[rescue] The best 'rescue' workstation

[Mike Meredith]

On Sat, 23 Apr 2005 21:44:23 +0200, Jochen Kunz wrote:
> On Sat, 23 Apr 2005 11:08:26 -0400 (EDT)
> adh at an.bradford.ma.us (Sandwich Maker) wrote:
> 
> > the sparc risc cpu architecture - and likewise sgi's mips, dec's
> > alpha, hp's parisc, and ibm and apple's power - is much more
> > efficient than intel's x86, so they get a lot more done
> > clock-for-clock.
> Depends. To my experience / knowledge the folowing CPUs have about the
> same INT compute performance (~ 15 SPECint95):

Just to be irritating, I've used SPECint2000 and crunched some numbers
:-

sys				Cint2K	int/mhz
IBM p5-550 (POWER5 at 1650MHz)	1200	0.72
Fujitsu PP900 (SPARC64 at 1890)	1174	0.62
Sun Blade2000 (USIII at 1200)	 642	0.54
Sun V1280 (USIII at 1200)		 608	0.50
SGI O3200 (MIPS R14K at 600)	 483	0.81
HP (Itanium2 at 1000)		 807	0.81
SGI Altix 3700 (Itanium2 at 1600)	1441	0.90
Dell PE2550 (PIII at 1300)		 561	0.43
Dell PE2650 (Xeon at 3200)		1244	0.39
Sun U10 (USIIi at 333)		 133	0.40

Seems that Intel have been learning some tricks whilst persuading that
old x86 architecture to go a little faster.

> As usual: Remember that the environment of the CPU adds a lot to this.
> i.e. cache size, latency, memory and IO bandwith, ... depending on the
> type of work load.

Just look at the difference between the V1280 and the Blade2000!