Why Pentium?

[Mxsmanic]

If the system has been recently booted and/or doesn't have
enough memory, true. Otherwise, not necessarily or rather,
it depends on the job. There certainly are a lot of things
bottlenecked by the HDD(s), but if you start timing how long
it takes some things to open, then look at the # of files,
total size additive plus the seek time of the HDD, there's
still quite a lot of time unaccounted for. The HDD is one
of the most common large bottlenecks but the CPU is also to
blame for some of it.

In the experiments I've tried, disk delays account for almost 100% of
the time spent waiting for many things to happen on the PC.

Depends on the task, Windows GUI navigation maybe not, maybe
not even email, websurfing, basic office tasks. Problem is
that although these are the most common PC tasks and do
respond well to a HDD upgrade, there's still plenty
stressing the CPU for brief peaks even if total % of
utilization isn't that high.

The peaks are so brief that they are not perceptible to the user.
Additionally, there really isn't any way to upgrade a disk. Even the
best disk drives are only slightly more performant than their
ancestors from thirty or forty years ago. Today's access times are
around 8 milliseconds, as opposed to around 30 milliseconds thirty
years ago. The improvement has been very small indeed. And the
increase in speed of CPUs, disk delay has come to dominate total
response delay in PCs, along with network delays.

True, but quite a bit of memory goes a long way even if it's
slow memory as it's practically always a faster filecache
than the hard drive.

Yes, but truly random access to the disk--as often occurs in
multiple-application environments--will rapidly make a file cache
useless.

 

[Mxsmanic]

... or if your motherboard had protection as the next
generation AMD platforms did.

Motherboard protection is less reliable, since it requires two working
components, not one.

You cite something that is no
longer true (with more modern AMD platforms) as a reason to
now not buy AMD.

The lack of this important feature in the past was a direct reflection
of decisions made by AMD, and showed that some of those decisions were
very poor. Unless all AMD employees from that period have been
replaced, the decisions of that era must be taken into consideration
even for choices made today, as they indicate the quality of decisions
overall.

This sounds inflated.

I lost two machines to overheated AMD processors, and each one cost me
a bit over a thousand dollars.

The then-aged and used CPU and board were worth $1000?

The entire machine, not just the motherboard and CPU.

Neither AMD nor Intel's 'sinks have
ever been known as high-end, but certainly a board and CPU
valued at $1000 should be considered just that.

These machines were SonBooks, and they used third-party heatsinks and
fans.

Again is it not a reason to avoid AMD now anymoreso than any
other manufacturer who has make something that failed (and
most have, nothing is perfect).

Bad experiences with AMD = 2, bad experiences with Intel = 0.
Conclusion: buy Intel.

This is very odd, the odds that such would happen are
astronomical against it (having both fail like this).

The odds were 100% if the CPU fans failed, which they did (both being
very cheap fans). One of the fans failed within days.

I don't think you can even say for sure that it was AMD's
fault rather than the fan manufacturer or motherboard
manufacturer, or the system assembler or ???

The damage was done by the overheated processor; it was thus an AMD
problem.

yes, you saw "result" rather than "cause". AMD didn't CAUSE
the failure.

The fan failed, and the CPU overheated. But an Intel CPU would not
have overheated. Case closed.

That doesn't decrease your loss, but perhaps it will make
you think about what heatsink (and particularly fan) you use
on (any) CPU, including Intel's.

I have thought about it. Now I buy Intel.

Something was wrong outside of the CPU or fan failure then,
an overheating CPU will not pose a serious fire hazzard.

Anything that produces the smell I encountered could have easily
caused a fire.

Are you really that dense? Obviously the problem was that
it overheated. If you want Intel, fine, If you want it
because of your unfortunate loss, also fine. It doesn't
begin to change the PROBLEM, which was the CAUSE of the AMD
based system overheating.

I prefer more than one fail-safe mechanism.

Nonsense. It did not almost set the machine on fire. If
your machine almost caught on fire I would suspect the CPU,
the fan, and everythign else was merely a casualty of
whatever DID overheat, but it wasn't the CPU because a CPU
will not produce that kind of heat to ignite anything.

Oh yes, it will. You've never seen those videos on the Web of AMD
processors overheating?

