[GLLUG] Building a new system
Sean
picasso at madflower.com
Thu Jul 24 12:46:51 EDT 2003
On Thu, 24 Jul 2003, Keyes, Randall wrote:
> As I said...the debate continues. :)
I don't think there is much of a debate, Intel has acknowledged but not
rebuted the claims. The G5/2ghz has a max theoretical of over 20Gflps per
processor with a dual independent 1ghz front side bus to each processor.
It does 64-bit calculations and all the benchmarks that I have seen are
all 32-bit benchmarks.
The only real rebuttals are that I have seen are:
A they didn't include the Intel MMX libraries for testing purposes.
However, they didn't use Altivec enabled instructions either.
B They used GCC instead of Intel's compiler which incidently makes a great
Free Unix (linux/freebsd etc.) comparison.
C They used unix instead of Windows for testing. =) *shrugs*
> I'll stick with PC's. Big Blue and M$ help keep me employed. :)
IBM makes the PPC970 chip. It wouldn't surprise me to see some IBM spin
off projects that use this chip with both Linux and AIX for the low-mid
range server products or for their blade packages since they are fairly
low power still and cool. not to mention they can be used in 8,16-way
configurations and they should be fairly compatible with the Power4 series
software.
As far as M$ goes, running Windows only through emulation is biggest knock
on the whole thing. If what Mike said is true and RealPC delivers. I don't
think emulation will be that slow. A LOT of speed is lost because
VirtualPC uses a trident/s3 as its emulated video card. Using the native
video instructions would increase the emulation speed literally by orders
of magnitude especially for gaming.
I guess I'm not sure what there is to debate from a performance aspect
unless you want to run Windows.
-------------- next part --------------
_______________________________________________
linux-user mailing list
linux-user at egr.msu.edu
http://www.egr.msu.edu/mailman/listinfo/linux-user
More information about the linux-user
mailing list