64bit or 32bit for my laptop

[Wendell Nichols]

Sharpe, Sam J wrote:
> Patrick O'Callaghan wrote:
>> On Wed, 2009-03-11 at 22:40 -0600, Frank Cox wrote:
>> > On Thu, 12 Mar 2009 11:45:26 +0800
>> > Meng Qiu wrote:
>> >
>> >> The "gears" for my GeForce4 MX440 on F10 is about 1700 FPS.
>> >> But the number for my HD1950pro on F11(pre) is never up to 100 FPS!
>> > 7074 frames in 5.0 seconds = 1414.687 FPS
>> > 7103 frames in 5.0 seconds = 1420.448 FPS
>> > 7102 frames in 5.0 seconds = 1420.320 FPS
>> > 7112 frames in 5.0 seconds = 1422.319 FPS
>> >
>> > That's on Fedora 10 with an ATI X1550 card and the video driver 
>> that's included
>> > with Fedora 10.  (Not the proprietary driver.)
>>
>> For comparison:
>>
>> 6799 frames in 5.0 seconds = 1359.743 FPS
>> 6961 frames in 5.0 seconds = 1392.055 FPS
>> 7015 frames in 5.0 seconds = 1402.842 FPS
>> 6660 frames in 5.0 seconds = 1331.944 FPS
>> 7034 frames in 5.0 seconds = 1406.609 FPS
>> 6960 frames in 5.0 seconds = 1391.793 FPS
>>
>> That's on an Intel mobo with a 2GHz Core 2 Duo and 4GB of RAM, using the
>> *onboard* 82Q963/Q965 graphics chip. So either glxgears isn't a very
>> representative benchmark, or installing one of these graphics cards is a
>> waste of time. I'm not sure which.
> Or it's badly setup, because my low-end NVidia Cards beat that easily. 
> (I have a pair of Quadro NVS 290 PCIe 64MB - admittedly using genuine 
> NVidia drivers)
>
> [sam at sam ~]$ glxgears
> 9648 frames in 5.0 seconds = 1929.527 FPS
> 9566 frames in 5.0 seconds = 1913.107 FPS
> 9679 frames in 5.0 seconds = 1935.721 FPS
> 9621 frames in 5.0 seconds = 1924.059 FPS
>
> FWIW, I see a definite difference between the card running 2 x 
> 1280x1024 monitors vs. the card running 1920x1200 + 1280x1024 - the 
> latter consistently scores in the 1700-1800s, the former (above) 
> scores in the 1900s. Each card is running NVidia's TwinView, but I 
> don't have Xinerama running to bind the  two sets of screens together.
>
> (note to others - don't try running Compiz on this setup, it really 
> really can't handle a 5760 x 1200 array of pixels yet)
>
> -- 
> Sam
>
In my opinion it has almost nothing to do with performance!  The actual 
differences are just too small to measure without a good ruler.  The 
real issues revolve around what software is available in 64 bit mode and 
how the lack of 64 bit versions affects you.
The main problem is the lack (for the moment) of a generally available 
64 bit java plugin for firefox.  This has just in the last few weeks 
been made available but because it was unavailable for so long many 
vendors didn't produce 64bit native code to go with their applets for 
linux.  Two that seriously affect me are Webex, and Juniper VPN.  These 
things both come with native jni code used by their applets so I have to 
run 32 fit firefox on my 64 bit laptop.  I also have to keep a 32bit jre 
around for its plugin and manage the link to the dll manually cause all 
my other java work is with the 64 bit jvm/jre.
So its a pita to manage 32bit firefox on a 64bit machine but its necessary.
The rest of the 32 bit world runs more or less invisibly on my 64 bit 
system and I'm quite pleased with how seamless it is.

If you compile software which is 32bit, you also have to make sure that 
you install both 64 and 32 bit versions of all the libraries and 
development packages.  The compiler is pretty good at managing the rest!
enjoy...
wcn