Yep, went to the site to check it out.
First of all, I noticed this little tidbit, the site disclaimer for the data:
"The results published by SPEC have been reviewed by the SPEC organization prior to publication. However, these are submissions by member companies and the contents of any SPEC reporting page are the submittor's responsibility. SPEC makes no warranties about the accuracy or veracity of this data."
In other words, the testing was done by the submitters, NOT by SPEC. Therefore, the data is to be taken with a grain of salt.
Secondly, I noticed from the names of the systems tested that the vast majority are meant for workstation or server use. AMD has only recently started entering this region, so the fact that there is little data on AMD systems listed, suggests that it most likely doesn't reflect the marketplace as a whole.
More importantly, while the benchmarks provide some comparisons for the different CPU's, the site itself says that they are meant to test the processor by itself. Unfortunately, the average computer user cannot simply sit down to a monitor, plug a CPU into it, and expect to run their programs. They have to have the full system: hard drive, CD-ROM/burner, video card, and other peripherals, all of which will come into use at one point or another. Hence, while synthetic benchmarks such as SPEC uses have their usefullness, they cannot completely replace benchmarks created by using real-world software -- that is, software that the average user will actually use.
Finally, after some careful comparisons of the numbers, while the P3's and P4's do tend to beat out the Athlons in these tests, they STILL show that the P4 is not that great of a chip. More importantly, the type of memory used is also going to affect the benchmark results. Ignoring costs, RDRAM is faster than even PC2100 DDR RAM, which means that you have the Athlons competing with a speed handicap in the RAM department. That's like taking 2 Olympian sprinters, and telling one of them he can't move his arms while running, or worse that he has to hold 1 leg behind him. Of course he'll lose the race
Anyway, just to hightlight the anaylysis, I used the SPEC CINT2000 results. 6 P4's, 32 P3's, and 2 Athlon 1.2GHz's were running Windows (NT4 SP5 or Win2K, depending on the manufacturer; Athlons all used W2K, the Pentiums were split). From what I could tell, the OS had very little effect on the benchmarks.
Fastest Base Rating: Fujitsu Siemens Celciius 460 (P4 1.5 GHz, 512 MB RD800 RAM) and Intel's D850GB mobo (P4 1.5 GHz, 256 MB RD800 RAM), both with a 524 rating. The fastest Athlon system was AMD's Gigabyte GA-7DX mobo (256 MB PC2100 DDR RAM) at 443. The fastest P3 1 GHz (Dell Precision Workstation 420, 256 MB RD800 RAM) was rated at 454.
Fastest Peak Rating: Intel's D850GB mobo with 536 (the Fujitsu Siemens was right behind with 535, though). The GA-7DX board had a 496 rating. The fastest P3 1 GHz came in at 462.
Examine those benchmarks closely, though:
-- While the P4 is 25% faster in processor speed than the Athlon, it only performed 18.3% better on the benchmark (8.1% on the peak rate).
-- Even worse, although the P4 is 50% faster in processor speed than the P3, it only had a 15-16% increase in the benchmark rating.
-- While the P3 at 1 GHz seems to perform at roughly equal to the Athlon, it also has much faster RAM to work with. To truly test the P3 against the Athlon, the RAM speeds would also have to be equal. This is especially important given that currently RDRAM is more expensive than DDR RAM.
Considering the cheapest costs for these processors (on www.pricewatch.com today: $577 for a P4 1.5 GHz, $236 for a P3 1 GHz, and $232 for an Athlon 1.2 GHz), the performance per dollar for a P4 (0.908 per $ base/0.929 per $ peak) is MUCH lower than that of a P3 (1.924/1.958) or Athlon (1.885/2.111).
And despite any comments you might have about lemmings around here, you're as much an Intel lemming as anyone else is an AMD lemming. So what if the P4 has a higher clock speed, or has a better benchmark? This site didn't deny that the P4 can get higher benchmarks than Athlon. What it DID point out (and what any half-wit that can cut-and-paste data into Excel will also see) is that the P4 doesn't deliver enough extra performance for the cost being charged. When a P4 system costs double or more the cost of an Athlon system, but doesn't deliver double or more of the performance, it's not a good deal. That's why the P4 is flawed.
You come back to me with a P4 system that uses PC2100 DDR RAM, a P4 chip with 64 KB of L1 data cache (and not the 8 KB it currently has) and 64 KB of L1 instruction cache (not the current 12 KB) [BTW: that would make it identical to AMD's Athlon L1 cache], and THEN show me that it can deliver more than double an Athlon's performance. Then I'll believe that the P4 is worth the money Intel is charging for it. Until then, take your feet out of your mouth and start using your brain.