Ok, note that I was never trying to talk you into buying
one, rather discussing your reasons for avoiding parts that
do not have the potential problem you are citing as the
reason not to buy them. You are just mad at AMD and that is
why you won't buy, today. You are entitled to be mad at
whoever you want I suppose.

I'm not mad at AMD; you are in love with AMD (or at least you seem to
be promoting AMD).

I lose nothing by abandoning AMD in favor of Intel. I've already lost
way too much by trusting AMD instead of Intel.

 

[Rod Speed]

Actually Intel's CPUs can and have burnt up _with_ their
CPU shutdown feature if they lose all external cooling
(the heatsink) and are powered on from cold off.
The diode can't always reach fast enough to cover this.

Still worth having for fan failure etc.

That doesn't make it an essential feature
at all, but for another more obvious reason.

We'll see...

Systems ran for years without it- if it had
been essential, they could not have done so.

Those were significantly lower power cpus.

Systems ran fine for years without cpu fans too.

So what ?

Any CPU since the early Pentium days needed a 'sink,
and most active cooling (or carefully engineered passive
cooling via chassis airflow, same diff.), or would fail,
yet they ran fine. It was not essential nor is it now.

No one said a word about ESSENTIAL.

Everyone has been rubbing your nose in the FACT
that when thermal shutdown COSTS PEANUTS,
only a fool wouldnt include it in the cpu design.

Because shit does happen, even with systems designed by
god's gift to system design and its terminally stupid to have
to replace the most expensive component in the system
because some fool like you didnt include thermal shutdown
in the cpu design WHEN THAT COSTS PEANUTS.

That doesn't make it a bad idea, it seems Rod would like to
imply I meant that even after stating otherwise multiple times.

You need your seems machinery seen to, child.

What it does mean, and what got the thread all boated out
of proportion, is that it is no substitute for forethought about
reliable cooling, that a cooling system failure during the lifetime
of the system is cause for reassessment of that design.

Or you could get real radical, use what appears to be a decently
designed cpu heatsink and fan, AND HAVE THE INSURANCE
THAT THERMAL SHUTDOWN PROVIDES when you see the
occasional cpu fan failure or just user stupidity, continuing to
use the system even when the system keeps falling flat on
its face because they dont check for fur in the cpu fan etc.

Like I said, thank christ you never get any say what
so ever on cpu design, or anything else at all, ever.

 

[Rod Speed]

Rod Speed writes

Do you have any feel for fan life on boxed Intel CPUs?

I havent seen one fail.

I presume they use fans of reasonable quality, but I'd
prefer to think that they use fans of very high quality.

The quality appears to be quite adequate
given that I havent seen one fail.

And I care more about quiet systems too.

In fact with the last Northwood cpu I installed, the cpu
fan was so quiet that I had to look carefully at the fan
to check that it was spinning up, and thats with a very
quiet power supply and no hard drives etc and no case fans.

 

[George Macdonald]

Any CPU will burn up if it loses all external cooling, so providing
for CPU shutdown if the temperature climbs too high makes perfect
sense, and is an essential feature.

It's not shutdown he's objecting to - it's silent throttling in a chip
which, at the time of its intro, could not be produced consistently enough
to guarantee it would not happen. What the situation is with current
product I've no idea... and don't care - there are better choices.

 

[George Macdonald]

Virtually no one is doing decision support, and many executives can
hardly understand e-mail, much less any other computer application.

Nonsense. Every corp. uses decision support and in just about every part
of their organisation to plan strategic and tactical moves; to not make use
of such valuable tools would be suicide. The tools may be embedded in an
application which presents results in, e.g., HTML but they are there... not
necessarily for use by "executives" though not necessarily excluding them
either.

As a general rule, the higher the rank, the more a PC is for looks
rather than work. The actual horsepower requirement shrinks, but the
desire to look good with a fancy and recent computer expands.

The fact that a notebook looks good and fits the executive image has
nothing to do with the horspower inside.

 

[George Macdonald]

True. And if the AMDs had just done that, I might still have them.

So you still choose based on prejudice and not on current results!

