Hello Charlotte++,

On Wed, Nov 23, 2016 at 2:40 AM, Charlotte Plusplus <pluspluscharlotte@gmail.com> wrote:

Edit devicetree.cb and set: register "max_mem_clock_mhz" = "666"

Good to know! Thank you for the tip, so Federico can try quick check and see if this does work as workaround for now. :-)

Federico, I have here also suggestion for you. You should try to run only one DDR memory stick, and then record the memory values (since I see from your logs that they are identical for both memories/DDRs), and then to try to hard-code them for both sticks, altogether avoiding raminit.c setup, and see if this improves your situation?!

I suppose there is one. I don't know. I want to investigate. I will certainly try again by adding the patches suggested by Kyosti. As soon as I can get the MRC blob to work, I can make some better guesses about what is going wrong (I tried so many various things already) by having some reference points.

Please, keep us in the loop. I really enjoy reading your posts. They are very technical, and very instructive. They force me to think, and recall various things I did once upon a time (referring to Y2012/Y2013).

At the moment, I want to make a first public commit for the W520, but with the RAM issues (only stable with the MCU at 666), and no native video init, and the power consumption issues, I'm not sure how helpful it will be.

YES, of course, it will be/is VERY/extremely helpful. You did excellent job, paving path to other platform yet, and leaving to the community to fix these bugs.

Yeah, the blob. I don't like blob, but I like to have the option of something that works if I need to investigate why something else is not working. Or just for an initial release.

You are talking about SNB/IVB MRC blob, I have noticed. About this blob (beneath), correct?

_______

This change was committed on Apr. 5th, 2012 into the main coreboot git repository. In fact, if you going to look into directory:

…/src/northbridge/intel/sandybridge/raminit.c you'll see the following:

http://review.coreboot.org/gitweb?p=coreboot.git;a=blob;f=src/northbridge/intel/sandybridge/raminit.c;h=696417f1615621bc9118c92b4717ce4afdcb4e5a;hb=HEAD

Please, find the line 213: void sdram_initialize(struct pei_data *pei_data)

And then, the line 245: entry = (unsigned long)cbfs_find_file("mrc.bin", 0xab);

_______

Yup, this was the MRC blob and the commit you are referring to, back then in April 5th, 2012. This was the MRC SNB BIOS blob developed (my best guess) by some BIOS CCG group for Coreboot/Google. And I also guess, this one was reverse engineered by two of Google people (Vladimir Serbinenko somehow comes to my mind). ;-)

Kyosti, my best guess is that you can find this patch/original SNB MRC blob API code, and blob itself, can you (since Charlotte++ would like to experiment with it)?

Thank you all for refreshing my old, patchy, full of black holes brain, giving me the hope that I can (much) better organize huge DATA info scattered all over my mind! :-)

Zoran