I personally haven't had any issue with Windows 95 OSR 2.5 and partitions beyond 32 GB. I'm using a 120 GB drive in more than one system, with an 8 GB NTFS partition for NT4 and the remainder of the 120 GB for Win95 in FAT32. Scandisk works as far as I can tell, and I don't have any access issues or boot up issues.

I do recall there being issues with older BIOSes not being able to see beyond certain sizes, which changed with time. There was the 512 MB limit, then some 2 GB limit for FAT16 DOS, then I think some controller cards only saw 8 GB, some up to 1 GB. Then there a motherboard BIOS limit of 32 GB for IDE drives, which there have been many custom and manufacturer fixes for to extend it to 64 GB or 128 GB. Back when the 32 GB limit was an issue, there would be a jumper on older hard drives which the user could set that would limit the usable size of the HDD to 32 GB. This was intended for motherboards which had the 32 GB BIOS limit.

Most of those were talking about partition sizes not the whole drive capacity.
I'm sure I confused different limits by reading about too many at the same time.
I've been using sub-8GB drives on the workbench for so long to avoid the problems that I forgot what the other limits are.
I was reading up because for a change I want to put a couple of retro machines together permanently and I want drives bigger than 8GB.
.

Someone had mentioned that the 430NX chipset had poor memory speeds, so during my preliminary examination of the motherboards, I did a quick run of Cachechk -v7 using a Pentium 100. CMOS timings left on auto for all the motherboards. All boards using FPM. I tried EDO in the FX board, but the cachechk results didn't change.

In the chart,
430FX = Asus PCI/I-P54TP4
430NX = Gigabyte GA-586IP
SiS 500 = Asus PCI/I-P54SP4

The SiS 501/502/503 chipset seems to have the worst memory timings, while the NX had a slight edge for L1 and memory read, but fell short with memory write times.