In these cases, though, I smelled a strong burnt smell when I walked
into the room, and some parts of the motherboard had been deformed or
discolored by heat. Clearly, these processors had overheated far too
much. They still worked, for a while, but then lots of highly
unbelievable errors started to crop up when I used the machines. When
I saw that they were increasing in number and occurring in the most
improbable places (the OS and programs that had run for years without
errors), I realized the processor was failing. Within a day or two,
the failures were so frequent that it couldn't get through POST. So I
replaced the machine ... with Intel. The price difference was
trivial, especially compared to the money I lost when the old machine
cooked.

What you are describing is not, at its root, a CPU problem. Intel systems
from the same mfr/assembler would have spent half their life shutting down
umm, unexpectedly. There are many K6 systems still running happily every
day with the correct assembly methods... with the original heatsink/fans.

If I understand your story correctly, to continue to run those systems for
a day or two(?) after smelling burning and seeing the motherboard
"discolored by heat", due to a failing fan(?), beggars belief... user
negligence appears to be a factor here. So why don't you tell us exactly
which models of AMD processors were involved here and which Intel ones
replaced them?

In all the years, I've had one CPU fail due to overheat... a Cyrix 686
system which I stupidly/naively bought preassembled "bare-bones" from an
on-line vendor. On replacing the CPU, it turns out there was no thermal
interface material at all.<shrug>

 

[George Macdonald]

No, they won't. The major delay in response time on a typical desktop
comes from disk drive access time, not any lack of performance in the
CPU. If the system is connected to a network, network delays may be
even worse, and CPU speed still won't matter.

Oh yes they will. Tony is right -- he usually is -- the snapiness of
task startup/switching on a dual core CPU is instantly noticable. The
startup time of Windows, with all the startup junk placed there nowadays,
is also noticably improved. Drive performance is, of course, also
important.

Additionally, even if you have a blazingly fast CPU, you need very
fast memory to get full performance from it. Cache is not a
substitute for fast memory.

Rubbish. The difference between DDR400 3-3-3-8 and 2-2-2-5 timings is
purely imaginary. Even a drop to DDR333 wil barely be noticed.

 

[George Macdonald]

I don't think anybody can guarantee a mechanical device like a fan
does not fail unexpectedly within a short time frame. Insisting that
you can obtain such a model that will NEVER fail within say the 2~3
years a system is useful, is really stretching it.

And yet how many fans have you had fail unexpectedly without making enough
noise to alarm first? In all the years I haven't had a single one.

I've seen one case of Intel fan fail within two weeks. I'm pretty sure
Intel did not pick a fan for their boxed processor expecting the
useful lifespan of the system to be less than 10 days.

That *is* surprising and indicates a severe lack of QA by err, somebody.

 

[Rod Speed]

It was very easy to build an AMD based system that was
quieter than Intel w/Intel-retail sink-fan combination.

Wrong with the Northwoods.

Intel's fans from the celeron and P2 era were not very quiet
considering the CPUs did't produce much heat. Their P3 fans
should have been deemed defective because they all started
making high pitched bearing whine noise after a few months.
We shelved most P3 heatsinks because of this, replaced
with aftermarket 'sinks. I still have a few brand new P3
'sinks lying around too, they didn't get reused as much
as most other 'sinks because of that oddball stepped
base on the metal and the plastic level-clips.

I was talking about P4s.

They buy fans from manufacturers that make "normal" fans of
high quality. These fan manufacturers spend a great deal of
time (all actually, it IS their core business) engineering
fans then Intel comes along and wants their own special spec
which is a degraded function fan over what the manufacturer
already had. Nidec and Sanyo Denki are the main two, both
making good stock fans and both worse after re-spec'd.

Easy to claim.


What matters to me is whether they are nice and quiet and adequately reliable.

Not interested in getting all obsessed about the cpu fan,
there is plenty of other stuff thats much more important
WHEN YOU HAVE THERMAL SHUTDOWN THAT
ENSURES THAT IF THE FAN FAILS, THE CPU WONT DIE.

 

[Rod Speed]

kony wrote

My daughter's P3-1GHz, with stock Intel heatsink/fan,
is 5 years old and has had no such problem.

Mine neither. About the same age.

 

[Rod Speed]

I used to think that too, but I've seen too many
grandmothers plopping pictures in their 2400x2400 scanners,
then firing up their auto-everything photoediting app that
chokes on processing a 200MB image.

