Intel, AMD...

On Thu, 18 Nov 2004 06:25:03 +0100, dion_b wrote:

Checksums? Checksums are normally done in software, so are completely
CPU-independent. If your data is sensitive, implement checksums in your
applications.

The software that I used to write, didn't do Checksums. That
was part of the CPU's job. Aren't you referring to the firmware of the
CPU?

Then what about the adage: You get what you pay for? Does
Intel COSTmore because it's WORTH more?

A non-argument. Cost is regulated by market forces which have little or
nothing to do with quality or performance. Take cars. A Toyota is
cheaper than a Ford, yet more reliable and fuel-efficient. But hey,
maybe the Ford is 'worth' more...

The adage is true, especially when the difference in price is
1/3. And BTW, Ford and other NA cars typically have better built
bodies than Japanese cars, which use recycled steel.

Or to stay on-topic for computing, take Rambus memory- it's been
consistently vastly more expensive than plain old SDRAM derivatives, yet
apart from a brief period in the early P4 era it only just managed to
equal (DDR-)SDRAM performance, it ran hotter and was less flexible (max
2 modules per channel with a performance loss for 2 as compared to 1).

I'm not familiar with the above. But if more expensive
materials are used, greater effort is expended in engineering and
manufacture, then all of that is factored into the price. It just
makes $en$e. :-)

That's not to say that cheaper is always better, it's just to illustrate
the pointlessness of such an argument.

Actually, SOMETIMES inexpensive, IS better. But usually, NOT.
If what you say is true, then nobody in his right mind would spend
more money, FOR NOTHING.

....or AUTOCAD, or 3D modelling, or photo and text editing?
I'm just wondering.

Text editing can be done perfectly satisfactorily on a Digital PDP-11
from the 1970s...

LOL!!! I saw a PDP-11! It was a really nicely built machine.
There were toggle switches to boot the brute up.

Photoshop is FPU intensive when running filters and utilises SSE(2)
extensions. Before AMD supported SSE2, Intel and AMD CPUs were about
equal, since SSE2 support AMD wins hands down, particularly when
measured on a performance/dollar scale.

See:
http://reviews.zdnet.co.uk/0,39023100,39164010-4,00.htm

In Autocad an Athlon 64 3400+ performs on average 8.3% better than a
similarly priced P4-3200

See:
http://www.xbitlabs.com/articles/cpu...4-3400_16.html

Good links above. I checked them out, Hmm... interesting.

Are we talking faster frame rates AND high resolution, at the
same time?

Whatever you want. Either a lot more of one or a little more of both.
It's just a ratio you can set as user. More power just gives you a
little more choice.

Usually, the higher the resolution, the lower the frame rates,
unless the processor-memory combination is powerful (and expensive),
enough.

In the computer world, SECONDS is an awfully long time! Such a
time lag is a VERY SIGNIFICANT difference.

So assuming there is a 1-second advantage for CPU x in Excel (Winstone
Business 2004 benchmark, say). Assume doing 12 such calculations per
day, five days a week, makes one minute's difference. That means a whole
lunch break per year (allowing for holidays etc), assuming the employee
in question never ever does anything parallel...

So what is all the hoopla about AMD or Intel being faster than
the other?

That must compare to nose-picking in terms of time wastage (plus the
obvious fact that an employee incapable of multitasking is a far bigger
problem in itself)

Sounds like an argument for slower processors. :-)

But isn't faster memory related to CPU speed?

No.

The most important factors are the speed of the memory chips (far slower
than CPUs these days) and the width of the bus with which to access
them, determining how many chips can be accessed in parallel. A close
second is the latency, how long it takes for a request for data from the
CPU to get to the memory and back.

Using two memory controllers in parallel can double maximum throughput,
a step Apple incidentally took in the mid 1990s, years before PC chipset
makers thought of it (and even then it was the rather obscure ALi who
first implemented it on the little-known Aladdin 7 chipset, which could
take both Intel and AMD Socket 7 chips). These days most performance
chipsets have dual channel access.

Latency can be addressed by CPU cacheing (mainstream CPUs now have a
Megabyte of L2 cache) and integrating memory controllers into the CPU
(which AMD has done in the Athlon 64 series).

And I suppose that Intel has not? Or has Intel resorted to
some other technique for increasing througput.

