[MINC-users] CPU Speed

[Alex Zijdenbos]

Yeah CPU speed increase definitely seems to be leveling off. There are some
hardware reasons for that of course (heat dissipation seems to be one of
the main issues,
http://electronics.stackexchange.com/questions/122050/what-limits-cpu-speed
).

But is the MINC benchmark perhaps not so valid anymore - at under 2 mins of
any MINC process, wouldn't the brunt of that time be spent on I/O, not
computation?

But yes, there's the increasing number of cores that do provide more
cycles, not counted in single-thread processes. Perhaps we should have
benchmarked with respect to $$ instead of time :-)

-- A

On Sun, Aug 9, 2015 at 10:28 PM, Andrew Janke <a.janke at gmail.com> wrote:

> Anyone elase find it somewhat annoying that things haven't really got
> all that faster in CPU's over the last few years. New desktop Xeon
> based Dell (Precision 5810) is about 14seconds faster than the 6 year
> old i7-950 it replaced. See the last line here:
>
>    https://en.wikibooks.org/wiki/MINC/Benchmarks
>
> We have more cores available for the same $$ spent, but they aren't
> getting much faster. They are however available in laptops. Either
> that or a 10% change is not impressive anymore. To put it all in
> perspective, a 2mm non-linear fit that used to take Louis about an
> hour when he wrote minctracc now takes just under 2 minutes.
>
>
> a
> _______________________________________________
> MINC-users at bic.mni.mcgill.ca
> http://www.bic.mni.mcgill.ca/mailman/listinfo/minc-users
>
>