Pentium 66Mhz performance test

Socket 4 Pentium Overdrive original Box

Test

Thanks for sharing those results. Is your overdrive marked as PODP5V120 or PODP5V133, given there is 120/133 on the box.

regarding the fdiv bug. does it prevent anything "normal" from happening? cause any instability? if so, why use it?

It doesn't cause instability. It gives slightly incorrect results when the two input values for a floating point division match a specific rare pattern.

foil_fresh wrote on 2020-03-25, 01:36:

regarding the fdiv bug. does it prevent anything "normal" from happening? cause any instability? if so, why use it?

The FPU is less accurate. You get wrong results when you divide certain numbers. Won't affect stability, but it has historical significance - a sort of curiosity.

Intel ran expensive chip replacement program and most chip got replaced and scrapped. That's why those original chips with FDIV bug are way less common.

The performance of this system has been great so far.

I was thinking about how to tune it even further. Then I found a datasheet of the clock generator (MX8315) with undocumented 80 MHz FSB option.

Obviously, 80 MHz is is too much for old 5V 60/66 MHz CPUs. However, it surprisingly works with more efficient PODP5V133. The 160 MHz CPU, 80 MHz FSB and generous 2MB L2 cache really gives this system wings. A Socket 4 system with asynchronous L2 can now compete even with much more modern 150/166 MHz Socket 7 systems with Intel Triton chipset. Even with more conservative 3-2-2-2 L2 timings.

Some results:

Cool. I have an Acer J3 motherboard with dual Socket4 CPU card. It originally came with P60s but I have dual POD5V133s for it. This system also has an 80mHz FSB setting, but I've always been a little nervous about testing it out.

Anonymous Coward wrote on 2020-04-19, 09:15:

Cool. I have an Acer J3 motherboard with dual Socket4 CPU card. It originally came with P60s but I have dual POD5V133s for it. This system also has an 80mHz FSB setting, but I've always been a little nervous about testing it out.

Is SMP possible with POD's? Wasn't that option removed?

The on- chip APIC only came with P54C. However, there were 2- or more way systems before that (even 486). Just vendors had to use their own glue logic to handle interrupt, maintain cache coherency, etc. They could also use external i82489DX which Intel provided long before P90.

Also EISA systems have APIC alreeady in the chipset already which might be the case of the Acer system

Anonymous Coward wrote on 2020-04-19, 09:15:

Cool. I have an Acer J3 motherboard with dual Socket4 CPU card. It originally came with P60s but I have dual POD5V133s for it. This system also has an 80mHz FSB setting, but I've always been a little nervous about testing it out.

Interesting. I have an Acer socket 4 motherboard setup with a POD5V133 and a Voodoo1 but I don't think I bothered to look for an undocumented 80 MHz setting. Something else for the list I guess.

FDIV is present on 60 to 100 MHz Pentiums on some steppings, except 486 Pentium Overdrives and not present over Pentium 120 MHz and up.

Pentium FDIV bug related!