[Info-vax] RealWorldTech on Poulson

[ChrisQ]

Michael Kraemer wrote:

> 
> Well, we don't know what the actual configuration of these
> machines were nor whether they were really comparable
> (i.e. same time frame, comparable RAM and disk etc.) and
> probably not even you will remember that after such a long time.
> That's exactly what benchmarks are for, provide reproducible
> results under controlled conditions, beyond "personal feelings".
> Spec doesn't test compile performance,
> but anyway this will depend less on CPU but
> much more on amount of available RAM,
> disk speed and size, whether the disk is local or NFS mounted etc.

I can only speak from the point of view of computing for embedded dev work,
where interactive and compile time metrics were quite valid indicators
of performance in the real world, even if a little subjective. In a way,
spec testing is an artificial environment which can bear little relation
to real application. I only ever saw spec testing as a rough guide,
especially as some of the results at that were listed seemed to bear no
connection with actual performance of machines seen on the bench. For
example, iirc, the Ultra 30/300 was listed as being of similar performance
to an Alphastation 500/400, but they were like chalk and cheese 
interactively,
with the sparc machine noticeably slower in every respect.

The Alpha boxes of the time were seriously fast in comparison with other
similar specced systems and usually had the disadvantage of having less,
not more memory. The base systems were parsimoniously light on memory and
dec memory upgrade prices outrageously expensive. I think the 3000 had 64
or perhaps 128K memory on a single, very custom, non standard module, that
ran half the width of the motherboard. It was never upgraded.

> HP snakes were much more "sexy" than any Alpha of 1993/94.
> Not only they were faster, but they also had an outstanding desktop,
> HP-VUE, which helped a lot in gaining popularity.
> 

I don't remember what the pa-risc machine model was, perhaps 720 or 725 
?. Big
19" monitor and base underneath. The display was, as you say, really neat.
However, nothing came close to Irix on the Indy for beautifull desktop, 
with
clever and usefull features. I ran hpux 10. something on some of the 
earlier
68k based technical computers and found the os to be hard work, with odd 
named
utilites and cryptic command line args that were so unlike mainstream unix
of the time. No doubt about the engineering quality though. If anything at
all came close to the Alpha on performance, it was sgi machines.

The early 3000 Alpha series were later considered to be underpowered, 
but it
wasn't noticeable at the time and the later ev56 and ev6 machines 
re-established
the lead for several years.

Regards,

Chris