Pure CPU speed can even be a handicap, because it means the CPU is
running so much faster than the memory that it spends an awful length of
time idle waiting for data. The first Celeron CPUs based on the P4
illustrated that, with the CPU running up to 28 times faster than the
bus that connects it to the outside world and 14 times faster than the
best memory it could be paired up with, yet a tiny 128KB cache which
couldn't compensate for it.

I guess you could say that a $200 difference may be NAME
related. But what does it REALLY mean when Intel costs about $800 MORE
than the equivalen AMD?

It means that the marketplace is prepared to pay $800 more for snake oil
brand x rather than snake oil brand y. Assuming (as I do, incidentally)
that the intrinsic value of Intel and AMD chips is almost identical, if
Intel can get away with charging more, who are we to complain. AMD would
do the same if they could- and that they can't is merely a product of
market perceptions. In fact, as savvy consumers our response should be
to favour the better value over the inflated repuation, a choice that
collectively will push up AMD prices and drop Intel prices. And if Intel
ever becomes cheaper whilst delivering comparable intrinsic value- hell,
buy Intel.

No company would stay in business long enough to pronounce
'snake oil', if they charged that much more for essentially nothing.
Come on Dion. There must be something REALLY better, for consumers to
pay that much more than the equivalent AMD. From the links you gave
me, it seems that Intel is more RIGOROULSY engineered and manufactured
than AMD. I've heard this from other pros in the business. In
addition, Intel is less likely to overheat than AMD.

But, isn't the Intel P4 a 64 bit processor?

No.

The only thing '64-bit' about the P4 is it's bus, but then Intel (and
AMD) buses have been 64b wide since the days of the first Pentium in
1993. What is commonly meant by '64b' is that is can address 64b of
(virtual) data and process 64b long instructions. In that sense the P4
is firmly 32 bit, at least, except for the very latest EMT64-P4 CPUs,
which (just like the Athlon 64, incidentally) consist of a traditional
32b chip with a set of 64b extensions bolted on.

'True' 64b chips are rare beasts, Intel has its Itanium, Sun the
UltraSparc, and HP the PA-RISC64 and Alpha CPUs. MIPS also had a few in
times gone by, and IBM has its Power-series (with derivatives in Apple
G5 systems). However these chips are rarely seen on the desktop (OK, I'm
typing this on an Ultrasparc IIi-based machine, but this is not a
typical desktop ), mainly because they don't run Windows (except the
Alpha which ran NT) and are prohibitively expensive.

64b is a double-edged sword too, because having to double the length of
all addresses seriously bloats code, which decreases performance. 64b is
mainly hype, the prime use for most mortals is that it allows more than
4GB of RAM on a fairly simple motherboard.


And just for the record, I own more Intel-based systems (4 in use) than
AMD-based ones (2). I greatly admire Intel CPUs from a technical point
of view, as do I AMD CPUs and other more obscure stuff as well- I
particularly like Intel's tendency to over-engineer slightly. But when
price becomes a factor (as it always is in the real world), you want the
most bang for your buck. And Intel usually (though not always!) does not
deliver.
As an excercise, take a good benchmark in an application you consider
important. Then look up the prices of the CPUs in a pricewatch or
whatever. Divide the performance by the price. Then list the CPUs in
terms of price/performance. Finally choose the minimum performance
(and/or the maximum price) you are prepared to accept and discard the
CPUs that fail that mark. Choose the cheapest of what is left. That is
the rational way to choose a CPU.
7 times out of 10 that will lead to an AMD CPU at the moment. The
remainder will usually be laptops, where the Intel Pentium-M and
Celeron-M lines are quite remarkable, particularly as they combine
excellent performance with extreme economy in electrical terms.

Interesting analysis Dion.

And as for reliability- there are two relevant points:

1) the economic lifespan of a system
2) the technical lifespan of a system

In business most computers are written of over 3 years, which means that
almost always the techinical lifespan is much longer. For home use 5
years is more common and then technical issues become more important. A
system is good if its technical lifespan is longer than its economic
one. The technical bottleneck is almost always the moving parts. CPUs
don't move.

You forgot the lifespan of one's patience (with software bugs)
and just general frustration! :-)

If you intend to play with your hardware, rebuild your computers
regularly and swap components, the physical strenght of CPUs becomes
relevant, and until the advent of the Athlon 64 Intel had a considerable
advantage due to their aforementioned over-engineering.
But idiots like me account for less than one percent of users. If you're
one of the 99%, your CPU will stay in one place and never be at risk
from the dreaded core crunch (which incidentally CAN happen with older
Intel P3 CPUs too, a fact I can unfortunately relate out of personal
experience- although the cause was a badly secured heatsink which
bounced around too much during a long drive. That (as with similar AMD
incidents) was user negligence, not a design fault.

