PC 4GB RAM limit

[Mxsmanic]

Only if you're stupid enough to think that you need a quad Xeon with
4TB of RAM just to serve files.

You can serve files with a 386 as long as you have sufficiently fast
disks.

The major offender in computer performance today is slow disk drives.
They've improved only moderately over disk drives from half a century
ago.

 

[Mxsmanic]

You're arguing a 'PC' should use the same architecture, which also means
expense, and that is equivalent to 'replacing' it.

The architecture of mainframes isn't what makes them expensive, at least
not today.

 

[Al Dykes]

In message <[email protected]> Mxsmanic


I'm about as far from a linux zealot as you can get (I run Windows both
on desktops and servers), but this isn't a matter of being poorly
written.

As long as the system requirements are documented, it's not poorly
written, it's design goals which don't meet your needs, and I'd
recommend you purchase another product.


--

99% of the binary code that is Linux is generated by it's C compiler,
which is told what the model of CPU is. If it's told to generate
lowest common denominator x86 fode it Linux will run an anything
anywhere close to an intel x86 chip. The other one percent is either
other compilers which do likewise or some drivers that only get
selected when requested. Linux has been ported to so many
fundamentally different CPUs that all the architecture-specific stuff
is long gone.

 

[kony]

You can serve files with a 386 as long as you have sufficiently fast
disks.

The major offender in computer performance today is slow disk drives.
They've improved only moderately over disk drives from half a century
ago.

You must buy your drives from the wrong places, as I recall
drives less than 15 years ago that weren't over 2MB per
second, and todays being over 60MB/s, is a 30X speed
increase. Relatively speaking other subsystems haven't
increased much more if that much.

BTW, what drives are you referring to from half a century
ago? About 35 years ago the typical mainframe was still
near the end of the punch-card era.

 

[Al Dykes]

You must buy your drives from the wrong places, as I recall
drives less than 15 years ago that weren't over 2MB per
second, and todays being over 60MB/s, is a 30X speed
increase. Relatively speaking other subsystems haven't
increased much more if that much.

BTW, what drives are you referring to from half a century
ago? About 35 years ago the typical mainframe was still
near the end of the punch-card era.

Huh ?

Lets see, The Univac FastTrand head-per-track drum, in commercial use
in 1970, was the size of a minivan, stored about 100MB, access time
was 90ms, and transfer rate was 100kB/sec. Cost started at $180k.

http://www.fourmilab.ch/documents/univac/fastrand.html

The first type of disk I bought was 3330-vintage the size of a washing
machine, stored 200MB (removalable disk packs), 30MS access, 800kB/sec
transfer. The size of a washing machine, weighed 400 pounds, cost
$50k (1979) plus $250/month for maintenance and needed 3 phase power.
I had a room full of these puppies.

The first PC HD disk I had was 5MB and the seek time and transfer rate
were comparable to a 1x speed CD reader. ISTR it cost about $800.

Here's a good general link on disk drive history.

http://en.wikipedia.org/wiki/Early_IBM_disk_storage

 

[Mxsmanic]

You must buy your drives from the wrong places, as I recall
drives less than 15 years ago that weren't over 2MB per
second, and todays being over 60MB/s, is a 30X speed
increase. Relatively speaking other subsystems haven't
increased much more if that much.

Unfortunately, the transfer rate is not the problem. The access time is
the problem. Today it is around 6-8 milliseconds; forty years ago it
was around 40 milliseconds. That's not much of an improvement, and the
improvement that has occurred is mostly just a happy side effect of
greater data densities.

Sixty megabytes per second doesn't help much if you have to transfer 100
small blocks from different places on the disk and it takes 10
milliseconds to access each one.

BTW, what drives are you referring to from half a century
ago? About 35 years ago the typical mainframe was still
near the end of the punch-card era.

The first disk drives date from the mid-1950s. Thirty-five years ago,
disk drives were fairly common, although tape was more so. Punch cards
survived well into the eighties and I'm sure some systems still use them
today. The same is true for tape, although most tape has moved from
open-reel tapes to cartridges. Disk drives have gotten smaller and have
greatly increased in data capacity; transfer rates have increased in
consequence. Access times have improved only slightly.

A quarter-century ago access times equated to about 17,000 instruction
cycles. Today they equate to about 21,000,000 instruction executions.
So hard drives are much more of a performance problem today than they
were 25 years ago, even though they are slightly faster in absolute
terms.

 

[David Maynard]

And? Can't have it both ways.

That was tongue in cheek just like the grok grep was.

 

[David Maynard]

David Maynard writes:




If that's the instruction the software needs to execute, it's a reality.

