Explain AMD CPU Speeds

  • Thread starter Thread starter Mista Fadedglory
  • Start date Start date
M

Mista Fadedglory

Please explain to me what AMD means when they advertise a cpu that is 2.17
ghz, but also 2700+. What does that mean? Is it equal to a 2.7ghz intel
machine?
 
Please explain to me what AMD means when they advertise a cpu that is 2.17
ghz, but also 2700+. What does that mean? Is it equal to a 2.7ghz intel
machine?
Click to expand...

Pretty much.
 
yes.

amd pr of 2700+ = intel 2.7ghz chip.

this is due to the chips' architecture. basically the pipelines are shorter
on the amd, so it can get the data out quicker - atleast, that is what i`m
lead to beleive.

tim
 
Mista said:
Please explain to me what AMD means when they advertise a cpu that is 2.17
ghz, but also 2700+. What does that mean? Is it equal to a 2.7ghz intel
machine?
Click to expand...

I recently read an article about that. The amd speed rating was at first
supposed to indicate the performance level of the processor to be the equal
of one of their older cores if it was run at that speed. It turns out that it
also matched the intel speed pretty well also. In today's world you should
believe only some of what you read and even less of what you here. Benchmarks
are all over the place when it comes to the performance of processors. A lot
of people that do the testing seem to have their own agenda when it comes to
the hows and whys of testing. Individual system components can have as much of
an affect on overall performance as processor speed. AMD is a solid performer
for sure and they are priced very reasonably. Compared mhz to mhz an intel
processor can't perform with an amd and when compared at their speed rating as
opposed to their actual operating speed an amd will often outperform the
equivalent intel processor. I like them both but amd cost considerably less.




Roger
 
I think it depends what your CPU needs to perform, for example me, I do not
play games, rather I need to do a lot of 3D rendering and video/audio
encoding. So I would prefer Intel's since Intel CPUs are much better for
scientific calculation that's what I need

http://www6.tomshardware.com/cpu/20030217/cpu_charts-25.html

Normally, AMD is $$better choice$$ I think
I recently read an article about that. The amd speed rating was at first
supposed to indicate the performance level of the processor to be the equal
of one of their older cores if it was run at that speed. It turns out that it
also matched the intel speed pretty well also. In today's world you should
believe only some of what you read and even less of what you here. Benchmarks
are all over the place when it comes to the performance of processors. A lot
of people that do the testing seem to have their own agenda when it comes to
the hows and whys of testing. Individual system components can have as much of
an affect on overall performance as processor speed. AMD is a solid performer
for sure and they are priced very reasonably. Compared mhz to mhz an intel
processor can't perform with an amd and when compared at their speed rating as
opposed to their actual operating speed an amd will often outperform the
equivalent intel processor. I like them both but amd cost considerably
Click to expand...
less.
 
I think it depends what your CPU needs to perform, for example me, I do not
play games, rather I need to do a lot of 3D rendering and video/audio
encoding. So I would prefer Intel's since Intel CPUs are much better for
scientific calculation that's what I need

http://www6.tomshardware.com/cpu/20030217/cpu_charts-25.html

Normally, AMD is $$better choice$$ I think
Click to expand...

You're right - it depends what your CPU needs to perform - and -
Normally, AMD is $$better choice$$ -.

But Intel being better for "scientific calculation" may be grossly
misleading, to say the least. Depends on your code. (But frankly I've
never seen the P4 more than match the AthlonXP, at the best, in a
scientific environment, even with SSE2 optimization.)

Best way to understand this is to look at the various parts of the cpu
instruction set. Basically we have:

'386 general & integer logic

'387 scalar floating point

MMX original 64 bit int SIMD extension using '387 registers

3Dnow/advanced 3Dnow (AMD only) 2x64/4x32 bit vector SP fp SIMD using
'387/MMX registers

SSE aka MMX2 2x64 bit vector SP fp SIMD using new registers.

SSE2 (P4 and AMDx86-64 only) 128 bit fp&int vector instructions.

