[coreboot] The vendors have caught up with boot times
c-d.hailfinger.devel.2006 at gmx.net
Sat Sep 26 03:01:55 CEST 2009
On 25.09.2009 18:21, ron minnich wrote:
> too bad their software is still closed and a security risk but hey ...
They do hardcode everything (SPD, PCI, HDD, ...) and rely on disks which
are already spinning and have fast link negotiation (which can only be
achieved if the machine is already on and you hit reset). Add some RAM
and the whole thing falls apart (you can't even boot anymore unless you
trigger a fallback mechansim which is not present yet AFAIK). Add a
PCI/PCIe/CardBus/Expresscard card and the whole thing falls apart.
Needless to say that this is focused on laptops where hardware seldom
changes. Ah, and of course they removed all BIOS compatibility to save
some time. Some people also reported that the fast Windows boot time was
achieved by rebooting Windows on shutdown, then going to Suspend-to-disk
with the freshly booted Windows to get an impressively fast "boot" speed
Essentially, this is a trick. We are faster even without these tricks.
And if they ever claim these numbers are not tricks, we can say that we
spend 0 ms inside EFI and 0 ms inside BIOS (in the
FILO/GRUB2/Linux-as-payload case) and it should be hard for them to beat
our perfect zero timing.
P.S.: I foresee claims about negative boot times once someone figures
out that resume is faster than regular boot and claims that resume is
sort of a special boot sequence.
More information about the coreboot