I think I have a 915/skt478 mobo too, though mine still runs classic DDR1 and is PCI-E + AGP. Asrock P4Dual-915GL.

As for Cedar Mill on 865 - IIRC they might use the same CPUID? I know I was looking at my MSI 865PE Neo3-V's CPU list and found no mention of the 631 being compatible.
Despite that, it did POST with the 631 and detected it properly (and enabled HT properly).

Socket3 wrote on 2023-08-29, 17:58:

On the other hand, I've had little to no luck upgrading OEM systems that use the chipsets you mentioned. I tried swapping dual core chips into Dell Optiplex, IBM Netvista and HP i915 or i925 prebuits with no success..... maybe this is mostly an OEM thing?

This may no longer be the conventional wisdom, but way back then, the general understanding was that the i915/925 had removed dual-processor/core support. The i865 had it, somewhat 'unofficially' (i.e. probably shared too much design with a Xeon chipset), then when somebody in Taiwan cooked up a way to design a dual-processor i865 board, Intel was very grumpy and made sure that you couldn't do the same thing with the i915/925. Then, of course, Intel launched dual-core chips (perhaps in response to AMD - I can't remember who came up with dual-core single-socket systems first) and this became a problem, so they launched the i945/i955 to bring dual-core support back.

(This is also what let... most of the Taiwanese manufacturers... make LGA775 i865 boards that supported at least hotburst 90nm dual-core CPUs to give people with AGP/DDR1 an upgrade path to dual-coreness.)

Did somebody in Taiwan actually find a way to get dual-processor/core support out of an i915 while I wasn't paying attention?

PcBytes wrote on 2023-08-29, 21:11:

Despite that, it did POST with the 631 and detected it properly (and enabled HT properly).

It may depend on what stepping that 631 was, the later ones had newer VR06 configuration and the earlier ones had VR05A which is closer to Prescott and may work without official support.

Socket3 wrote on 2023-08-29, 17:58:

I also remember messing about with an Asus - can't remember what model, but it had an intel 915 chipset - P5GD or P5GD2 perhaps? - the system I liberated it from came with a Celeron D, and despite no mention of compatibility on Asus's website, the board booted up and ran fine with a Pentium D 925....

I guess it could be ASUS P5GD2-X based on i915. But I've never had one to test it with Pentium D 65nm

Is it safe to use Cedar Mill with Gigabyte GA-8IPE775-G (rev. 1.1) ?

RiP wrote on 2024-01-04, 19:54:

rev. 1.1

Only 1.2. They share BIOS so it’s a hardware limitation.

I have a ThinkCentre A51 USFF coming tomorrow, model 8107-71G. This is an early-ish 775 machine with a 915G chipset onboard, but it's also a transitional model, Lenovo taking over most of the IBM stuff past this point. Somewhere in 2005 there was a microcode update added to a previous BIOS release (2GJT17A) from the release notes.

12GKT17A/2GJT17A
2- Updates video BIOS to build number 1233
3- Updates microcode update files
4- Fixes a problem where the system reboots continuosly with new
5  processors
6- Fixes potential 0162 errors
7- Fixes a problem where the system hangs during S3 resume 
8- Fixes intermittent 0197 error
9- Implements Intel 915/925 specification update

Using MCExtractor: https://github.com/platomav/MCExtractor I was able to figure out I have the following CPUIDs in the latest update file:

F62 is a Cedar Mill, so sufficed so say I should be okay there getting an earlier stepping.
The interesting one is F61. I can't for the life of me figure that one out, but CPU-World's CPUID DB has F61 down for a Xeon 5060? That makes no sense as the 915 doesn't support dualcore.

Did some of the Xeon codes overlap the Extreme Editions?

BitWrangler wrote on 2024-03-22, 03:44:

Did some of the Xeon codes overlap the Extreme Editions?

They might have but I don't think this applies to this, the only extremes that can work on that machine are the Prescott ones and those are a separate cpuid I believe.