H11DSI rev 1.01 => biod mod rev 2.0 issue

[kramer]

Hello,
I bought a H11DSI unfortunaly a rev 1.01 so I tried to made it a rev 2.0.
For this, I put the rev 2.0 microcode for Rome CPU instead of the Milan microcode, but the motherboard still don't boot (with 2 7742)

The original BIOS (v1.0c) had this microcode (infos from MCE)

║ # │ CPUID │ Revision │ Date │ Size │ Offset │ Last ║
║ 1 │ 00800F12 │ 08001213 │ 2017-12-05 │ 0xC80 │ 0x26F500 │ No ║
║ 2 │ 00800F11 │ 08001129 │ 2017-07-14 │ 0xC80 │ 0x270200 │ No ║

I put these microcode (4kb each) with an HexaEditor instead of the original ones (it's not the same offsets because it's a v1.3 version)

║ # │ CPUID │ Revision │ Date │ Size │ Offset │ Last ║
║ 1 │ 00830F00 │ 08300027 │ 2019-04-01 │ 0xC80 │ 0x26FE00 │ Yes ║
║ 2 │ 00830F10 │ 0830107C │ 2023-12-18 │ 0xC80 │ 0x270B00 │ Yes ║

I tried the customs bios from https://forums.servethehome.com/index.php?threads/finally-overclocking-epyc-rome-es.28111/ but same result.

If someone has an idea... I must put tomorrow an original bios again and try the motherboard with a 7001 epyc to be sure everything works.

In fact, it always starts with a 7001 series, so my microcode changes are no good.

BIOS Vendor: American Megatrends Inc.
Version: 1.3
Release: 06/25/2019

CPU1 Processor: AMD EPYC 7261 8-Core Processor
Speed: 2500 MHz

P1-DIMME1
Max Capable Speed: 3200 MHz
Operating Speed: 2667 MHz
Size: 16384 MiB

@RolloZ170 It looks like you've already done it, can you help me please ?

And I'm using a programmer to flash the bios because IPMI is already in 1.52.21 and says "BIOS can't pass the secure validation." even with an original BIOS.

 

[kramer]

In the bios it sayd Microcode Not loaded for this 7261 so I think, changing the microcode isn't enough to make a 7002 series works on a rev 1.01 ?

 

[kramer]

I think I found what's wrong, I can't just change the microcode part of the bios. I need to take an original 32Mb rev 2.0 and delete the first half (the Napples part)
And the second part, after save it in a 16Mb file has :

║ # │ CPUID │ Revision │ Date │ Size │ Offset │ Last ║
║ 1 │ 00830F10 │ 0830107C │ 2023-12-18 │ 0xC80 │ 0x4F1B00 │ Yes ║
║ 2 │ 00830F00 │ 08300027 │ 2019-04-01 │ 0xC80 │ 0x4F2800 │ Yes ║

Let's flash it

 

[kramer]

I tried in UEFI now that I have a screen but "ROM file ROMID is not compatible with existing BIOS ROMID" ^^
I'm gonna retry with /X

Edit: With UEFI, I'm only able to flash with an original BIOS, even if the ROMID is the same in my "R2 16MB custom BIOS" (check with /U)
And with the original BIOS v1.3 I see the 7261"s microcode in bios (8001230)

So let's try what RolloZ170 sayd, flashing the ROME part of the rev 2.0 bios with my programmer https://forums.servethehome.com/ind...0-versus-rev-2-0-comparison.40041/post-376736 (now that I have understand it's not only a microcode change)

 

[kramer]

After flashing with the ROME part, it still doesn't work. I used the v3.0 bios version for the Rev 2.0 motherboard, should I try with a v2.x version ?

 

[RolloZ170]

(now that I have understand it's not only a microcode change)

you missed the AGESA.

i tried similar with a H11DSU-iN with no luck.
on "some" boards are additional changes required, i don't know what.

 

[RolloZ170]

i would try now a 32mb chip. why?
one the dual BIOS in any case the reset vector points to the first, the code checks the CPU gen and jumps to the correct BIOS part.
maybe the Rome part doesn't have this code anymore, means the 2nd 16mb of the BIOS don't work.

