Shanghai CPU support for Tyan S3992 (not S3992-E)

[brainbone]

According to Tyan's chart, the Tyan S3992 can only support up to the Barcelona, where the S3992-E supports Shanghai.

The two appear to have very similar (nearly identical) hardware. Any chance of getting the S3992 to support Shanghai?

[1234s282]

Try this at your own risk , i replaced all the microcode in the S3992 with the microcode from the latest S3992-E BIOS

http://dl.dropbox.com/u/747181/Tyan%20S3...34s282.zip

Regards and Good Luck!

Admin

[paradigital]

Try this at your own risk , i replaced all the microcode in the S3992 with the microcode from the latest S3992-E BIOS

http://dl.dropbox.com/u/747181/Tyan%20S3...34s282.zip

Regards and Good Luck!

Admin

I know I'm necro-bumping a very old thread, but I tried this BIOS in my S3992, and whilst it will POST, my 6-core CPUs (2427) are incorrectly identified as dual-core Opteron "2319" (which don't exist). Then the board hangs with "CPU mismatch".

These are Istanbul CPUs rather than Shanghai, but Tyan's CPU compatibility matrix shows them as being compatible with the S3992-E (no idea from which BIOS revision though).

Is there anyone else that can take a look at the latest S3992-E BIOS, and inject the CPU microcode into the S3992 vanilla?

Latest S3992-E BIOS: here
Latest S3992 BIOS: here

Many thanks to anyone that can help!

EDIT: It would also be nice if I could have support for the Opteron 2419 (or 2419EE) CPU. No idea if this is possible or not.
EDIT2: No idea what AGESA code is, or whether the version is important, but it's perhaps worth noting that the S3992-E (and other TYAN boards that support the Opteron 2419 and Opteron 2427) have been updated to AGESA 3.5.0.0, whereas the S3992 is stuck with 3.3.0.0

[paradigital]

Well, had a little mess around last night (not really knowing what I'm doing, but nevertheless, I do have 20+ years of IT experience).

I took the latest S3992 bios, and found another 1MB AMI bios that has support for the Opteron 2419/2419EE CPU, taken from the Tyan S2932-SI (different chipset though). Using mmtool 3.19, I extracted the "CPU Patch" data for the lines that were missing in my S3992 bios, and then injected them into the 3992v205.ROM.

Flashed using "afudos 3992MOD.ROM /P /B /N /C /X", but no luck booting. Onboard diagnostic gets stuck at either d0 or d1, both of which are bootblock microcode errors I think. So is at least one of the added CPU lines is a no-go? Or have I gone about this the complete wrong way and this will never work?

As an aside, the Opteron 2419 will boot to POST with a BIOS that doesn't contain the correct CPU microcode, but then fails with "CPU Mismatch Detected!" and hangs. So the board itself can obviously do "something" with the CPUs.

Can anyone shed any light on the subject?

[paradigital]

Right, well I'm pretty certain that the problem is within the 1B module, and that I need to update the AGESA code to 3.5.0.0 or 3.5.3.0, in order to bring "Six-Core AMD Opteron™ Processor 24" (Istanbul) support, along with the microcode "CPU Patch" data I've already extracted and added.

No idea how to change the AGESA version in an AMI BIOS though, and module 1B appears to be read-only.

Is there anything else I should be using to update AGESA? Is this something I can flash with AFUDOS, and only update the AGESA portion of the BIOS? Do I need another version of MMTool (currently using 3.19) to edit 1B?

Anyone with any knowledge on AGESA updates would be much appreciated.

---

Bit more of an update. Found MMTool3.19Mod, which has the ability to replace the 1B module, however I'm guessing what I actually need is the AGESA code (v3.5.3.0) adding to my S3992's 1B module, and not what I did, which was replace the entire of 1B with an extracted module from an S3992-E (which already has v3.5.0.0, as of the latest BIOS update).

Doing this still resulted in the board not booting beyond d0 (bootblock microcode I think), even with known working CPUs (Quad-Core Opteron™ Processor 2356).

---

[paradigital]

Has this community died a death between December 2012 and now? Seems to be a lot of posters complaining about lack of response, both on the forums and the live chat.

I'm pretty sure my request is a simple one (from reading other threads) but no one ever seems willing to describe the process involved, instead just posting up a modded BIOS.

I'll happily do the mods myself if someone can post up a few pointers on tools and processes for swapping out AGESA code.

Come on guys I'm desperate, I've got the CPUs sat here wasting money, money that I could donate to the community (if there is anything left of it) rather than buy a new board.

[paradigital]

Seriously? No input at all?

[manisteinn]

I don't have a solution, but found a few things that may be helpful:

The SuperIO chips differ on E/non-E version (SMSC SCH430/National PC87417). It probably doesn't make a difference on CPU support though

A quote from http://www.overclock.net/t/1148324/dual-...nder-s2932

Hmm, well I've done some more looking and found that for Shanghai split-power planes are optional, while for Istanbul they're required.

The only mention of power in the manual is "Onboard dual 4-phase VRM" for both versions. No idea if there's a physical difference.

The final BIOS update that added Istanbul support indicates that AGESA 3.5.0.0 is enough

2009/06/26 S3992-E_v106 v1.06

Problem Fix Description:

-Updated AGESA code to v3.5.0.0.
-Added AMD Istanbul CPU support
-Updated AMD PowerNOW! Options

Could Istanbul support require some power-related BIOS changes?

Can someone confirm that Shanghai CPUs work?

[paradigital]

No idea if Shanghai CPUs work or not using the BIOS found in the 2nd post, as unfortunately, I only have Barcelona or Istanbul CPUs to hand.

I'll see if I can find any Shanghai CPUs to try.

[mermalion]

I have recently purchased a Tyan S3992 (non-e) and would like to get 6-core support. I'm wondering if anyone has made progress on this? It would be greatly appreciated.