🤣, i think quake is the only fpu-related test of the four.
cyrix surely has very good integer performance, however its not 100% 586 compatible, so some tests like linx would not run at all.
cyrix running at 1x clock is also the slowest performance achievable on socket 7 boards, although it requires a 2.5x multiplier and therefore not available on socket5.

noshutdown wrote:

🤣, i think quake is the only fpu-related test of the four.

Yes it is, but kept it nice and easy. Also by using Phil's Benchmark you can compare with several hundred results in his database.

noshutdown wrote:

cyrix surely has very good integer performance, however its not 100% 586 compatible, so some tests like linx would not run at all.

Minor problem really. Most people will use it for DOS and Windows (Win98 was slow on it btw!)

noshutdown wrote:

cyrix running at 1x clock is also the slowest performance achievable on socket 7 boards, although it requires a 2.5x multiplier and therefore not available on socket5.

You can go slower on Socket 5 boards as many of my boards can run a Pentium on 25-33mhz with undocumented settings. 2.5x multiplier is also available on some boards with a Powerleap or equivalent adapter. On Socket 7 you're more restricted, and the Cyrix running at 1x50mhz(which is normally the slowest FSB you can find on Socket 7) is probably the slowest you can get on a regular motherboard.

vetz wrote:

You can go slower on Socket 5 boards as many of my boards can run a Pentium on 25-33mhz with undocumented settings. 2.5x multiplier is also available on some boards with a Powerleap or equivalent adapter. On Socket 7 you're more restricted, and the Cyrix running at 1x50mhz(which is normally the slowest FSB you can find on Socket 7) is probably the slowest you can get on a regular motherboard.

with what settings can i make a pentium(old socket5 models) run at 0.5x clock? and would it work on socket7 boards?

I surely overlooked this thread somehow. Aside from the MMX-enabled chips, it seems Cyrix did not make any improvements whereby these 4 benchmarks could benefit from. Two tests even indicate some performance loss. I wonder if Cyrix had to sacrifice a feature for stability at higher frequency chips. Do you have a Cyrix MII, 2.2 V chip to test on the 430FX board? Basically, and MII-400 or 433. There is a rare MII-366GP at 2.2 V as well.

Is write allocate enabled on your 80 MHz Cyrix chip? Did you check any of the Cyrix register settings?

Based on the socres, I am tempted to conclude that your SiS-based socket 7 board is a piece of junk.

Try running the Cyrix at 1x25, unset the turbo, turn off L1 and L2 cache, and try to run the doom timedemo.

The recent socket 5 talk has stumbled me onto this thread again. Vetz, did you try to run any enhancements for the Cyrix 6x86, or to ensure that No_Lock was set? The settings I use for this chip can be found in this thread, Register settings for various CPUs

Basically, for the original 6x86, the only noticeable performance benefit was from NO_Lock. Run this,

1FASS.com /C

It is remarkable how the Cyrix 6x86 outperforms the Pentium by the amounts shown. I wonder had it not been for gaming, would the outcome of the MHz war been different. What percentage of people were into games in late 1997 compared to those who didn't care about FPU? I mean, I too would have focused on integer performance rather than floating point.

vetz wrote:

Note: The AMD K5 (new) could not be tested on the P54TP4 since it requires a 2.5x multiplier on the motherboard to run at the internal 2x setting which the board does not support.

Have you actually tried new K5 on the board? I guess your board has one jumper for multiplier? You might get:
1) 1.5x if BF1 is defaulted at high
2) 2x if BF1 is defaulted as low and jumper is open.
3) 1.75x if BF1 is defaulted as low and jumper is closed.