Your described problems pointing to an issue with L1:WB in combination with DMA access. This was a common problem in 486 platforms with introducing of WB-CPUs. The WB-mode needs a tight management of changed and writen adresses in RAM and caches. A two layer WB-mode (L1+L2 in WB) in combination with a component, which writes into the RAM in parallel to the CPU and the cache line (called DMA) is not easy to manage. The world of 486 were early pioneers ...

feipoa wrote on 2019-12-28, 21:01:

No I don't, sorry. I thought SV2GX4 had the greatest potential for L1:WB?

There may be another user here, who has a board with SIS 85C471 + POD + L1: WB + Adaptec SCSI in use and can give us the hint ... please! 😀

Hate to bump this thread but could anyone of you SV2GX4 owners check whether your board works with a POD and a GUS (Max or PnP) for Duke3D/ROTT/Raptor? Duke and Rott have issues with Midi, while sounds work fine. With Raptor it's the other way around.
I have a SV2GX4 Rev 1.7 here which butchers the sound for above games whenever I try to use the GUS with my POD83 - so far the only fix I found was disabling L1 cache altogether.
Also noted that CTChip hangs when attempting to fix the dirty-bit whenever I have 1024K L2 cache installed - however it works fine with 256K in place. I'll check the modified BIOS later today.