gerwin wrote:

Totempole wrote:

I can safely say my CPU speed dropped by atleast 1/3 after doing so. It's running at about the same speed as a Pentium 200MMX, even slower than a Pentium Pro.

I suppose. Looking back, PC player bench becomes 33% slower without L2, but Quake 1 just 15% slower:
133 MHz Challenge - 5th/6th gen CPU per clock performance

Results definitely vary. In many old DOS games, it has no effect at all. I actually just used MDK's performance test as a guide. Probably not the most accurate way to measure. 🤣

I've got an unlocked p2 333 from 1998 week 30 running at 300 mhz (3x100) and i can downclock it by software in windows to 200 (softfsb or setfsb). Disabling l2 cache doesn't make much of a difference. The advantage of running below 233 is not getting the turbo pascal divide by zero error, but it's not that important anyway and programs are easy to patch.

By the way throttle works by switching the cpu to a sleep state during brief periods of time (the southbridge sends the STOPCLK signal to the cpu) so it won't see some clocks and it's like it was running at an effective lower clock speed but it's not actually reduced. Those were the old t-states used in old laptops to prevent overheating, later they were replaced by p-states wich actually reduce the working speed of the cpu. Throttle works very well as long as you don't go too far, not below level 5 i think, otherwise time spent in those sleep states becomes important and performance takes a nosedive, specially for graphics and you get slow and distorted sound. So i think it's only useful to some extent to get the performance level of the previous cpu generation.

Benchmarked a 266 Klamath today and it supported multipliers for 133, 166, 200, 233 and 266 MHz.

dondiego wrote:

The advantage of running below 233 is not getting the turbo pascal divide by zero error, but it's not that important anyway and programs are easy to patch.

Running at 233MHz with L2 Cache disabled fixes the Turbo Pascal error for me, but the games themselves usually still run way to fast... eg. Highway Hunter. Moslo works great for that, but I still run into problems with Fatal Fumes. It starts up okay, but runs too fast. If moslo or Throttle gets used, even if only minimally, the game's scrolling stutters severely.

PhilsComputerLab wrote:

Benchmarked a 266 Klamath today and it supported multipliers for 133, 166, 200, 233 and 266 MHz.

Would be very interesting to see the results. Have you posted them online anywhere yet?

I have a 440LX based board here that supports 66 and 60mhz fsb. When selecting 2x multi it does say 133 and 120mhz at post, with a p2 266mhz cpu.

Didn't test it any further (just wanted to see if the board would post and wasn't broken).