Why Pentium?

wrote

Yes, but the designers of systems that are aimed at that market do.

And not that burn out question either. Funny that.

I'm not either, amazing tho that may be.

Irrelevant to what is being discussed, why some
prefer pentiums over athlons or later amd cpus.

 

Not necessarily for those who design systems aimed at that particular market.

Sure, it was always a very crude approach.

And I preferred the intel cpus just because
they were generally quieter for quite a while too.

Yeah, I generally prefer to use intel chipsets, because
of problems in the past, even with intel cpus.

And the most demanding game I play is Freecell Pro.

Yep, which is why stuff like the chipset can matter instead.

 

Well, migration is a thermal issue along with other 'not good' mechanisms
(not all of which are completely understood) but 'it locked up' isn't an
indicator. Well, other than indicating you're running at high temp and high
temp accelerates those mechanisms ala the Arrhenius activation energy model
(and others).

Unfortunately, no one outside the manufacturer knows what the numbers are
so while one can say a processor will last longer at a lower operating temp
it's anyone's guess how much longer or, conversely, how much operating at
high temp degrades it. You can use the halving per 10C guesstimate but we
don't know where the MTBF is starting at either. I mean, if running it at
50C instead of 70C cut MTBF down from 400 to 'only' a hundred years would
you really care? (Not that I'm saying this is the case but that we don't know).

Thermal *cycling* is another significant failure stress mechanism that I
mention only because it's one that deviates from the activation energy
model and can throw the 'halving per 10C' guesstimate off in the other
direction. (Note that even 'normal operation' induces thermal cycles from
varying CPU loads)

It's also interesting to note that when speaking of ion migration texts
will point out it can occur 'at modest temperatures of 100C', well outside
our 'lock up' discussion, but maybe not outside an internal hot spot. And,
of course, the Mil Spec temperature range goes to 125C.

Short version is I don't have a definitive answer

Btw, another 'rule of thumb' is a 10C drop in temp gives a 2% speed increase.

 

That's a good point because 'locked up' doesn't necessarily mean the
processor quit doing something, just not something 'intelligent', nor that
it quit consuming power.

 

Well, for general office use but lets not exaggerate too much. There are
still plenty of things business does where computing power makes a
difference, like CAD, modeling, and animation just to name a few.

Just recently did a dual, dual core, system so the user would only have to
wait a few hours for their animations to render.

 

I worked for a company on projects that bought thousands of computers over
the last ten years. In some cases the company bought them and delivered
them to the customer, in other cases the customers bought the computers.
The goal was always to buy the cheapest computers that would do an adaquate
job. There was a lot of discussion about the computers, but no one ever
discussed brands of CPUs.

The company researched the pros and cons of models of computers that were
commercially available. They never considered having computers custom
built. Most of the computers bought were from Dell. It was a matter of
processor speed, amount of RAM, hard drive capacity, USB ports, Built-in LAN
speed, never processor brand. In fact I don't remember the computers we
were choosing among ever having any other processor brand than Intel.

As others have said, this had nothing to to with the respective performance
of Intel vs AMD, it was simply a matter of the exclusive contracts that
Intel was able to make with the computer manufacturers.

 

Chip to chip, or more likely lot to lot.

Yes, but power plays into this too. It must run at speed under a
specified power. Sometimes one can over-clock processors if
extraordinary means are used to cool them.

No, as others have said, the device may not work but it's not
destroyed.

Good plan.

 

Nonsense. Executives are *very* careful with their pension plans.
ESRA and all that.

Oh, that's why companies have CIOs; to read glossies.

You read too much Tom's.

 

You got one of those right. The Pentiums beat out Athlons in speed and
reliability according to most tests. Besides, the big businesses
probably used Intel from the time they were 486's; they aren't going to
change now. Also, as "badgolferman" says, corporations generally buy
hundreds of computers at a time, all the same from Dell, HP, whatever.

hih,
poly-p man

 

That sounds like extremely outdated (by a couple of years) information
about the state of modern AMD thermal management.

 

Thanks for the message. I'm no expert on this, I'm afraid.

I found the Intel web site completely obscure on this issue. I understand
that 60 or 70 degrees will still work without problem, but the CPU starts to
throttle itself down if it gets too hot. I have never had any instability,
so I think that 50C is O.K. No doubt it will be lower in the winter!

 

Just to expand on this, there are steps in semiconductor processing
that work on single wafers, partial lots of wafers, full lots of
wafers, and multiple lots of wafers (a lot is typically 25 wafers).

Many of these steps are nonuniform across the wafer, sometimes edge to
edge, sometimes in a bullseye, and sometimes in a more complex
pattern. Likewise, results from wafer to wafer across a lot can be
different, and there can be a combination of the two effects. A
diffusion furnace is a good example of equipment that has both wafer
to wafer and across-the-wafer uniformity variations.

How a given piece of equipment performs at a given time depends on a
large number of variables, so there's always a performance window for
each process step for both target and uniformity.

As a result, chips will perform differently depending on where they
are on the wafer, within the lot, or what pieces of equipment they
were run on and when.

They try to shake this out when they put the chips in the speed bins,
but there will always be some chips that are near the performance
borders and hold up differently to heat variations than other chips
from the same batch.