 

[kramer]

i would try now a 32mb chip. why?
one the dual BIOS in any case the reset vector points to the first, the code checks the CPU gen and jumps to the correct BIOS part.
maybe the Rome part doesn't have this code anymore, means the 2nd 16mb of the BIOS don't work.

Thank you for your feedback.So I'm going to order a 32Mb chip and test it out. (It's maybe the only difference on all these H11DSI on ebay with a rev 2 sticker on the 1.0)
To be continued ...

 

[RolloZ170]

(It's maybe the only difference on all these H11DSI on ebay with a rev 2 sticker on the 1.0)

definitely not.
supermicro itself makes this way with stickers....
heared about two SMD resistors had to bechanged, sorry i don't know.

 

[RageBone]

I am not aware of those resistors you mentioned.
I didn't have any issues running Rome on H11 boards in rev one.

AMD uses 16MB uefi images on epyc with each generation having its own image.

To make a Rev1 work with Rome, it only needs the correct (upper 16MB) image from the rev2.

What are you trying to achieve modifying things?

 

[RolloZ170]

I am not aware of those resistors you mentioned.

info from a sm technician.

I didn't have any issues running Rome on H11 boards in rev one.

thats the point, mostly it works, but not all the time.

AMD uses 16MB uefi images on epyc with each generation having its own image.

and who switches ?

To make a Rev1 work with Rome, it only needs the correct (upper 16MB) image from the rev2.

i know, was not working with a h11dsu-in.

 

[kramer]

I am not aware of those resistors you mentioned.
I didn't have any issues running Rome on H11 boards in rev one.

AMD uses 16MB uefi images on epyc with each generation having its own image.

To make a Rev1 work with Rome, it only needs the correct (upper 16MB) image from the rev2.

What are you trying to achieve modifying things?

Thx, so I will try again with a v2.x REV2 bios (of course only the ROME 16MB part)

 

[RageBone]

info from a sm technician.

I am sorry but on the interwebs i have heard many "technicians" claim many different things.
So without further elaboration, ill put this down as speculation.

thats the point, mostly it works, but not all the time.
and who switches ?

"Switching" is done by the On chip Bootloader in Rome and Milan.
The Embedded Firmware Structure "EFS" has bits indicating generational compatibility.

The PSP turns on, executes the OnChipBL who will fetch the EFS and decide from there.

The only CPUs that i know of that do not check those bits are Naples because it is a later feature and early* Rome ES + maybe the odd OEM model.

I am sad to hear that you had issues on the H11DPU, first time i am hearing of such issues.
Can you elaborate how those issues looked like?

 

[RolloZ170]

I am sad to hear that you had issues on the H11DPU, first time i am hearing of such issues.
Can you elaborate how those issues looked like?

tried 16mb and 32mb BIOS, Naples work with 32mb dual BIOS but rome not, no matter if 16 or 32mb.
BIOS and BMC chips had sockets, maybe some kid of different board.

 

[kramer]

My CH341A is dead, I did a new one...

 

[RageBone]

My CH341A is dead, I did a new one...

you should be able to flash through the ipmi.
No external flasher needed

 

[RageBone]

tried 16mb and 32mb BIOS, Naples work with 32mb dual BIOS but rome not, no matter if 16 or 32mb.
BIOS and BMC chips had sockets, maybe some kid of different board.

How did the Rome CPUs behave? Did you get *any* postcodes in the postcode snoop?

 

[kramer]

you should be able to flash through the ipmi.
No external flasher needed

I'd like to but it won't work even with an original bios because the ipmi firmware 1.52.21 needs a signed file. (even with sum)

 

[RageBone]

I'd like to but it won't work even with an original bios because the ipmi firmware 1.52.21 needs a signed file. (even with sum)

H11DSi?
Must be a recent addition then.
I remember that it didn't care before.
Downgrade the ipmi?

 

[RolloZ170]

How did the Rome CPUs behave? Did you get *any* postcodes in the postcode snoop?

no life, the board turns on, not more. code FF
its over, the board is gone.