Re: NEC-LIST: Wanted: Speed comparison PIII 550 vs PIII 550 Xeon

Keith:

Your Xeon results are consistent with the PC-NEC benchmarks testing
I've previously done and my understanding of the Xeon processor
architecture. Please note that:

1. In Table-9 of the PC NEC4.1 PERFORMANCE DATA that turning the L2
    cache on/off had no significant impact on the NEC factor time for
    the P-II 300-MHz CPU being tested. Only the fill time changed
    very slightly. Thus going from a half-clocked to a full-clocked
    L2 cache would seem to have even less effect. Of course, this
    could be different in the faster processors, but your tests would
    indicate not.

2. If my feeble memory is any good, I recall that the Xeon is
    optimized for Windows-NT multiple processor server applications.
    I seem to recall that each Xeon CPU can share access to the
    secondary cache and thus improve the performance of Windows-NT
    server (which is a slow dog to begin with and needs as much CPU
    horsepower that one can afford to throw at it).

    Now maybe someone has a compiler that will generate optimized
    parallel code for multiple Xeon CPUs and provide some application
    speed improvement. Lahey was looking at something like this some
    time ago, but I've not heard anything more of it.
    
Thus a single Xeon CPU doesn't seem to offer much over a P-III of the
same clock frequency. Multiple Xeon CPUs without an optimizing
parallel processor compiler and a suitable application also wouldn't
seem to benefit much -- which is what your testing seems to confirm.

--Larry, W7JYJ
Received on Wed Jul 21 1999 - 21:03:00 EDT