[Vm-dev] historical memory sizes (was: PICs)

Phil B pbpublist at gmail.com
Sun Jul 29 19:38:21 UTC 2018


Sorry I can't do inline right now (stupid Gmail Android client must have
'fixed' something)...

Ok, I wasn't clear that the study had that much lag (my fault for not
reading it). What threw me for a loop (and still does) was the claim that
the 'average' Sun machine had that little RAM at any time in the 90s.  I
don't ever remember using a non-x86 Unix system in the 90s with less than
16M.  And I went into 1990 on 16mb PCs (I was a day 1 adopter of Win 3.0 as
it was the easiest sell to get business people off of DOS... to all human
beings who have suffered as a result: sorry, but as bad as Windows is/was,
DOS was worse).

An aside and minor correction to your point re: PCs: *most* people didn't
have software that could use more than 1m+64k (himem weird feature).  But
there were commonly available exceptions for example: iirc Lotus 123 v3 was
still a DOS app but was able to use extended memory (subject to the
constraint that only labels could go into extended memory) so you could
effectively use up to 2-3 meg in a DOS-based spreadsheet.  And there were
CAD systems and databases that could directly use up to 4-8meg or so via
DOS extenders. There were also crude task switchers, TSRs, etc that
clumsily allowed using more RAM.  It was pretty hellish, but possible to
break the 1meg barrier in the DOS days.  And there was also Windows 286/386
pre-1990.  (It was bizarre, you couldn't convince most business people to
look at anything that wasn't DOS or later Windows usually citing cost as a
reason (really it was fear of the unknown) but they'd spend a small fortune
having their config.sys/autoexec.bat optimized to find an extra 50-100k of
low memory so that they could keep limping along in DOS)

No doubt RAM was rediculously expensive in the 80s and 90s (I remember
paying ~$1000 for the last 128meg in that 192 meg machine) but Sun machines
weren't cheap either.  That is why I was surprised that apparently so many
people skimped on RAM in the workstation world... seems like putting a 4
cylinder engine into a high end luxury car.  The low end of the PC world on
the other hand has always been terrible for skimping on everything.


On Wed, Jul 25, 2018, 10:27 AM Jecel Assumpcao Jr. <jecel at merlintec.com>

> Ben Coman wrote on Wed, 25 Jul 2018 12:36:18 +0800
> > On 25 July 2018 at 08:00, Phil B wrote:
> > > On Tue, Jul 24, 2018, 7:18 PM Jecel Assumpcao Jr. wrote:
> > >>
> > >> One of the papers in that list is the 1997 techical report "The Space
> > >> Overhead of Customization". One of the reasons that Java won over Self
> > >> was that its simple interpreter ran on 8MB machines that most of Sun's
> > >> customers had while Self needed 24MB workstations which were rare (but
> > >> would be very common just two years later). Part of that was due to
> > >> compiling a new version of native code for every different type of
> > >> receiver even if the different versions didn't really help.
> > >>
> > >
> > > Just had to interject... are you sure about the year and RAM sizes
> below?
> > >  If correct, was this a lowest common denominator RAM size or perhaps a
> > > report from several years earlier?
> Self 3/4 was indeed from 1993 to 1994 while they paper that did that
> space analysis was from 1997. The decision to drop Self and TCL in favor
> of Java was from late 1994 and cited the installed base at that time,
> not the machines that were being sold then.
> > > I remember in 1997 (might have been early '98) buying a Sun Ultrasparc
> 2
> > > with 256MB RAM for work (for a workstation, not our servers which had
> far
> > >  more) and having 192MB RAM in one of my machines at home.  Seems
> > > strange that 'most' of Sun's customers would be at 8 meg in that
> timeframe
> > > as 16-32 was fairly common even on PCs on the low end of what I was
> working
> > > with.
> >
> > I got curious...
> > 1993 = 66-MHz machine 16MB RAM
> > 1997 = 233-MHz machine 64MB RAM
> > 1999 = 500-MHz machine 128MB RAM
> > https://www.pcmag.com/article2/0,2817,2390914,00.asp
> When I talk to students about researching computer history, I always
> caution them to take old magazines with a grain of salt. I point out
> that these tend to focus on the very best products and not what typical
> people actually bought. If they looked at car magazines, for example,
> they might conclude that in 1978 everone was driving around in Ferraris
> and BMWs.
> And like I said above, you have to take into account that most computers
> at any moment in time are older models and not what is being sold then.
> Though we like to make graphs that show smooth growth in computer
> history, that isn't always the case. For disks, for example, we got
> stuck with 5 and 10MB for nearly half a decade before stepping up to
> 20MB (and then a quick jump to 30MB with the switch from MFM to RLL).
> Soon after that an exponential growth started that is still going on
> today.
> In the same way, two factors slowed down memory size growth between the
> late 1980s and early 1990s. One was the antidumping actions by the US
> against Japanese companies and on PCs there was a problem that people
> didn't have software that could use more than 1MB even if they were
> buying machines with 4MB. Workstations and Macs (and Ataris and Amigas)
> didn't have that problem nor did PCs after Windows 3.1 become popular.
> -- Jecel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/vm-dev/attachments/20180729/5d95f2f7/attachment.html>

More information about the Vm-dev mailing list