TMI for most, I'm sure, but that's how it works in a nutshell.

There are also cases where higher performing chips are put into
lower-speed packages to meet sales commitments. This usually happens
in the case of a solid process design where the majority of the chips
are yielding in the high-speed bins. There was a lot of this during
the original Pentium production.

max

 

Outdated for sure.

AMD64 is supposed to shutdown the PC if the CPU gets too hot. According
to guy from AMD they are also thermally protected to not burn up.

K7's were different, they can burn up but most motherboards had a
CPU-Temp/Shutdown setting in the BIOS so the system should shutdown
before the CPU gets damaged.

I think Intel's slowdown feature is a joke, sure it might be useful but
to me it seems more like they were just covering up a flaw in a badly
designed CPU.

Ed

 

Your comparing a single core chip based on old technology to a
dual core on newer technology. I would hope the Intel could out
perform the AthlonXP.

 

Utter rubbish.

You "prefer" to live by ancient folklore too?

 

Which Pentiums and which Athlons are you talking about? Athlon64 has had
P4 beat for most business and gaming tasks for near 3 years now - same for
Opteron vs. Xeon. As for reliability, there's no basis for that conclusion
at all; in fact AMD has had better thermal behavior and management for a
similar amount of time. Any reliability concerns, e.g. Dell's massive
write-off on the capacitor problem, are independent of CPU & chipset.

 

I'd expect silicon chips to last long enough that their MTBF is to all
intents and purposes irrelevant - most solid state electronics has a
pretty good lifespan - the only thing that tend to let the side down
are electrolytics.

Whilst obviously we can't discuss modern machines in this context, as
a case in point I have an old Apricot LS Pro (something of a
collectors item now) powered by a a Cyrix 486SLC. This is a tiny,
_plastic_ chip with no heatsink and no fan. It was in daily use until
I picked it up maybe 18 months ago. I've no idea how hot the chip
gets - far too early for a motherboard sensor - but it's too hot to
touch in operation so that's probably at least 60C. It still works a
charm, maybe 14 years after it was built. Most people won't be using
the machine they have today in that amount of time, unless of course
like me they have a interest in old kit.

 

As others have pointed out elsewhere in this thread, there's more to
life than sheer performance. As you say, they run cool, making passive
cooling a realistic option, and the case can be smaller too. I chose
a VIA chip when I wanted a small, quiet machine for my bedroom - I use
it for exactly the same tasks as my other machines and most of the time
I don't even notice the lack of power.

 

Another reason is the increase in resistance at higher
temps. One could increase the voltage in such cases and
regain stability, but obviously it would then require even
more cooling to remain at a stable temp. threshold.

 

That is a reason to more carefully scrutinize the failure
point, which was not an AMD processor but another factor
like fan or grease failure, chassis cooling problems. One
should never buy a system with the idea that one of the
basic fundamental needs will fail and thus Intel's
last-resort shutdown would matter. Certainly that shutdown
feature is better than NOT having one, but it is not
something that should be among primary considerations in any
remotely normal system, selection.

I suggest that you drew the wrong conclusion. A system
built with an Intel CPU but same problem the AMD one had, is
not trouble-free either. You saw the result of the problem
as a focal point instead of the cause. Whatever that AMD
CPU was, that it was a past generation CPU is a sign that
many alternatives from either manufacturer produce more heat
today, we can't just write-off the AMDs as hot-running, and
contrary to urban myth, many Intel alternatives actually had
a higher TDP but merely idled cooler.

Put the $ towards the problem instead. If it overheats the
problem was the cooling system or maintenance (lack of)
towards cleaning out dust, replacing poor thermal compound,
or relubing junk fans (if for some reason it isn't viable to
replace them with good quality fans instead).

Some have had trouble, for example Intel southbridge USB
issues/burnout. Pointing to one past chipset used on AMD is
no evidence against AMD itself. Even in the past some
chipsets for Intel posed problems, like Sis 620 (or was it
630) refusing to use UMDA for HDD on NT/2K/XP in many cases.
So long as the system proposed doesn't use the specific
chipset, there is no point considering that past generation
chipset. It brings up another prudent practice though,
buying mature platforms where there is ample feedback about
issues.

That would be a silly random conclusion. One could argue
the same thing for an Intel Celeron w/Intel-integrated
video, as it is the most popular combination for the highest
selling market segment (OEM low-end).

Not at all. Many popular benchmarks make a ridiculous
assumption that one would only run a few of the premier
applications, newest versions of those. How many people do
you know that pay thousands of dollars every time a newer
version of their apps come out? Most people don't, only
getting newer versions when it happened to ship with their
new OEM system (which tends not to have premier apps on it
at all, except perhaps MS Office).

Take the typical apps of a few years ago and even Athlon XP
beat the P4 though online benchmarks suggested P4 beat it
most of the time (towards the end of the Athlon XP era at
least).

There are reasons to choose either alternative, it would be
most valid to choose based on the specific, most common or
most demanding use the system will encounter... as it is
with a comparison of any two CPUs having different
architectures. Nothing wrong with a P4 or Pentium D where
it excells but the very specific use, not even a newer
version of the same application, must be considered.