On Tue, Nov 22, 2016 at 9:30 PM, Charlotte Plusplus < firstname.lastname@example.org> wrote:
I have had similar issues with Corsair ram on the W520 recently: sometimes not booting at all, sometimes being unstable (in memest) after a succesfull raminit.
The only way I could get the 4 dimms to work was to hardcode some SPDs, or set the MCU to a much slower speed.
Like you, I found removing even 1 stick did help a lot: the raminit succeeded much more frequently as a higher MCU, even if this MCU still lower than the one the RAM is rated for, or that worked in the factory bios.
I tried using the MRC blob to compare the timings, but I must have done something wrong in my code as it didn't work at all
See following commits: 5c10abe nb/intel/sandybridge: increase MMCONF_BASE_ADDRESS 0306e6a intel/sandybridge: Fix builds with System Agent blob f9c4197 northbridge/sandybridge/raminit_mrc.c: fix missing include
My guess is something is really wrong in the raminit code. I read up too
much specs and code for no result, so I just gave up on this.
Hopefully more people getting similar problems will mean the MRC will be added back as an option next to the native raminit. This would facilite comparison on all boards, and identification of whatever bug there may be. (imagine figuring out native video init issues if there was no way to use a VGA option rom)
Was MRC blob ever really an option for lenovo/xxx boards? I did not check history.
Anyways, currently I have no trouble on samsung/lumpy changing between MRC and native raminits. And with native raminit it should now boot much faster than two weeks ago.