Granted, the average job(s) done on a PC are not CPU bound,
but at the same time the average jobs aren't memory bound,
or even hard drive bound. The user is the bottleneck on a
typical desktop system.

Precisely, so there isnt any point in obsessing
about the cpu choice performance wise.

 

[Rod Speed]

kony writes

In the experiments I've tried, disk delays account for almost 100%
of the time spent waiting for many things to happen on the PC.

The peaks are so brief that they are not perceptible to the user.
Yep.

Additionally, there really isn't any way to upgrade a disk.

Cant agree with that at all.

Even the best disk drives are only slightly more performant
than their ancestors from thirty or forty years ago. Today's
access times are around 8 milliseconds, as opposed to
around 30 milliseconds thirty years ago.

It isnt the access time that matters with most personal desktop
system use, its the thruput and that has increased dramatically, just
because of the sectors per track and rotation speed improvements.

Laptop drives are noticably slower than decent 7200 rpm desktop
drives with the sort of stuff people do on personal desktop systems.

The improvement has been very small indeed.

Nope. The effect of the speed of a 1G drive was quite
noticeable at boot time on a system that is essentially
a network terminal. A 4G drive from the same manufacturer
improved the boot time very noticeably, Fujitsu drives.

And the increase in speed of CPUs, disk delay has
come to dominate total response delay in PCs,

That was true even in 486 days.

along with network delays.

Not too many of that class of user do that much over networks tho.

 

[Rod Speed]

I was done before I started, because I
don't have fans dying at 2-3 year intervals.

Neither do I.

Havent seen that bearing whine you claim is
universal with Intel P3s boxed fans either.

Still using one in a PVR too.

 

[Rod Speed]

First, you'd have to be assuming the failure rate was
independant of the number of operating hours which it won't
be. Second you have not calculated for a MTBF of 50K above.
Third if the bathtub curve you propose next holds true, the
majority of failures would be clustered towards the 50K hour
mark rather than dispersed inbetween. Forth I hardly ever
use 50K fans, they're almost always 100K. Fifth you'd be
assuming 24/7 operation which isn't necessarily applicable
to anything but a server, but many servers tend to use two
distinctly different fan setups- either 120mm in the PSU
exhaust or in shorter rackmounts, arrays that "could" be
small diameter fans. I will grant you that in all my
projections about fan lifespan thus far, I have completely
ignored the small diameter fans in rackmounts that run at
very high RPM. On these particular fans, they should be
swapped out every year or two, even more often on a valuable
(function) system.


With some gear it's possible, though with good QC and sample
testing this can be mitigated.


Depends on sample size, it could indeed be reasonable if for
example, any fans making a strange noise or vibration were
immediately discarded. Another real example, a few years
back I bought three cases of a specific NMB fan, a model I'd
not bought previously nor since then so they are known
unique and can be discriminated against other NMBs I've
bought, and none of which have failed yet AFAIK... but quite
a few of the systems they're in have been brought back for
other reasons like spyware/viri removal, OS or hardware
upgrade, etc., plus I have several of the fans here in
systems too. I can't be 100% certain that none have failed,
but there is no evidence of it and I do tend to hear about
sytems... usually spyware but occasionally the ones with
popped caps because I got stung on a bunch of Gigabyte
boards from the P3 era, ended up selling 3 different
versions of their boards at the time, all with the bad caps.
@#$% Tom's Hardware liked them, I should have known not to
trust a quick review.

What I know about fans I know from real world experience
with good, and bad, brands. I've hauled off enough dead PCs
and relubed enough fans to have a really good grasp of which
ones stand the test of time. That doesn't mean I can expect
a 100.00% success rate, but where is there a guarantee of
100% in ANY part you'd buy?

Might just be why all cpu manufacturers
with a clue include thermal shutdown now.

 

[Rod Speed]

"IF" it's run long enough, every single part will have
failed, or rather, that first part that takes the whole
system down. That part does not have to be, and usually is
not the CPU fan if you used a good, well implemented fan.

Many years later, as with any mechanical device (containing
"thing") it is prudent to do maintanance. This goes for
PCs, cars, houses, anything... you replace parts when at the
end of their useful lifespan, when it is convenient or
scheduled, not as an emergency "accident" scenario. You
service the rest, clean out dust, etc and so forth, as
required.

