40 FPS in GLQuake sounds pretty good indeed. You could have tested something like 800x600 as well perhaps.

It's interesting how Speedsys hints at the Y2k bug, yet the date still gets listed correctly below.

A small hint. With 3DBench as you approach 100fps you want to be on the 1.0c version for higher accuracy. And for low scores the 1.0 is more accurate.

Which TNT card and drivers are you using? There was only one driver version I found for the Creative TNT card which worked well on a 486. I was not able to get OpenGL working with the Creative TNT card. For OpenGL on a 486, I like the Voodoo3 in Windows 9x and the Matrox G200 in NT4. I have generally found the Voodoo 3 card to be a lot easier to deal with on 486 boards than the original TNT. Refer to this thread, Modern graphics on a 486

You upgraded from how much cache and in how many banks? If your previous cache was using 9 chips and your 512K upgrade is still using 9 chips, I don't see how this would enable cache interleaving. However, if you previously had 5 cache chips and went to 9 cache chips, I can see how you might get benchmark improvement for motherboards which support cache interleave. Can you show a Speedsys screen shot of before and after the cache upgrade?

Your double-banked L2 cache speed with 512 KB was 49 MB/s in speedsys. For single-banked L2 cache on a UMC board and a POD83, I get 47 MB/s. Your memory throughput is 43 MB/s, whereas for my board, I get 37 MB/s. Your board definitely shows, both, L2 and RAM improvements over a UMC board. I am left wondering what exactly is causing the improvement. I would be interested to see the Speedsys chart for your system with single-banked and double-banked cache, as well as with 1 memory SIMM filled and with 4 memory SIMMs filled. This will hopefully shed some light on memory and cache interleaving.

Do any BIOS settings stand out to you as being unique to the ALI board?

I tried installing cyrix 5x86-120 chips without much success, at first the mainboard would'nt turn on at all or have any beep code errors.
I tried a gainbery cpu with the full voltage setup for 5V and it did power on but its unstable, some of the problems would be hanging after the ram count or running better if I changed the Write Back cache to Write Thru in bios setup.

http://motherboards.mbarron.net/models/486pci/m918i.htm I had the mainboard jumpers set to "Intel P24D/M1SC" whilst I had the POD83 running all those settings and it ran fine, obviously it was meant to be set for P24T but it had been running perfectly? Their is even a setting in the bios for "M1SC Linear Wrapped Mode" so you'd think the cyrix 5x86 would run fine but it does'nt.

For what its worth this is the unstable speedsys report for the cyrix 5x86-120, note how it doesn't report L2 cache speed and not entirely sure if all video memory/memory bandwidth numbers are correct, shame it would interesting to put the cyrix through its paces.

Perhaps my mainboard is too early a revision to work properly with the cyrix chip 21/7/95?

Actually, I was just hoping for 4 Speedsys screenshots of your system with the POD83. These are the numbers I was using for comparison to see if your board has some type of interleaving going on.

I would be interested to see the Speedsys chart for your system with single-banked and double-banked cache, as well as with 1 memory SIMM filled and with 4 memory SIMMs filled.

I'm not sure why the Cyrix 5x86-120 is not working. What voltage are you feeding it? Try it at 100 MHz and 3.45 V first.

I actually removed 4 cache pieces and changed to jumpers for 256K but the machine wouldn't post, I then decided to put the 4 chips back in but not before removing the cpu to create enough room, also removed power and data for the cd-rom drive as well, I thought I F##ed the machine after I tried all this putting it all back together and it still not working, one of the cache chip pins had bent completely but I was able to fix it and now the machine is working again with 512k, Im never going to mess with cache chips again unless im upgrading it, if its not broken....

As for the cyrix, the first chip was a barebones model and set it for 3(3.3V?), no response however with the gainbery chip it did sort of work at 5V, windows would crash etc, I might look into it again but I'll keep an eye out for another ALI board If I can get one for a decent price.