The point is it isn't.

No, it's saying that coding for specific hardware platforms dramatically
impairs portability and is thus usually a poor design decision.

It is precisely the same thing: keeping backward compatibility for an
infinite period of time when technology has gone through over 10 years of
progression.

 

[David Maynard]

David Maynard writes:




No, I argue that software bloat consumes the horsepower of faster
processors. Almost all the improvements in hardware are absorbed by
software bloat.

Except that it doesn't, as has already been proven.

As a result, newer systems perform about the same as
older systems, from an end user's standpoint.
Poppycock.

If software were not so
bloated, new systems would have performance thousands of times better
than old systems.

And you can type text that fast?

It doesn't support my hardware.

Then use old hardware since nothing is faster anyway, from your "user
standpoint."

 

[David Maynard]

David Maynard writes:




Not in the mainframe companies I worked for.

Yes it is because it's clear you don't have a complete handle on what costs
are.

 

[David Maynard]

David Maynard writes:




The architecture of mainframes isn't what makes them expensive, at least
not today.

Then why don't you build them and get rich?

 

[Conor]

Except that it doesn't, as has already been proven.

Except that it does. Compare the average spec of a box from 1996-1998
and what we have today yet they feel no quicker.

And you can type text that fast?

Typing is a minor part of what goes on.

Then use old hardware since nothing is faster anyway, from your "user
standpoint."

I actually sell old hardware. I make a decent amount selling refurb P3
systems. I stick Win98 on them and they're great for web/office.

 

[David Maynard]

Huh ?

Lets see, The Univac FastTrand head-per-track drum, in commercial use
in 1970, was the size of a minivan, stored about 100MB, access time
was 90ms, and transfer rate was 100kB/sec. Cost started at $180k.

http://www.fourmilab.ch/documents/univac/fastrand.html

The first type of disk I bought was 3330-vintage the size of a washing
machine, stored 200MB (removalable disk packs), 30MS access, 800kB/sec
transfer. The size of a washing machine, weighed 400 pounds, cost
$50k (1979) plus $250/month for maintenance and needed 3 phase power.
I had a room full of these puppies.

The first PC HD disk I had was 5MB and the seek time and transfer rate
were comparable to a 1x speed CD reader. ISTR it cost about $800.

Here's a good general link on disk drive history.

http://en.wikipedia.org/wiki/Early_IBM_disk_storage

Yes.

Mxsmaniac seems to have mastered not only the appropriately inappropriate
measure and comparison but the fine art of nonsensical numbers as well.

 

[Conor]

Maybe running stuff out of the box isn't a good idea.

Absolutely but just trying to do the best apples and pears comparison
possible by using the OOB experience.

 

[Conor]

You must buy your drives from the wrong places, as I recall
drives less than 15 years ago that weren't over 2MB per
second, and todays being over 60MB/s, is a 30X speed
increase.

But they've stuck at that point for the last few years.

 

[Mxsmanic]

And you can type text that fast?

No. Such systems would simply not be necessary at all, except for a
handful of applications.

Then use old hardware since nothing is faster anyway, from your "user
standpoint."

Old hardware and software are no longer available.

 

[Davinderjit Markandeya]

Except that it does. Compare the average spec of a box from 1996-1998
and what we have today yet they feel no quicker.

BWAHAHAHAHAHAHAHAHAHA! Cooner the stupid ****tard **** strikes again.

 

[Mxsmanic]

Lets see, The Univac FastTrand head-per-track drum, in commercial use
in 1970, was the size of a minivan, stored about 100MB, access time
was 90ms, and transfer rate was 100kB/sec. Cost started at $180k.

Access times today are about 6 ms. That's an improvement of barely over
10 to 1, in forty years. When compared to processors and memory, disk
drives are actually orders of magnitude _slower_ than they have ever
been in the past.

 

[Mxsmanic]

Then why don't you build them and get rich?

There are not enough applications to run on them for the mass market,
and they generally are not single-user, GUI-oriented machines.

UNIX is close to a mainframe system, though, and a lot of people run
some version of that. They do so because UNIX is cheap.

 

[Al Dykes]

Access times today are about 6 ms. That's an improvement of barely over
10 to 1, in forty years. When compared to processors and memory, disk
drives are actually orders of magnitude _slower_ than they have ever
been in the past.

You forgot the fact that a disk, in 1970, was being used by the
several/many jobs or user tasks run concurrently on a mainframe. That
could be in the tens to hundreds.

Transer rate counts too. A Photoshop image can be 10s of MB and I'd
hate to have to wait each time that got read or written to an old
disk.