[KLUG Hardware] EETimes has more on 64-bit PowerPC (and other
MDF goodies) ...
Adam Williams
hardware@kalamazoolinux.org
16 Oct 2002 12:43:11 -0400
>Intel believes its 32-bit Xeon processors with 36-bit addressing (64GB) will
>continue to be the flagship product for workstations and servers for the next
>few years. Both Intel "Xeon" P3/P4 chips (as well as all 32-bit AMD Athlon
>processors) support an extended mapping of 4 extra bits, for an additional 60GB,
>into the limited 4GB of 32-bit addressing. For those that have been around
>awhile, if this seems similiar the old "Lotus, Intel, Microsoft Extended Memory
>Services" (LIM EMS) that allowed upto an additional 15MB to be mapped into the
>20-bit addressing on 286 processors with a 64KB window in that first 1MB, you're
>right. Not surprisingly, that too also used 4 extra bits to allow upto 16x
>memory to be addressses -- but not very efficiently either.
I think they are right, and can afford to take their time deciding on
their 64 bit path. The *VAST* (I'd wager >95%) of organizations can
find their needs met with properly configured 32-bit architectures,
usually with ample capacity to spare. And for that 5% there is the
zSeries, iSeries, etc....
Sure people said that a i386 32-bit desktop would be all anyone would
ever need, but the transition to 64bit is on an entirely different
scale. Certainly not 64bit on the desktop where the volume to bring
down cost is. And even most servers are WAY below 4Gb of physical RAM
above which 64 bit starts to strut its stuff.
I install steroid enhanced servers, not because people need them, but
because management has got a yin for deploying new boxes on their pet
projects. And that seems to be the norm for most organizations I visit
or admins I talk to.
Unless you need HUGE scale or have some ferocious new application there
just isn't any reason to be troubled with the immature state of 64 bit
platforms.