The Intel marketing department has much to say about what "numbers" are assigned to their products. When they say 3.6 GHz as they do regarding some of their CPU's, I think they're describing optimal equivalent performance based on some arbitrary value not necessarily reflecting the performance in a single-threaded task. They've added more pins, more data paths, and lots of other features, and, in the interest of producing a "number" that people can "fairly" compare, they've come up with numbers that reflect the somewhat parallel event processing that occurs in the modern parts that didn't occur in the originals.

While it's possible that they've built 64-bit synchronous counters that operate at the high rates implied by their marketing department numbers, I have my doubts. If you attempt to aggregate the numbers reflected in their video data paths, memory data paths, and other communication paths within the CPU, you can undoubtedly come up with high parallel transfer rates like that, but few tasks attempt to use all these resources at once. There's a potential performance, and there's a real performance. When you're running a high-performance video game with sound and all manner of video effects, you're probably using as much of these resources as the architecture will allow. If you'd tried to get that original Pentium to do all that, it would probably have had to run at the high rate that the marketing department suggests they're delivering now. I'm not convinced they have the technology, yet, to deliver that in raw, single-thread, performance, e.g. even a long string of no-op's.

If these CPU's were really capable of that 3.6 GHz, it probably wouldn't take Windows XP >90 seconds to open a file, which it often does.

RE