AMD ES ZS1406E2VJUG5 microcode version?

ExecutableFix

Active Member
Nov 25, 2019
114
45
28
Hey all,

I recently got my hands on this EPYC Rome engineering sample, according to the previous owner it was fully working on an HP ProLiant server. The physical appearance of the cpu is as expected, no capacitors (or transistors) have been damaged.

I have a rev 2 supermicro H11SSL-i, which supports Rome cpu's, but I can't get the thing to boot. I've already tried flashing a bios with updated microcode for CPUID's 00800F12, 00830F10 but no luck so far. Does anyone know the CPUID or stepping of this engineering sample? Or perhaps any other method to get this CPU to post/boot?
 

nero243

Active Member
Oct 28, 2018
108
76
28
Can you ask the seller which CPUID the ES has? What kind of server exactly was the CPU running in?

I've got to say this doesn't look good, i was always able to post to bios even with no Microcode loaded on Naples.

Edit:
The only MC related difference i found between a H12DSTB (which this ES was seen running on) and your board is that the H12 has the MC 27 120008 and 30 120008 and your board only has the v30.
 
Last edited:
  • Like
Reactions: Tha_14 and RageBone

ExecutableFix

Active Member
Nov 25, 2019
114
45
28
Can you ask the seller which CPUID the ES has? What kind of server exactly was the CPU running in?

I've got to say this doesn't look good, i was always able to post to bios even with no Microcode loaded on Naples.

Edit:
The only MC related difference i found between a H12DSTB (which this ES was seen running on) and your board is that the H12 has the MC 27 120008 and 30 120008 and your board only has the v30.
He doesn't know the CPUID as he's claiming he's not in office at the moment, but he did say they got it to boot fine in a HP ProLiant Gen10. My naples ES works fine in the H11SSL-i, even if it's rev2, but that could be because all the supermicro boards use microcode version 08001129 for CPUID 00800F11.

The naples engineering samples don't have an epyc stepping, so maybe this one doesn't either?

EDIT: What even are the symptoms of missing CPU microcode? Currently the system is powering on, it's just not booting.
EDIT2: Just checked the POST code via IPMI, FF. Probably an F for me?
 
Last edited:

nero243

Active Member
Oct 28, 2018
108
76
28
Searching for an server bios like th HP proliant or Dell R7515 is a dead end, i'd need a direct dump from chip to check for differences.

Theres not a whole lot of information out there of what AMDs Microcode is even doing. On my H11SSLi + Naples ES i could post to bios without any microcode and with newer versions than 29, i Just couldn't boot to windows.

There are only 2 MCs in the Supermicro bios as far as i know and those are for CPUID F11 and F12.

I think theres more required than just the correct microcode to make these ES work. You could try replacing the 30 120008 in your bios with version 27.
 

ExecutableFix

Active Member
Nov 25, 2019
114
45
28
Searching for an server bios like th HP proliant or Dell R7515 is a dead end, i'd need a direct dump from chip to check for differences.

Theres not a whole lot of information out there of what AMDs Microcode is even doing. On my H11SSLi + Naples ES i could post to bios without any microcode and with newer versions than 29, i Just couldn't boot to windows.

There are only 2 MCs in the Supermicro bios as far as i know and those are for CPUID F11 and F12.

I think theres more required than just the correct microcode to make these ES work. You could try replacing the 30 120008 in your bios with version 27.
I was already looking into replacing that microcode, but I don't know how. The UBU tool for AMD can only update to newer versions and playing around with the versionname/date doesn't help. Any suggestions?
 

nero243

Active Member
Oct 28, 2018
108
76
28
I PMed Patriot about a similar problem with A1 Naples ES CPUs

I could replace the the microcode from the retail CPU with the one for the ES CPU.

Rebuilding the bios with ubu fails when i change the microcode so i'm doing everything with hexedit by hand. Worked for me when i tried the newer version 37 Microcode.

All the epyc and tr microcodes have a length of 3199, so what you do is open the whole .rom in hexedit and search for 120008 (for retail mc) or 110008 (for es). Than mark everything from the start of the mc to 3199 postions further. Check if the end is correct (after the MC theres a few lines of FF FF FF) and insert the the other MC.

This only works if the MCs are the same size!

Here you go:
supermicrodual mod

I also included the MC extracted from my bios incase you want to do something on your own. The bios has now no retail CPU MC but the ES MC from my board.

I also have to say, that my CPU posts without any MC loaded, but fails to boot windows liek that. So take your chances but don't have to high hopes on this.
You have to extract the 27 MC from the H12DSt-B bios, search 27120008 in ubu, extract the padfile and grab it from there.

edit:
If theres a size difference just add or delete a few FF FFs
 

ExecutableFix

Active Member
Nov 25, 2019
114
45
28
I PMed Patriot about a similar problem with A1 Naples ES CPUs
Awesome, hopefully he can help or knows something we don't

You have to extract the 27 MC from the H12DSt-B bios, search 27120008 in ubu, extract the padfile and grab it from there.

edit:
If theres a size difference just add or delete a few FF FFs
I got the 27 MC already. Sorry, I might sound dumb, but how do I extract the padfile from UBU? I've read through multiple tutorials on how to edit the microcode, but none have really helped so far. All seem to be for Intel cpu's

EDIT: Big NVM, just did it. I'll try to flash this bios now and see if it works
 
Last edited:

nero243

Active Member
Oct 28, 2018
108
76
28
I think that never worked for him.

