I think Win95 had issues with CPUs faster than 350MHz, not sure if it was fixed in later versions like OSR2. Here's the fix.

As to the hardware part, I think 440BX should be a pretty solid testbed. My Evans & Sutherland Freedom (early OpenGL PCI accelerator with an onboard S3 968 chip) wouldn't work on a VS440FX Socket 8 board (even though it's 1 or 2 years newer), but would happily work on a Chaintech 6BTM0. Go figure.

The OS doesn't directly interact with the CPU, rather with the motherboard chipset. If there are (good) chipset drivers for Win95, it will work fine. If not, less so. i440BX is very nicely supported, so should work just as well as older motherboards. Consider that Windows 95 was current when the i440BX was released, so there's nothing strange about that combination.

Thanks, eventually I might try using an older slower cpu like a Celeron 333 but any time I install a new cpu is absurdly stressing on this Slot mechanism when switching them. So I preferred to use a single cpu for most tests of different vga periods and the P3-500 seems like a good balanced choice.

My P2 400 duel booted Win95c and NT4 for the first year I had it, This is back in late 98 when drivers weren't so mature as well.

There was a thread here that investigated what happens if software tries to use SSE with Windows 95. The OS is not aware of SSE registers. I think problems can happen. I would probably avoid SSE CPUs with Win95.

A non-SSE aware OS should not enable SSE, which is disabled at boot time. If some buggy application still thinks that SSE is available, you will simply get an illegal instruction exception at a very first SSE instruction.

If does bad thing when do thing, don't do thing which make it do bad thing.

Ok yeah I found the other thread about SSE and 95. They did do testing of it, including using a TSR to enable SSE use in 95. It is indeed not available without using an external enabler.

Re: Windows 95, 3Dnow, and SSE