[PlanetCCRMA] 64bit vs 32bit

[Henrik Frisk]

>> A recent test I read proved that a 64 bit (Linux) system has better
>> performance than a 32 bit (I'll dig it up if someones interested)
>
> It'd be nice to see it. Is the test relevant to the workflow you have?
>
> How big is the improvement and under which conditions?
>
First of all, apologies for the slow response on this...

WRT the test, it was published in a serious Swedish computer magazine
in December 2009. The tests were done on a Intel Core i7-975 under
Windows and Linux Mint and the tests were performed with Peacekeeper
(http://service.futuremark.com/peacekeeper/), DBench and LBench
(http://www.bitmover.com/lmbench/whatis_lmbench.html). Tests were done
on typical browser activity (rendering, graphics, DOM operations, text
parsing, etc) as well as on more low level operations such as
integer/float/double calculations measuring caching speed and latency
in parallel calculations.

IMHO the results are not very clear. The 64 bit system perform
slightly better in general but some tests are better in the 32-bit
system (the 32-bit system is remarkably more efficient at doing
parallel calculations (add/mul/div) on floats and doubles). All in
all, the author seems to draw the conclusion that a 64-bit system is
faster (i.e. that the 64-bit system he tested is faster). Revisiting
the article and looking at the test results again I'm not so sure.

best,

/Henrik