I have had a look at the uCode block of the bios on my Asrock EPC621D4I-2M but it seems to be surrounded by stuff on either side so i’m not too sure how to proceed and test if it will run skylake-SP ES.
Ideally I’d like for it to have the ucodes from 50650 to 50657 so I can test different chips, but if there is some constraint, 50652 is probably the most important one.
Cheers.
@coldwove - Here you go, good luck! 
We fought BIOS updates for a while on this with Asus WS-C621E-SAGE and ES CPU’s, but finally got it working, and then also working on latest BIOS too now
If latest BIOS does not work, test with the initial release one instead, although version 2.20 does not sound too initial… 
Would be nice to have 1.00 or before
Wait! Please run spsInfoWin64.exe -verbose
And show me the end of the report, if Measured or Verified Boot is enabled on Left/FPF side, you cannot edit microcodes on these BIOS
Here is SPS V4 ME System Tools you’ll need to run above command - http://s000.tinyupload.com/index.php?fil…211391216051973
@Lost_N_BIOS Can’t run SPS Cause the board can’t train memory for the 50654 chips I have (At least that’s what the BMC tells me). Udimms only work on 6 and 8K series xeons or something like that. Finding registered sodimms is a bit of a challenge.
I did make myself an adaptor for the 1.27mm ICP Header, so i’m able to recover failed flashes and have succesfully tested that out.
@coldwove - Did you try one stick, in all slots one by one? Are you 100% sure CPU functions properly?
Ohh, OK, if you are sure you can recover, then we can go ahead and test, but if it’s all bricks, you wont know if it’s due to CPU not compatible still, or Boot Guard brick due to BIOS edits.
So, if these both fail, you wont know if it’s due to the CPU still not allowed to work, or Boot Guard, until you get another single stick of memory to test with. Grab a cheap small single stick, surely you can find on ebay cheap.
Also, due to Asrock Protection methods (new or old), you may (read will not) not be able to flash in via Instant flash
This model may not be new enough, or have the newer form of asrock protection, but due to it’s BIOS style/type, UBU removes WAY too much and edits WAY too many areas when doing it’s thing to remove protection (so not compatible for this BIOS, for that thing)
Due to this, and the protection area not even being flashed in possibly, it may need removed/passed over in programmer (ie currently not in the onboard BIOS). So if these both fail, please send me your original dump too, so I can edit that, and send back, so you can test to be sure this is not part of the issue too.
Here is four BIOS to test, the ones without FIT in name have FIT left as-is by MMTool edit, broken locations for Boot Guard stuff (does not matter for this, if Boot Guard enabled at PCH)
The ones with FIT, I fixed FIT table locations, so then Boot Guard hash is allowed to be found by UEFITool and thus Boot Guard hash mismatch - image may not boot
This warning applies to all these BIOS, due to how Boot Guard works, and where we’re editing >> Microcodes inside PEI volume, extreme cover by boot guard = RED/Yellow)
Normally we do correct FIT, but in this case, we don’t even know if Boot Guard is enabled or not, so FIT Fixed or Broken will not matter if Boot Guard enabled.
http://s000.tinyupload.com/index.php?fil…955701924518255
Also, remember to send me that original dump too, in case those protection modules do not belong flashed/programmed in, so I can edit your dump this way x2 as well, as that should be tested too.
In the end, it may just be your memory is not compatible is the only issue, 50654 is already in the stock BIOS as it is, so those should run provided they are not blocked by some other means.
@Lost_N_BIOS Thanks for that. Yeah I’m fairly sure (hope) the 50654 (3106) works fine, certainly works on the sage, and Asrock themselves say that bronze chips only work with the unobtainable ECC REG Sodimms.
Anyways I quickly chucked the 2.20FIT bios onto the board and with a 50650 A1 step chip the BMC now is actually alive and tells me that no mem is detected which is perfect cause I haven’t put any in, i need to go find my sodimm tray, idk where it went. But at least superficially it is working better than before. I suppose it may still die on 79h DXE initalisation but I’ll find out probably tomorrow. (Assuming one of the chips will work with normal ram).
Just a side note, the asrock BMC implementation on these boards is great, even before post they tell you what the board is doing and so on.
Bios dump is here incase you want to go through that: https://drive.google.com/file/d/1uCpXoSA…iew?usp=sharing
So as I understand MMTool takes care of moving the bios around to make ucode fit, but doesn’t fix the FIT table? I’ve always moved ucode around by hand using HxD so don’t particularly have any experience doing it with mmtool.
