Microkernels (WAS Re: New to the list)

[Zane H. Healy]

At 10:12 AM -0700 8/5/06, Al Kossow wrote:
>  > Current Macs share much more with PeeCees
>  > than SGI IRISes and Sun SPARCs, so there is most likely a non-insignificant
>  > amount of suboptimal hardware that eats processor cycles
>
>Please give concrete examples of this "suboptimal hardware" you speculate
>about, and give examples of how this is done better on SGI or Sun DESKTOP
>CLASS Hardware.

You know, I was about to comment on that one myself.  I personally 
consider my PowerMac G5 dual 2Ghz to be very closely related to an 
SGI Octane, but an improvement.  It is most definitely a better 
design than my SGI o2's.  I'd say the design is better than any of my 
better PCI-based Sun's (Ultra 60 and SunBlade 1000).  Construction 
wise, my G5 beats any of them.  The only place it is likely to loose 
out, is in Graphics.

Unfortunately this does not include the OS.  Solaris and IRIX don't 
hang the way that Mac OS X does.  I've got so far as to try to get a 
IRIX system with very low spec's to keel over, and the best I could 
do was slow it down significantly but it was still very responsive.

Of course I am talking about a Professional Mac, and not a consumer 
grade Mac.  I must confess that I'm not that impressed with the 
consumer grade models.  But then the closest I've ever been to a 
consumer grade Mac was my PowerBook 520c.

Also that "suboptimal hardware" is used in Sun hardware as well. 
Take a look at all the cards on Workstations and you'll see that 
they're basically PC cards.  It is things like the chipsets that make 
or break a computer.  For example, last year we got a new server in 
that was purchased for our team by another group without talking to 
us.  It was the most pathetic piece of **** I've ever seen come out 
of Sun.  It totally crumbled under the I/O load that it was expected 
to handle, and could only handle 1/16th of the load it was expected 
to handle.  It would have made a nice workstation (if it had a UPA 
slot), but it simply couldn't handle receiving data on one PCI card 
and sending it out another.  It wasn't the PCI cards that were the 
bottleneck, it was the way the system was designed.

For truly well designed hardware take a look at some of the IBM 
systems, they use PCI as well, but WOW!  Also take a look at 
something like an HP DL380 G4, for a fairly nice Xeon-based system, 
though the onboard RAID stinks.

			Zane


-- 
| Zane H. Healy                    | UNIX Systems Administrator |
| healyzh at aracnet.com (primary)    | OpenVMS Enthusiast         |
| MONK::HEALYZH (DECnet)           | Classic Computer Collector |
+----------------------------------+----------------------------+
|     Empire of the Petal Throne and Traveller Role Playing,    |
|          PDP-10 Emulation and Zane's Computer Museum.         |
|                http://www.aracnet.com/~healyzh/               |