If you have the MC 27, you don't have to extract the padfile. But for educational purposes:
Open the Rom in UEFITool, search hex 27120008, click the first result, right click on the padfile extract as-is.
 

ExecutableFix

Active Member
Nov 25, 2019
114
45
28
I think that never worked for him.

If you have the MC 27, you don't have to extract the padfile. But for educational purposes:
Open the Rom in UEFITool, search hex 27120008, click the first result, right click on the padfile extract as-is.
Just updated the bios, unfortunately no luck. Still doesn't boot and the postcode is still FF.
 

nero243

Active Member
Oct 28, 2018
108
76
28
That was kind of what i expected, i don't think it's the microcode that keeps you from posting.

You could try to flash a H12 Bios and look if you're getting any further than FF, which is highly experimental and by no means usable. If that works i think you could get the CPU working with bios modding, but i don't know what exactly has to be changed.
 

ExecutableFix

Active Member
Nov 25, 2019
114
45
28
That was kind of what i expected, i don't think it's the microcode that keeps you from posting.

You could try to flash a H12 Bios and look if you're getting any further than FF, which is highly experimental and by no means usable. If that works i think you could get the CPU working with bios modding, but i don't know what exactly has to be changed.
That could be a good idea, because of IPMI there's no way to brick the board by flashing a different bios (or is there?). I could give it a shot
 

nero243

Active Member
Oct 28, 2018
108
76
28
I forgot about that, i usually use a flasher. I think i found that somewhere before, but i need to look for it again myself. The problem is if your changing the id and it flashes you might not be able to flash back via ipmi.
 

ExecutableFix

Active Member
Nov 25, 2019
114
45
28
I forgot about that, i usually use a flasher. I think i found that somewhere before, but i need to look for it again myself. The problem is if your changing the id and it flashes you might not be able to flash back via ipmi.
Hmm, that might be a bit too risky then. How would that brick the IPMI exactly? Isn't that a standalone controller?
 

nero243

Active Member
Oct 28, 2018
108
76
28
You don't brick the controller, but it could just think your board is not compatible with the bios you want to flash and won't let you.
 

ExecutableFix

Active Member
Nov 25, 2019
114
45
28
You don't brick the controller, but it could just think your board is not compatible with the bios you want to flash and won't let you.
Hmm, that could be right, but couldn't you just change the version of that bios as well? Idk if it's worth it, I mean maybe it'll brick the lan and then you can't even access the IPMI.

Maybe someone else has any suggestions, maybe the chip is just a dud. I'll definitely wait till the seller provides some screenshots.

EDIT: Might be worth buying a Supermicro H12SSW-iN to try and run it on there. According to the screenshots this exact chip ran on a H12SSW-NT/iN. If it doesn't work it's always possible to return the board.
 
Last edited:

nero243

Active Member
Oct 28, 2018
108
76
28
I was blind but now i can see...

All the 32MB Bios versions for Rome contain 2 Bios images 1 for naples and 1 for Rome, the first padfile belongs to naples doing anything there won't help. (UEFITool fails to display 30% of the 2nd image btw).

Your ES has the CPUID 00830F00 MC in hex (27)003008, but the microcode is present in the second image of your rom.

But this information is pretty useless since i still think it's not a mc issue...

Some ebay sellers vaguely suggest that these ES chips are compatible with H12 boards.
 

ExecutableFix

Active Member
Nov 25, 2019
114
45
28
I was blind but now i can see...

All the 32MB Bios versions for Rome contain 2 Bios images 1 for naples and 1 for Rome, the first padfile belongs to naples doing anything there won't help. (UEFITool fails to display 30% of the 2nd image btw).

Your ES has the CPUID 00830F00 MC in hex (27)003008, but the microcode is present in the second image of your rom.

But this information is pretty useless since i still think it's not a mc issue...

Some ebay sellers vaguely suggest that these ES chips are compatible with H12 boards.
I could still try to patch that microcode as well, but I've also ordered the H12 board. Will give it a shot and if that doesn't work then the chip might just be dead
 

nero243

Active Member
Oct 28, 2018
108
76
28
I was blind but now i can see...

All the 32MB Bios versions for Rome contain 2 Bios images 1 for naples and 1 for Rome, the first padfile belongs to naples doing anything there won't help. (UEFITool fails to display 30% of the 2nd image btw).
Okay i got to retract that, seems more like someone didn't bother taking the old unused stuff out. This Bios is a utter mess, maybe thats the reason why it's not posting :p

I don't think your chip is dead, pretty sure it's just a issue with the H11 Boards and Rome ES CPUs.
Verkaufen ist -! AMD epyc 7702 es 2 S 1404 E 2 vjug 5 1.4ghz 64c Sockel sp3 7nm | eBay

This seller for example says that the chips are only compatible with the H12DST-b, altough his chips are not in a good physical condition.
 
  • Like
Reactions: ExecutableFix