 requests per second.
Now, you can either have a system that can handle many requests simultaneously, or one that handles them all sequentially.
The first system may be slower in the absolute sense that it takes longer to process all 1000 requests. But the average response time will be rather good because it can run multiple requests at the same time. It doesn't take very long for a slot to open up for your request to be processed. It doesn't matter that a lot of other requests are still processing at the same time, you don't have to wait for them.
The second system will have to queue up all requests, and if you're in the front of the queue, you may get very good response time... but if you're at the back of the queue, you get a very poor response time, because you have to wait until EVERY request before you is complete.
Sometimes people choose the system that is 'slower' in the absolute sense, but gives a better response for your website on average. It's nicer if all users have to wait ~10 ms for the page to load, than it would be to have one user get the page in 1 ms, and another having to wait for 100 ms.
In the first case, nobody will feel that the site is slow.
In the second case, some will feel the site is fast, but others will complain that the site is too slow for them.
But there's no easy way to measure that.