Titanas: συνέντευξη στο BloGR

George Moschovitis on 23.05.2007 @ 12:48 am.

Τι να λέμε τώρα, είναι θεοί οι άνθρωποι. Ωστόσο, και οι αναμενόμενοι επεξεργαστές της Sun έχουν ενδιαφέρον…

Constantinos Sfyris on 23.05.2007 @ 11:59 am.

It is remarkable that IBM has increased the clockrate up to 4.7 GHz,
the promised 5GHz weren’t reached. IBM has forgotton that there are
some physical laws which couldn’t be overcome (also the IBM has it’s
limits….). The IBM hasn’t increased the speed of the main memory
in the same ration as they increased the clock rate of the CPU.
Therefore a high increase in speed can only be seen for ‘cache-resident’
algorithms. Programs with heavy date movement between main memory and
CPU do not benefit from the high clock frequency. The Sun SPARC
Enterprise server are very competitive in this area. As an add-on the
CMT technology takes care about the memory latency so that this
technology
fits much better today – and expecially in the future – for highly
parallel algorithms. IBM didn’t made a clear positioning for their
new pSeries server, the xSeries, zSeries and iSeries. The choice
is much more difficult for the IBM customer after this announcement.
The workload oriened approach of Sun gives the customer a clear position
for each Sun server. Choose the right server for the right workload is
a much better approach for the customer than a ‘gigaherzomania’ which
is proposed by the IBM.

At least it’s clear that IBM will push their benchmarketing like
crazy to win the deals – they will provide their results and will
offer much cheaper (and slower) configurations. To win in this
situations a clear positioning of the operation system would be
the best strategy:

“You date your server – you marry your operation system”

and Solaris is much better than all AIX, Linux, zOS, iOS and all
other OSes.