So in conclusion then, AMD is the better overall value for
most consumers, is that right?

JW


"FAILURE IS NOT AN OPTION is certainly true for
Israel in its conflict with the Arabs. Now Microsoft
has adopted the slogan as follows: FALURE IS NOT AN
OPTION. IT COMES BUNDLED WITH THE SOFTWARE!"

On Sun, 21 Nov 2004 22:23:41 -0800, JW
wrote:

On Thu, 18 Nov 2004 06:25:03 +0100, dion_b wrote:

Checksums? Checksums are normally done in software, so are completely
CPU-independent. If your data is sensitive, implement checksums in your
applications.

The software that I used to write, didn't do Checksums. That
was part of the CPU's job. Aren't you referring to the firmware of the
CPU?

CPU performs the calculation but it is a software routine
doing it.



Then what about the adage: You get what you pay for? Does
Intel COSTmore because it's WORTH more?

A non-argument. Cost is regulated by market forces which have little or
nothing to do with quality or performance. Take cars. A Toyota is
cheaper than a Ford, yet more reliable and fuel-efficient. But hey,
maybe the Ford is 'worth' more...

The adage is true, especially when the difference in price is
1/3. And BTW, Ford and other NA cars typically have better built
bodies than Japanese cars, which use recycled steel.

Not necessarily true, ... You certainly don't always get
what you pay for, Japanese cars of same size are not less
expensive than a Ford, and recycled steel is not much of an
issue relative to (anything else).


Or to stay on-topic for computing, take Rambus memory- it's been
consistently vastly more expensive than plain old SDRAM derivatives, yet
apart from a brief period in the early P4 era it only just managed to
equal (DDR-)SDRAM performance, it ran hotter and was less flexible (max
2 modules per channel with a performance loss for 2 as compared to 1).

I'm not familiar with the above. But if more expensive
materials are used, greater effort is expended in engineering and
manufacture, then all of that is factored into the price. It just
makes $en$e. :-)

Sometimes, but it's often a matter of licensing fees and
scale, the market demand for the product.



That's not to say that cheaper is always better, it's just to illustrate
the pointlessness of such an argument.

Actually, SOMETIMES inexpensive, IS better. But usually, NOT.
If what you say is true, then nobody in his right mind would spend
more money, FOR NOTHING.

You might be surprised how many people are not in their
right mind, and/or just drink up marketing because they
don't know the difference.



So what is all the hoopla about AMD or Intel being faster than
the other?

Boys and their toys, everything has to be a contest.


And I suppose that Intel has not? Or has Intel resorted to
some other technique for increasing througput.

Oh, this thread was about CPUs?
No wonder it's already off track, "Intel vs. AMD" threads
are a waste of time... read a few hundred and be content
that you've wasted an evening.



No company would stay in business long enough to pronounce
'snake oil', if they charged that much more for essentially nothing.
Come on Dion. There must be something REALLY better, for consumers to
pay that much more than the equivalent AMD. From the links you gave
me, it seems that Intel is more RIGOROULSY engineered and manufactured
than AMD. I've heard this from other pros in the business. In
addition, Intel is less likely to overheat than AMD.

You might want to spend more time learning and less time
speculating. There are spec sheets for Intel, AMD. There
are benchmarks. There are all kinds of datum available yet
you seem intent on just making wild assumptions that are
already known contrary to human nature. Humans buy what's
advertised, it's just that simple. They may not need it,
may not know what it is, or even if there are better
alternatives but to many people, "TV" or the written word
seem like magic, it MUST be true/great/best if it's made it
to media. AMD has never had strong advertising but Intel
has.



So in conclusion then, AMD is the better overall value for
most consumers, is that right?


In conclusion, the individual user's specific, most
demanding uses must be weighed and alternatives from each
processor manufacturer can be compared at the needed
price-point. It's pointless to consider who made the CPU,
you might as well compare AMD Athlon XP vs Athlon 64 or
Intel P4 vs Tualatin Celeron. There is a broad knowledge
base, either do the hard research specific to your needs to
just buy whatever is advertised or promoted at the right
price... like most people do (Dell, Compaq, etc, etc,
buyers).