The computer was used a lot maybe, the battery never discharged much and now it has had plenty time for it and still hasn't gone bad yet. Battery in my PCI 486's RTC module was 24 years old and to my surprise it still read full 3V !

Now I wonder if this board can't switch turbo states while the power is on, similar to Gigabyte implementation. But "Boot Up System speed" option in BIOS also does nothing.

My 430FX motherboard doesn't have header for turbo switch and if I set the boot up system speed to low in bios it just disables all the cache, so if I have caches already disabled in bios settings setting boot up system speed to low does absolutely nothing. So I don't think 430FX motherboards can do anything else to change speed of the cpu.

Time to raise my thread back from the Necropolis. During my recent test I have found some oddity with async (pre-installed on the board) L2 cache on this board.

SpeedSys results:
L2 cache ON (Asynchronous)

L2 cache OFF

As you can see, L2 cache is hampering my memory performance.

And here's some other benchmark results.

L2 cache ON (Asynchronous)
3Dbench 1.0c - 104.1 fps
Doom - 1238 realticks / 60.3 fps
PCPlayer 320x480 - 31.3 fps
PCPlayer 640x480 - 14.0 fps
Quake 320x200 - 28.8 fps
Duke Nukem 3D VESA 320x200 (rooftop scene) - 70 fps

L2 cache OFF
3Dbench 1.0c - 118.0 fps
Doom - 1282 realticks / 58.2 fps
PCPlayer 320x480 - 30.2 fps
PCPlayer 640x480 - 13.8 fps
Quake 320x200 - 30.3 fps
Duke Nukem 3D VESA 320x200 (rooftop scene) - 67 fps

Quake and 3Dbench are slightly slower and everything else is slightly faster, with L2 cache enabled.

So what is the problem here? Iwill BIOS bug? 256Kb cache is not enough? Async cache can't keep up with EDO memory? Ancient aliens?

And so I've switched to pipeline burst cache (COAST module).

SpeedSys results:
L2 cache ON (Pipeline Burst)

And other benchmarks:

L2 cache ON (Pipeline Burst)
3Dbench 1.0c - 123.3 fps
Doom - 1174 realticks / 63.6 fps
PCPlayer 320x480 - 36.7 fps
PCPlayer 640x480 - 16.2 fps
Quake 320x200 - 34.5 fps
Duke Nukem 3D VESA 320x200 (rooftop scene) - 86 fps

In conclusion, looks like asynchronous cache is really slow for a Pentium system. It provides very little benefit in most cases and may harm performance in some. While pipeline burst cache will boost performance by 20-30%.

Was curious about this and tried to benchmark the same with some extra tests.

So yes. I f you have a 430FX motherboard with asynchronous cache it might be beneficial to disable it. In order workloads it still improving things slightly

More details

https://dependency-injection.com/intel-430fx- … riton-l2-cache/

I speculate that async cache support was more of an afterthought. Intel motherboards had only pipeline cache or no cache option.

In fact one of the most popular Intel 430FX motherboards used in by many OEMs - Intel Zappa was designed exclusively for asynchronous SRAMs. It is true that it often came without any SRAMs.

I think COAST and PB cache was introduced by later boards like Intel Advanced/EV...

I think they couldn't make slow and fast options for both async and PB, so they had to reserve one for each. And PB cache didn't required slow option anyway.
But to be fair, I had similar behaviour on 60-66 Mhz bus overclocked 486 system: due to timings, performance boost from 15ns L2 vs no L2 cache is almost negligible. And 430FX boards were usually equipped with common 15ns SRAM too.