The comp tech guys often suggest that people get new computers. My attitude
is that often slowness is often the result of too little RAM.
Often: yes
Always: no
It has to be taken on a case-by-case basis.
For CPUs, how slow is too slow? Context here is just the usual
not-intensive apps: email, web browser, some simple documents and
spreadsheets.
That's largely a matter of personal use/habits. These
things could be done acceptibly on any system new enough to
support the amount of memory required to keep all operating
system and application data in memory instead of paged to
disk.
In other words, some people might be able to use a decent
(not low-end all-integrated) Pentium 2 /233MHz, but would
notice the speedup going to a CPU with 100MHz FSB & Memory,
(late K6-300 or Pentium 2/350), and notice another plateau
of improvement moving closer to 1GHz. More than 1GHz CPU is
not needed for these tasks in general, unless one of them
was atypically needy, like viewing video online, or giant
documents (especially non-100%-zoomed-PDF scrolling).
Often if someone has only these basic needs, the more
significant factor is system age as it relates to stability
or failure, particularly if it wasn't built with good parts
initially (like fans, PSU and motherboard), or if it were
subject to dirty environment like dust clogging up PSU or
heatsinks, or cigarette smoke fouling electrical contacts).
Some are better than others though, some of the most popular
early Pentium 2 systems with Intel motherboards are still
running fine today as they had very good construction and
low current/higher-voltage, quality capacitors and more
tantalum/other solid caps in use still... this was before
the environmentalists and anti-millitant types started their
campaigns againt tantalum mining, funding guerrilla warfare,
etc.
I have an 800 MHz PIII at home, and it seems fine. (Doesn't hurt that I
have 512 MB RAM.) At work, I've been using machines that are even
slower---550 MHz---and they still seem fine (after I dumped more RAM into
them).
Many people do still use machines of that era, and want
nothing but to keep them running as cheaply as possible.
Some do break though, and if the user (or a friend) doesn't
have the technical ability and desire to find and fix any
faults, it can be cost prohibitive to spend (perhaps) $150
to fix a system with only 30% of it's expected lifespan
remaining, when that $150 could go towards a newer system
with it's whole life ahead of it. Plus, some types of
failures are more problematic than others, for example a
hard drive failure may wipe out ALL of the user's data, and
they should have been making backups but might not have done
so in a timely manner. In other words, on a well built
system that was maintained properly too (dust cleaned out
and "if" sleeve bearing fans, they were relubed every few
years), the hard drive would fail before the board and PSU,
unless these latter parts had a particular defect.
Once the drive fails, again it is a technical issue of how
to get it running - some boards won't even recognize today's
modern capacity HDDs, and if you told someone they needed to
flash their bios if possible, they'd look at you like you're
speaking Greek... and again it becomes a matter of whether
they can get it done cost-effectively in addition to the
purchase of the new hard drive.
Of course, the stock line is that "computers are so cheap these days, you
might as well get a new computer."
Sometimes it's true, but personally I value the data more
than the systems, and considering that you can get a
barebones OEM system from the P2-P3 era at surplus 'sites
for under $100, it's still a matter of time to set it up,
install applications, etc- if those apps are even licensed
or installable on another system.
But given that prices seem to have a
floor, if you replace every two years, you're spending about twice as much
money as replacing every four years.
Non-applicable, anyone doing the tasks described above
should not need to replace every two or four years. If they
find their system failing in that interval, it would tend to
suggest they bought something TOO cheap, too many corners
cut and they encountered a failure mode as a result.
Typically in real-world terms this means to avoid the
cheapest things sold by most local shops- systems built with
generic came-with-case PSU, came-with-case or profit-leader
generic fans, PCChips/et al. motherboard, generic heatsink
(fan), or a poor case that overheats everything. Knowing
where to spend the extra 20% or so to improve the weakest
areas, most common failure points, may distinguish a system
that lasts for a decade from one that is trash in less than
half that.
BUT, some people really do need more performance, or just
want something faster for vanity's sake, or to keep up with
the Joneses, or because they heard WinXP is what they need
(and is "sometimes", somewhat true) and has a higher
hardware requirement.
Not to mention that ordering and
dumping more RAM into a machine seems a lot simpler than ordering an entire
computer, migrating the HD contents, ...
Maybe, but in general I seldom recommend upgrading ram
instead of outfitting the system with (that much) it in the
first place. IF someone went cheap and didn't get enough
memory at first, sure it makes sense to upgrade ram
sometimes, IF it's reasonable to believe the system would
otherwise remain viable for long enough to get the value out
of the memory upgrade.
So in summary we're back to what I first wrote;
It has to be taken on a case-by-case basis.
