[Beowulf] Sidebar: Vista Rant

Robert G. Brown rgb at phy.duke.edu
Tue Jul 17 08:31:14 PDT 2007

On Mon, 16 Jul 2007, Bruno Coutinho wrote:

>> >Consumers will forgive a lot, but not poor interactive performance.
>> >That's why Linus has made excellent interactive performance a design
>> >mandate from the very earliest days of the kernel (and why linux plus X
>> >on 486's was peppier -- much peppier -- than Vista on multi GHz multi
>> >cores).
> When machines start coming with 8-16GB RAM this will be forgotten.
> More than ten year ago everyone said that NT was bloated, but when
> machines went with 128MB RAM, everyone switched to 2K, XP.

I don't believe that Vista's slowness has anything to do with hardware
memory footprint.  My experience is limited to a dual core 1.6 GHz CPU
with 2 GB of RAM, running "nothing" but the OS itself and a single
bo-ring application (e.g. a game) that runs under XP FAST in 512 MB (and
runs perfectly acceptably in half of that).  Vista is configured with
transparency off and all that -- obviously computing translucent windows
is a COSMIC and STUPID waste of cycles.  In fact, it is configured for
maximum performance (it has a lovely little configuration interface that
lets you do it with a single mouse click, basically).

Slow, slow, slow.  To call it a pig does disservice to pigs.  Move the
mouse and a few seconds later the mouse pointer moves.  Type into the
keyboard and 0.5 seconds later (with irregular luck) characters appear.

I have no idea "what it's doing in there", but with two CPU cores with
an aggregate 3.2 GHz of cycles to dispose of and 1 GB of high speed RAM
apiece, to be unable to manage what amounts to a single threaded
foreground task (on say one CPU) and the thumb-twiddling of kernel
idleness (on say the other CPU) isn't something that will be solved by
16 GB of RAM.  At least not until the software itself is fixed.  This is
just plain broken behavior.

The real question is -- is it broken by design -- something that will be
VERY expensive for them to fix, as it basically back to the drawing
board to start over again, and as I alluded to in a previous note, they
might have to go ALL THE WAY BACK and re-engineer Windows itself from
the ground up if they've encountered a scaling limitation in their
generic design.  Or it may be a matter of retuning, fixing bugs, falling
back to XP and relabelling a working but much more humble advance on XP
"Vista" on an emergency basis.  Those things would be expensive too and
of course the whole situation is already embarrassing, and embarrassment
is expensive in its own right to a company that relies on the confidence
of its customers that however painful and expensive Windows might be, it
can be made to WORK!  It isn't clear that Vista can be made to "work",
no matter what.

> But Linux plus X on a 486 with a 640x480 VGA is a very different
>> matter from digital rights managed 1600x1200 HD on that
>> multicore.  It's not really an apples/apples comparison.  You have to
>> burn a lot of compute cycles to make sure that your content is
>> legitimately being viewed <grin>.
> Really this is a very unfair comparison. :-)
> I read that Vista multimedia system has 7 modules:
> 2 to playing multimedia content and 5 to ensuring DRM.

I didn't even mention Apples, because they are a different OS and issue

My point was that to a computer user -- even a sophisticated and
somewhat jaded one like me -- the "speed" of a computer has nothing to
do with its clock, the amount of memory it has, the quality of its
network, the vast ocean of its disk.  It is measured in very simple
terms.  When I type, do characters appear "instantly" or after a lag?
When I move the mouse, does the pointer or view scroll smoothly, or
jerkily and after a lag?  When I click an execution box, or run a
program from the command line, does the process execute "now" so that I
can see the new process hop up on the screen, or is there a 2-3 second
delay?  When I load a file to work on it, save a file to disk, or do
ANYTHING AT ALL does it happen NOW or does it happen LATER, after an
annoying interruption while the system "thinks"?

SunOS (4.x, say) was amazing in the way it would make interactive use of
its GUIs (sunview, X11R4 and up) happen NOW.  Solaris was amazing in the
way that it would make interactive use of its GUIs on much more powerful
hardware happen LATER.  Linux on 486's and beyond was amazing in that it
was pretty much as good as SunOS 4.x at making things happen NOW, a
property that has been carefully preserved in later version of Linux by
deliberate design.  As long as the system has enough memory not to swap
a linux box works NOW.

This is where Vista is a major screw up.  Nobody cares about DRM
compliance but the DRM police.  Nobody cares about multimedia systems,
and since functioning multimedia now fits in $500 telephones people are
understandably cynical about needing multicore multi GB systems to make
it work.  What people DO care about, very much, is having a system that
happens NOW, click to keyboard, at their actual UI.  Microsoft is in
deep trouble...


Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu

More information about the Beowulf mailing list