Another Intel vs AMD Comparison

Semperon @ 1.8GHz vs Celeron D @ 3.3GHz

Running a floating point program using the
..NET 1.1 complier, running trig, power, mul, add
functions, nothing too fancy.

The Semperon beats the Celeron D by 20%.
Wow, nearly half the clock and still outperforms.
Something seriously wrong with the P4 architecture!!!

Oh God, not again. No offense to you, the thread starter, but I
participate in another forum that is
FULL of AMD fanboys who take it upon
themselves to bash the ever living crap out of Intel on a daily
basis. As true as your findings may be, would you mind stating some
specifics? For example, what model of Celeron and Semperon processors
were you running? Are we talking new generation versus old
generation(SKT:754 Vs. SKT 478)? Also, what setup were you using? ie.
Motherboard, RAM, HDD, etc. I'm not trying to start anything, or am I
challenging you in anyway. I'm just curious as to how AMD's Semperon
beat out Intel's Celeron by 20%; because that's quite a substantial
amount for a budget processor.

These results are real, and I'm no AMD fanboy.
In fact I've been an Intel fanboy like for ever,
until the P4 that is.

This is a highly FPU intensive app which I've
written myself.. It's using .NET JIT compiler,
so it's not as if it's compiled with very old compilers
either.

HDD has nothing to do with this result whatsoever.
RAM and motherboard are virtually irrelevant also.
Yes it's SKT 478 vs 754. It's a very apples to apples
comparison using comparable processors, and
AMD wins it hands down. It's really astonishing
given the clock difference.



"Mike L" wrote in message
...
Oh God, not again. No offense to you, the thread starter, but I
participate in another forum that is
FULL of AMD fanboys who take it upon
themselves to bash the ever living crap out of Intel on a daily
basis. As true as your findings may be, would you mind stating some
specifics? For example, what model of Celeron and Semperon processors
were you running? Are we talking new generation versus old
generation(SKT:754 Vs. SKT 478)? Also, what setup were you using? ie.
Motherboard, RAM, HDD, etc. I'm not trying to start anything, or am I
challenging you in anyway. I'm just curious as to how AMD's Semperon
beat out Intel's Celeron by 20%; because that's quite a substantial
amount for a budget processor.

To give another comparison, the same FPU intensive app
performs only 35% better on a Celeron D @ 3.3GHz
than a Coppermine Celeron @ 1.1GHz. When you
go up 3 times in clock and get only 35% better on P4
architecture, on a CPU intensive app doing ANYTHING,
you know there something seriously screwy with the P4.
The AMD Semperon results are just further confirmation of
the same.

Raymond wrote:
Semperon @ 1.8GHz vs Celeron D @ 3.3GHz

Running a floating point program using the
.NET 1.1 complier, running trig, power, mul, add
functions, nothing too fancy.

The Semperon beats the Celeron D by 20%.
Wow, nearly half the clock and still outperforms.
Something seriously wrong with the P4 architecture!!!

Don't know what your test program looks like or does but the premise of
your comparison is only valid if one considers work per clock cycle some
sort of holy grail. It's not if one just wants to get the work done as the
P4 accomplished it by doing less work per clock at higher clock speeds
while AMD processors do it by more work per clock but can't clock as fast.

"David Maynard" wrote in message
...
Raymond wrote:
Semperon @ 1.8GHz vs Celeron D @ 3.3GHz

Running a floating point program using the
.NET 1.1 complier, running trig, power, mul, add
functions, nothing too fancy.

The Semperon beats the Celeron D by 20%.
Wow, nearly half the clock and still outperforms.
Something seriously wrong with the P4 architecture!!!

Don't know what your test program looks like or does but the premise of
your comparison is only valid if one considers work per clock cycle some
sort of holy grail. It's not if one just wants to get the work done as the
P4 accomplished it by doing less work per clock at higher clock speeds
while AMD processors do it by more work per clock but can't clock as fast.

When it comes judging CPU architecture, work per clock cycle
is exactly like some kind of holy grail. Granted,
Intel makes better silicon, so then can clock faster and make
up some of the difference. That's why they're still business.
Otherwise it would've been no contest.

Wow, with that kind of attitude I can see you running a
very unsuccessful review site. Let's
take a look at a few flaws shall we:

This is a highly FPU intensive app which I've
written myself..

So you're admitting that you've compiled the test
yourself, ok I'll give you some credit
there. But exactly how many systems did you test your program on? How
many different possible hardware and software configurations were
taken into consideration when you benched this program on those two
test beds? Did you ever stop to think that there was a margin of
error involved, or that maybe there'd be some sort of conflict due to
an unlimiting number of possibilites? Therefore I'll ask you this, how
do you know that 35% increase over the Coppermine core really wasn't
65% or even 18%? See any good software programmer would know that
software is unintentional, and very bare during it's initial stages
of development. Therefore a plethora of tests must be run to ensure
maximum efficiency and a truthful denotation before it's released.

HDD has nothing to do with this result whatsoever.

