@ mOBSCENE
Have you tried testing 1 stick at a time of SDRAM at cas3 then cas2 timing and benchmarking. Maybe more issues with the chipset caching area? Nice thing about SDRAM is it didn't have to run in pairs 😀
I have a Socket 370 board that will recommend cas3 timings over cas2 when I force it in the bios. It won't change it on its own though. It was common for manufacturers back in the day to promise the hardware could do better than it would really do and could be the bios programmed to play it safe for stability.
@Tetrium
The Tualatin was killed off by Intel because they didn't want it to compete with the Pentium 4 line since they were being produced and sold at the time period. There was nothing stopping them running it up to 2GHz or more. As results would show that a Tualatin could beat a Williamette clocked around 400Mhz higher. Netburst was rushed out anyways to compete with the athlons and didn't compete well until the northwoods showed up. The Tualatin as well didn't support higher bandwidth for DDR and Rambus which didn't help its case.
The Pentium 3 1.13 coppermine was just OC'ed from factory as Stated. TomsHardware and HardOCP found it to not run properly then Intel recalled the CPU as to why you never see one in the flesh. The coppermine used aluminum interconnects which did limit clocking a bit before the dieshrink.
Celeron was a line of its own and I do own a few of them. The celeron 1100 coppermine core, 1300/1400 Tualeron. 366 Celeron Mendocino and a 500MHz.
Edit: Didn't mean to break the OPs topic with my post, sorry.