Essentially while what you wrote might seem right on the
surface, in reality it is wrong. If you choose to properly
implement a good quality fan, the system will have been
retired before the fan had aged enough to fail, whether by
choice or another part failure. Only when using lesser
quality fans or bad design/implementation choices would you
find the CPU 'sink fan failed prematurely.

You have already demonstrated your inability to do that and
the resultant damage that you now blame on a 3rd party.
Sorry but you are the last person on earth I'd consider to
have knowledge about fans, except maybe on the one or two in
particular, to avoid using.

Perhaps you are like Rod, pretending to know something without
having ample evidence, only looking at a small subset of all factors.

You have absolutely NO idea of what evidence I have, child.

AND I have enough of a clue to have noticed the evidence
that even amd now has thermal shutdown in their cpus,
and have manage to grasp why they chose to do that too.

There are plenty of devices out in the wild built with good
fans that don't break down in a few years time. Perhaps
you never even noticed them because they keep on running
fine... but use pretty much the same fans you could have
in your pc if you made better choices.

No perhaps about the FACT that anyone with a clue has
thermal shutdown in their cpus WHEN THAT COSTS
PEANUTS, even when cpu fan failure is unlikely, if only
because users who let their cpu fan get clogged with
fur are much more likely and it isnt exactly practical
to only sell systems to those who have been tested
to be obsessively anal about cleaning their cpu fan.

 

[Mxsmanic]

Every corp. uses decision support and in just about every part
of their organisation to plan strategic and tactical moves; to not make use
of such valuable tools would be suicide. The tools may be embedded in an
application which presents results in, e.g., HTML but they are there... not
necessarily for use by "executives" though not necessarily excluding them
either.

I don't think you have the same thing in mind when you talk about
decision tools. Classic decision tools--the kind hyped by the
industry and trade rags--are scarce. Excel spreadsheets don't count.

Decision tools are extremely difficult to create because they require
an understanding of business that almost no executive actually
possesses.

 

[Mxsmanic]

So you still choose based on prejudice and not on current results!

I choose based on past experience. After being burned twice by AMD
(in more ways than one), I've decided to switch to Intel. I know that
Intel doesn't have the specific deficiencies that caused a problem for
me with AMD, and there is no _disadvantage_ to switching to Intel, so
why not?

What you are describing is not, at its root, a CPU problem.

Yes, it is. The CPU should not continue to run as it overheats.
Intel CPUs don't.

If I understand your story correctly, to continue to run those systems for
a day or two(?) after smelling burning and seeing the motherboard
"discolored by heat", due to a failing fan(?), beggars belief...

The one that burned up wouldn't boot reliably, as I recall.

... user negligence appears to be a factor here.

Defective design is a factor here. The processor was a fire hazard.
In a different context, it would not have been able to get a UL rating
because of the fire risk.

So why don't you tell us exactly
which models of AMD processors were involved here and which Intel ones
replaced them?

I don't remember. Athlon Thunderbird something. I only recall that
it was AMD.

The current processors are Intel Pentium 4 processors running at 3
GHz.

 

[Mxsmanic]

It isnt the access time that matters with most personal desktop
system use, its the thruput and that has increased dramatically, just
because of the sectors per track and rotation speed improvements.

No, it's the access time that matters. Very few disk I/Os are large
enough to profit from greater throughput. The disk spends most of its
time seeking for I/Os of only small blobs of data.

Nope. The effect of the speed of a 1G drive was quite
noticeable at boot time on a system that is essentially
a network terminal. A 4G drive from the same manufacturer
improved the boot time very noticeably, Fujitsu drives.

30 milliseconds / 8 milliseconds = 300% improvement for disks

5000 MIPS / 4 MIPS = 12,400% improvement for processors

That was true even in 486 days.

Yes, but it's much worse now, because nothing else is holding the
system back as much, and because applications have bloated to the
point where they do hundreds or thousands of disk I/Os even for the
simplest functions.

 

[Mxsmanic]

Oh yes they will. Tony is right -- he usually is -- the snapiness of
task startup/switching on a dual core CPU is instantly noticable.

I've used multiprocessor systems before, and snappiness is not a
function of processor speed.