So basically you're telling me that if I was to take a setup which had
a Bigfoot drive installed compared to another setup which had a SATA
HDD installed, that the results would be totally dismissable because
hard drives have nothing to do with the outcome at all? Quite an
interesting use of logic. My friend, though a HDD may have next to
nothing to do with 'this' test, a HDD still has something to do with
everything. Take a hard drive running at 7200RPM and the exact same
model running at 5400RPM, and tell me if you notice "any"
difference at all, because you will. Doesn't the hard drive have to
somehow run the program off of itself, at the same time trying to
manage the best way possible for the system as whole to utilize the
application at hand? C'mon man..

RAM and motherboard are virtually irrelevant also.

Again, more apalling use of logic. So let me just ask you, are you
being serious? If so, this is just as worst as your example above. I
don't even know if I should bother going into detail about this one
because it's almost not even worth it. I know you're talking about
motherboards and RAM being irrelevant in
[b:dfab9b917e]your[/b:dfab9b917e] test, but once again, just for
comparison sake, a motherboard and RAM are probably the two single
most vital components sitting in your case right now.

Yes it's SKT 478 vs 754. It's a very apples to apples
comparison using comparable processors, and
AMD wins it hands down. It's really astonishing
given the clock difference.

That's a fair comparison in your eyes? Comparing AMD's next generation
processor to a previous generation one from Intel? I take it that
you've totally disregarded the fact that the Semperon is an all new
redesigned core, with an on-die memory controller and the ability to
run 64-bit code; albeit totally useless by today's standards, but
known for [i:dfab9b917e]sometimes[/i:dfab9b917e] enhancing 32-bit
programs(once again, margin of error). I'm not surprised that the
Semperon beat out the previous generation Celeron "hands
down", what did you expect. Secondly, clockspeeds mean nothing
here. AMD processors have a higher IPC throughput then Intel
processors, and have a better performance-per-watt rating as well.

This is ridiculous, I've debated out every single one of your
points... and badly. What was going through your mind when you posted
this topic? Like I said, I'll give you credit for writing your own
program and performing some tests, but your results aren't viable by
a long shot.

"Mike L" wrote in message
...

silly drivel snipped

That's really funny stuff Mike. Your ignorance
is showing! I wrote this program myself. I know
exactly what it does, what it uses, and what it doesn't.

The HDD is a complete non-issue. Because
99.99% of the time the program doesn't use
the HDD at all. Not to mention that my Semperon
system has the slower HDD anyway - as if it
mattered.

As I explained initially, it's an FPU intensive
app that does a lot of math processing. Do you
know what that even means??? If you did, that
should've been enough. It means it's not memory intensive!
Over 95% of it's time is spent processing small code-blocks wholly
inside the CPU. The effect of RAM and motherboard
on this program is infinitesimal. But if it makes you happy,
they're both running DDR400 at roughly the same bus speed.
Again, the Semperon is actually on the cheaper more inferior
board, disadvantaged lower memory-CPU bandwidth there,
no dual-channel setup at all on the Semperon system-
as if it mattered.

And yes it is a fair comparison, both processors
were released about the same time, have comparable
prices, and markets.

"Raymond' wrote, in part:
| And yes it is a fair comparison, both processors
| were released about the same time, have comparable
| prices, and markets.
|
_____

And that 'comparable market' is concerned about 'small code-clocks wholly
inside the CPU' and 'FPU' intensive? A little knowledge is, well, a little
knowledge.

Phil Weldon

"Raymond" wrote in message
news:Tv3Wf.5412$Od7.2959@trnddc06...
|
| "Mike L" wrote in message
| ...
|
| silly drivel snipped
|
| That's really funny stuff Mike. Your ignorance
| is showing! I wrote this program myself. I know
| exactly what it does, what it uses, and what it doesn't.
|
| The HDD is a complete non-issue. Because
| 99.99% of the time the program doesn't use
| the HDD at all. Not to mention that my Semperon
| system has the slower HDD anyway - as if it
| mattered.
|
| As I explained initially, it's an FPU intensive
| app that does a lot of math processing. Do you
| know what that even means??? If you did, that
| should've been enough. It means it's not memory intensive!
| Over 95% of it's time is spent processing small code-blocks wholly
| inside the CPU. The effect of RAM and motherboard
| on this program is infinitesimal. But if it makes you happy,
| they're both running DDR400 at roughly the same bus speed.
| Again, the Semperon is actually on the cheaper more inferior
| board, disadvantaged lower memory-CPU bandwidth there,
| no dual-channel setup at all on the Semperon system-
| as if it mattered.
|
| And yes it is a fair comparison, both processors
| were released about the same time, have comparable
| prices, and markets.
|
|

"Phil Weldon" wrote in message
link.net...
"Raymond' wrote, in part:
| And yes it is a fair comparison, both processors
| were released about the same time, have comparable
| prices, and markets.
|
_____

And that 'comparable market' is concerned about 'small code-clocks wholly
inside the CPU' and 'FPU' intensive? A little knowledge is, well, a
little
knowledge.



My CPU intensive application doesn't care what the "comparable
market" is concerned with. My point was that the Celeron D vs Semperon
CPU comparison was fair, in case it went over your head.