I've just spent a bit of time benchmarking different cache timing configurations on my 486 and thought I'd share the results in case it's interesting or useful to anyone.

Specs:
UM4980 socket 3 motherboard
Intel 486 DX4 100MHz Overdrive SZ959
2x 16MB 72 pin FPM
256KB 15ns external cache
1MB Hercules Stingray Pro
SB Pro 2.0

For testing I ran timedemo demo1 in Quake. I didn't test every single possible combination of cache settings, I was just experimenting as I went through to see what worked and what actually improved performance. Settings in bold are what changed from the previous test.

BIOS defaults
External cache: Write through
Cache read hit state: 3-1-1-1
Cache write wait state: 1
DRAM wait state: 1
DRAM page mode: Normal
ATCLK: PCLK/4
Quake timedemo demo1: 6.8

External cache: Write back
Cache read hit state: 3-1-1-1
Cache write wait state: 1
DRAM wait state: 1
DRAM page mode: Normal
ATCLK: PCLK/4
Quake timedemo demo1: 7.8

External cache: Write back
Cache read hit state: 2-1-1-1
Cache write wait state: 1
DRAM wait state: 1
DRAM page mode: Normal
ATCLK: PCLK/4
Quake timedemo demo1: 8.9

External cache: Write back
Cache read hit state: 2-1-1-1
Cache write wait state: 0
DRAM wait state: 0
DRAM page mode: Normal
ATCLK: PCLK/4
Quake timedemo demo1: 9.8

External cache: Write back
Cache read hit state: 2-1-1-1
Cache write wait state: 0
DRAM wait state: 0
DRAM page mode: Fast
ATCLK: PCLK/4
Quake timedemo demo1: 10.1

External cache: Write back
Cache read hit state: 2-1-1-1
Cache write wait state: 0
DRAM wait state: 0
DRAM page mode: Fast
ATCLK: PCLK/3
Quake timedemo demo1: 10.4

External cache: Write back
Cache read hit state: 2-1-1-1
Cache write wait state: 0
DRAM wait state: 0
DRAM page mode: Fast
ATCLK: PCLK/2
Quake timedemo demo1: 10.6

So just from increasing the timings from the BIOS defaults, I went from 6.8fps to 10.6 fps which is a 55% increase. I have 1024KB cache being delivered which I'm going to test to see if that gives any improvement over 256KB.