There's four things to really notice here:
First - AthlonXP doesn't have SSE2 and every application, that
supports SSE2 and can make heavy use of it, tend to run from as fast
to significantly faster, on the P4. For illustration, consider this
page:
http://www.anandtech.com/cpu/showdoc.html?i=1834&p=5
For video and mp3 creation, enthusiast or professional, the P4 is
undoubtably a better choice. Even considering $.
For mediaplay and occasional media encoding, AthlonXP does well
enough.

Secondly - The AthlonXP, on the other hand, really stomps on Intel P4
when it comes to '386 and '387 code. Have a peak at these benchmarks:

Raw FPU power, Sciencemark 2.0 BLAS double precision.
(Matrix multiplication)
'387 FP, higher is better
A-XP3200 3178
3.2 P4C 2015

ScienceMark 2.0 Cypher AES.
(good ol' integer code)
lower is better.
A-XP3200 14.7
3.2 P4C 23.9

That's 85% faster/clock on '387 FP and 92% faster/clock on integer for
AMD!
The AthlonXP is fundamentally a much, much faster general logic cpu.
The Intel P4 excells in bandwidth and nonconditional repetetive
operations on large blocks of data. On many current applications,
these two sides seem to average out, so that XP+ rating = P4 MHz in
very rough terms. But when PCW early '03 took a set of applications to
the P4 and Athlon, ostensibly to test HT, even the slowest AthlonXP
-2600+, outperformed the best P4 - 3.06 GHz@533 HT on all. Another
eyeopener might be this page:
http://www.anandtech.com/cpu/showdoc.html?i=1834&p=6

Best XP+rating/P4MHz correspondence seems to be enjoyed by games.

The third thing to notice is that currently published crop of
benchmark suits, really put massive emphasis on SSE2. Not only do we
see a lot of media encoding, but a lot of synthetic benchmarks, like
'content creation' and 'cpu-performance' are really just other SSE2
jpeg/mpeg exercises underneath. Only other thing that outfits like
toms hardware seem interested in, are things that bottlenecks in
ram-bandwidth, another P4 strong show. Also game benchmarks usually
only sends renders down the 3D pipe. In a real game there's a lot of
AI and pathfinding going on. Something the AthlonXP does better.

This does seem somewhat biased to me, but the really important lesson
here, is that you have to figure out what it means for you and your
apps. If you have purchased a P4 on basis of the predominant sse2mpeg
benchmarks, and haven't bothered to upgrade your apps to the latest
P4-optimized versions, - well, then the above two Sciencemark
benchmarks (fpu & int) show you just how fooled you are.

Anyone looking for a faster PIII - the AthlonXP is exactly that.
Anyone looking more for a high throughput 'mediachip' in a mediarich
desktop PC, never mind old style computing speed - well, that's how
Intel figured it too.

The fourth thing to notice is that P4 GHz are not equals either. The
800FSB versions are faster by a very good measure!


ancra
 
Please explain to me what AMD means when they advertise a cpu that is 2.17
ghz, but also 2700+. What does that mean? Is it equal to a 2.7ghz intel
machine?
Click to expand...

Clockrate doesn't have more to do with cpu speed or power, than revs
have to do with the horsepower of internal combustion engines. You
don't hear the 10,000 hp diesel of a large motorship exactly
screaming, do you? Revs and clockrate are related to power, but they
aren't it.

AMD wants to get away from public perception that MHz or GHz is the
"speed" of the cpu. Thus they are allocating numbers based on some
kind of performance rating instead. This practice continues on AMDs
other cpu families, Opteron, AthlonFX, though with other numbers.

Performance characteristics of AthlonXP and Pentium4 are so dissimilar
that no simple figure can compare them. And AthlonXP ratings are
supposed to correspond to the older Athlon MHz, if anything. To me it
looks a bit like a numbering system. It takes no genius to figure out
that, for example, the Barton 2500+ does better than it should, after
studying some benchmarks, so the system can't be strictly relative.

Comparing AMD numbers with P4 MHz work well enough as a rough guide
for game performance though. Nitpickers will find it depends on the
game and which P4 you're comparing to.


ancra
 
Back
Top