On Wed, 15 Mar 2017 16:17:31 -0500 Timothy Pearson firstname.lastname@example.org wrote:
On 03/15/2017 04:09 PM, Daniel Kulesz via coreboot wrote:
I wanted to try caching of the MRC training data. As described in Timothy's post, I commented the following line:
allow_config_restore = 0;
However, I was not able to measure any effects regarding boot time. Does this setting only work with cbfs enabled? And if so, is it necessary to set additionally any cbfs variables?
You will need NVRAM enabled for this setting to work. Additionally, there is some instability when this line is uncommented; when I have some free time available I'll take another look at why.
Many thanks - this did the trick! Boot time decreased from 71s to 30s now. Finally, coreboot boots up twice as fast as the vendor bios on the KGPE-D16!
@Timothy: Regarding the mentioned instability: Have you tested for its presence after we applied this "revert fix" for the MCT failures? I've been running "memtester 10G" + "stress-ng --cpu 0" for around 20 minutes now but could not observe any failures, instabilities or dmesg entries so far.
@Paul: Have you tried this setting as well? If yes - would you like to share your experience?