Cheers.
----------------
@Lost_N_BIOS Alrighty so 50652 is all good, 50651 and 50650 hang on EE at some point in post mem pei, just before post code 4F, same results for both bioses. From memory the X299 dark used to do a similar thing with 50652 chips on the earlier bioses, now it just gives DXE errors.
Is there a known fix for this kind of error?
Regardless, thanks a lot for the bios c:
Cheers.
@coldwove - You’re welcome! Also, good to know you have SAGE and have tested CPU’s on that which you plan to play with here, at least that way you know the CPU’s are OK and not just DOA causing these issues.
MMTool does the ucode inserts OK, I remove all microcodes except one, then insert one I want first, then remove that last old one, then insert rest I want in there.
But yeah, it does not fo FIT correctly in several ways on this BIOS, it messes up the layout, and messes up the locations of the Boot Guard info at the bottom.
You can compare the ones I fixed (and similar to stock) vs the ones I did not fix, to see what MMTool does to FIT on this BIOS edit. Compare side by side with UEFITool 57 NE, or 51 (does not show one item, but still OK to see what I mean)
Similar edit can be done by hand with UEFITool 25 (insert) / 51 (Extract body) and hex editor, if you wanted, but you’d then also still have to Fix FIT manually too. I wrote a guide about that a while back, I need to update it badly now!
Now, you do not have to do all the math, you can use UEFITool 57 to get the correct microcode starting offsets (address) via direct click on the first one that show up and the get correct address to fix FIT by selecting the ones below it one by one
If none show up, check side by side with stock BIOS, to make sure you go to the correct microcode volume that is used, then on the mod one go to that area and expand the microcodes to see each ones address to fix FIT
Make sure you leave footer, or add it back once done, in the microcode body.
[GUIDE] Update CPU Microcode + Fix FIT Using UEFITool / Hex
So your EE hang, is that with 2.20 Fixed FIT and non-fixed, or did you mean 2.20/2.30 fixed FIT only you tested? I really only sent you the non-fixed FIT ones purely for testing purposes, just in case.
If BIOS does not = brick period, due to Boot guard enabled, then the fixed FIT ones should be what you use. As for the error, sorry, no, I don’t know what would fix that other than possibly getting an earlier BIOS than 2.20 (probably need SEC Core from earlier BIOS, and or other modules)
See if you can find 1.0 or earlier BIOS in google, if you do let me know and I will edit for you. Meantime, also try 2.30 mod BIOS if you have not yet, maybe it will work for 50650/50651?
So, how about 50653, tested yet? Do you think maybe the 50651/2 issue could be related to not having that certain memory, like 50654?
Thanks for the Info.
I was indeed using the 2.20FIT and 2.30FIT bioses you made. I don’t think this board has earlier bioses as it launched quite late, I believe CCL was already out when it did.
Haven’t tested 50653 as I don’t have any here right now, but as far as I know, they’re custom amazon skus and should in general work fine if ucode is present.
I don’t think the issues with 50650 / 50651 are related to memory as i’d expect the board to hang up before it gets that far into post, based on my experience with Skylake-SP it certainly looks like it’s bios related. I might have a go swapping a few modules around from other asrock 3647 boards to see if that goes anywhere.
I’ll report back if I solve anything.
@coldwove - Thanks for confirmation on which BIOS you tried
Due to the number/name of BIOS alone, there has to be earlier BIOS for this model, but maybe not in public, or at least maybe never linked on their page, but possibly someone dumped or posted package somewhere. It would be nice to find.
Maybe if you send them email and ask, they will either send you earlier than public BIOS by request, or make you special BIOS to allow 50650-50654 directly compatible? Never hurst to ask
SEC Core is what helped on SAGE for a bit, and on some other SuperMicro’s that I’ve seen with similar issues too.
But, I’m not sure if that can be swapped in from other BIOS, especially not without editing it first. And due to it’s PEI volume location, ideal method to do this kind of edit would be either straight hex swap on entire BIOS, or insert with 25.0, then extract from saved outfile, then straight hex into stock BIOS target
I got a 50653 to test on this exact motherboard. If the links here could be revived, it will be great.
I’m desperate to get this motherboard working for my CPU. I had been digging around to try to modify the bios ROM on my own, but it seems impossible without first having a CPU that can actually boot.
Its user private links and they r been away for a long, not possible to recover links.
Try address to the user coldwove with: @username