That is what it supports yes. I would test to see if a higher clocked coppermine will run at reduced speed on the 66 bus. However, on this platform a 66 bus is a real bottleneck.

Buy a few Mendocino Celerons 400. One of them will work at 500Mhz (6x83.3Mhz). Get a Riva TNT and it will serve you well.

st31276a wrote on 2024-01-27, 18:55:

That is what it supports yes. I would test to see if a higher clocked coppermine will run at reduced speed on the 66 bus. However, on this platform a 66 bus is a real bottleneck.

I didn't think that you can run Coppermine voltages on these older boards. I also read that the 440LX chipset is not compatible with these instructions. I experienced this issue previously with a 440EX:

Windows 98 can hard lock when pairing a Pentium 3 with an older 440 chipset

LX cannot address SSE. I have 566 and 633 Celerons here - both will POST on a Zida LX98-AT but hang up randomly. (sometimes even when on BIOS screen.)

Out of curiosity, does anyone know how the chipset interacts with SSE instructions to cause them to fail?

Interesting. I do not think the chipset touches instructions directly. Perhaps it is more of a microcode problem.

I found that my motherboard (ASUS MEL-M) had a BIOS update which states that Coppermine stepping B0 is specifically supported. I thought that this was very interesting, since it was mentioned previously that the 440LX cannot address SSE.

Well, it works and it doesn't. I think that it's "supported", but SSE is stills broken. When testing a TNT2 with driver version 43.45, the system hangs at startup. This is not the case with Mendocino CPUs. I tried using a Coppermine Celeron 533A (SL46S), the lowest spec one I could find.

https://www.asus.com/supportonly/mel-m/helpdesk_bios/

Add microcode update support for Coppermine B-0 stepping

I suspect that it might be possible to get it working with an older driver that doesn't use an SSE codepath, but I feel like I would end up encountering other random crashes in similar instances where SSE instructions are accessed.

AlexZ wrote on 2024-01-27, 19:08:

Buy a few Mendocino Celerons 400. One of them will work at 500Mhz (6x83.3Mhz). Get a Riva TNT and it will serve you well.

When testing a Mendocino Celeron 400Mhz overclocked to 500Mhz with a 83.3Mhz FSB and a TNT2 Pro I got the following results in 3DMark99:

3660 3DMarks
4593 CPU 3DMarks

With a Mendocino Celeron 533Mhz (not overclocked) I got the following results:

3426 3DMarks
4477 CPU 3DMarks

I wasn't able to overclock the Celeron 533 to 666Mhz, but I was able to get a small boost to 600Mhz (8*75Mhz):

3829 3DMarks
5071 CPU 3DMarks

There used to be adapters that allowed any 66mhz celeron to install-766mhz. I’ve never had much luck with adapters

Kahenraz wrote on 2024-02-22, 22:25:

I wasn't able to overclock the Celeron 533 to 666Mhz, but I was able to get a small boost to 600Mhz (8*75Mhz):

3829 3DMarks
5071 CPU 3DMarks

You may be able to get Mendocino 433 to work at 6.5x83.33 (541Mhz) provided you have sufficient samples (at least 5). 500Mhz is the safest OC though. 600Mhz will not be stable.