[geeks] value of PIII PC servers

[nate at portents.com]

> I've actually considered installing XP on my own personal game machine,
> for one and only one reason:  To get the 64-bit kernel.  Not that I have
> any games that need 64-bit, but when I built it, an Athlon64 made sense,
> but I've since learned that a 32-bit Windows kernel on Athlon64 hardware
> is actually slightly slower than a regular Athlon of the same clock
> speed running the same 32-bit kernel would be.  The 64-bit kernel on
> Athlon64 is supposedly significantly faster.

Well, in theory.  x86 64-bit (aka x64) doubles the register count, and
makes SSE2 mandatory, dropping the old x86 floating point altogether.  So
code compiled for x64 can take advantage of a more sane number of
registers and SSE2, which isn't bad SIMD.

The overhead of running 32-bit code in the 64-bit enviroment isn't that
high, thanks to the processor design where the mode switch isn't that
costlyand the WoW64 thunking layer that handles function-call argument
translation and interprocess communication.  However I wouldn't say from
experience that it is 'signficantly faster', and I've seen some odd
behavious in person which I haven't seen reported elsewhere.

First of all, and perhaps this is a side-effect of running a 32-bit dev
environment and debugger in Windows x64, but the GUI is not what I would
call consistely responsive when running 32-bit apps alongside 64-bit apps.
 Also, the shell is 64-bit, which means that most shell extensions won't
work (because most people are still writing 32-bit shell extensions). 
Thus, things like TortiseSVM, WinZip, etc. shell integration just won't
show up (at least, there weren't 64-bit versions when I last tried).

> I've also wondered whether it might get around the problem of being
> unable to use many games except using a No-CD patch (and unable to use
> any game at all for which no No-CD patch exists) because the copy
> protection mistakenly thinks the machine, which has SATA disk and PATA
> DVD writer, has no optical drive installed at all.

I'm confused what is causing your system to be "unable to use many games
except using a No-CD patch", unless you are using a SCSI CD-ROM drive or
somesuch.  One of the more notorious game copy protection systems,
StarForce, is driver-based, and if you go to the Device Manager in the
management console and select "Show hidden devices" you will see StarForce
under the new "Non-Plug and Play Drivers" section that will appear (don't
try to remove it manually, however).  StarForce protected games have
encrypted object code and the StarForce driver (which runs in ring 0)
checks your system to make sure you don't have a debugger running, an
emulated CD/DVD drive presenting disc images as real discs, etc.  In some
instances I also believe there is an additional filter driver that gets
installed between a CD/DVD drive and the OS.  Anyway, last I heard there
wasn't even a 64-bit version of StarForce, and thus StarForce protected
games just don't work in Windows x64 *unless* you no-CD patch them.  And
FYI, a lot of demos use StarForce too (under the philosophy that
unencrypted object code in a demo exe could be used to reverse engineer a
working executable from a protected retail game more rapidly).

> (Don't even get me started on the festering LeadTek piece of shit I
> purchased under the mistaken belief that it was a usable and stable
> video card for any game more graphically demanding than, say,
> Minesweeper.)

Unless a person is willing to memorize lists of letter-number combinations
and what they actually correlate to in terms of GPU architecture, graphics
memory architecture, and how they correlate to real-world quality and
performance in both a theoretical and practical sense to both OpenGL and
DirectX, then buying a video card is the equivalent of playing slots. 
That part of the industry has kind of gone off the rails.

- Nate