First Athlon64 X2 benchmark

[Ed]

The test I have data for is CineBench 2003's CPU test, which is one of
those rare benchmarks which can make use of multiple CPUs and thus shows
a dual core part in its best light. For comparison, we can see how an
Athlon64 X2 4800+ fares against a single core Athlon FX-53, as well as
two Intel Extreme Edition Pentium parts, including the dual core 840
CPU. Check out the scores below!
http://www.elitebastards.com/#news9950

Ed

 

[Rob Stow]

The test I have data for is CineBench 2003's CPU test, which is one of
those rare benchmarks which can make use of multiple CPUs and thus shows
a dual core part in its best light. For comparison, we can see how an
Athlon64 X2 4800+ fares against a single core Athlon FX-53, as well as
two Intel Extreme Edition Pentium parts, including the dual core 840
CPU. Check out the scores below!
http://www.elitebastards.com/#news9950

Ed

Can you provide a direct link to this review ?

I couldn't find a link to it anywhere on the page at that URL.
It does not even contain a single instance of either "AMD" or
"Athlon".

Using their "search" tool to look for "840 and cinebench" finds
nothing.

 

[Derek Baker]

Can you provide a direct link to this review ?

I couldn't find a link to it anywhere on the page at that URL.
It does not even contain a single instance of either "AMD" or "Athlon".

Using their "search" tool to look for "840 and cinebench" finds nothing.

They've deleted it. Got this from Google's cache:
http://64.233.183.104/search?q=cach...rds.com/+athlon+x2+benchmark&hl=en&lr=lang_en

 

[Rob Stow]

They've deleted it. Got this from Google's cache:
http://64.233.183.104/search?q=cach...rds.com/+athlon+x2+benchmark&hl=en&lr=lang_en

The Google cache doesn't have the benchmarks. It tries and fails
to load a picture from the original location.

Someone probably violated an NDA and got slapped down by Intel or
AMD.

 

[Yousuf Khan]

The Google cache doesn't have the benchmarks. It tries and fails to
load a picture from the original location.

Someone probably violated an NDA and got slapped down by Intel or AMD.

Well, whenever the NDA expires, the both dual-cores (the Intel EE & the
AMD X2) were both miles ahead of the single-cores, and the X2 was
significantly ahead of the EE too.

Yousuf Khan

 

[Ed]

Can you provide a direct link to this review ?

I couldn't find a link to it anywhere on the page at that URL.
It does not even contain a single instance of either "AMD" or
"Athlon".

Using their "search" tool to look for "840 and cinebench" finds
nothing.

I copied the benchmark, posted here....
http://forums.amd.com/index.php?showtopic=45108

 

[Tony Hill]

Well, whenever the NDA expires, the both dual-cores (the Intel EE & the
AMD X2) were both miles ahead of the single-cores, and the X2 was
significantly ahead of the EE too.

If by "significantly ahead" you mean that the X2 was all of 3.9%
faster than the P4 EE, then yes... Personally I'd call it a tie
between the two dual-core chips, though I suppose AMD can take some
comfort if their 'mainstream' Athlon64 (ie not the FX) can match
Intel's Extremely Expensive Edition.

I am rather disappointed though that AMD is sticking with their
current naming convention with dual-core chips. To me it just really
doesn't do justice to what sort of performance you're going to see
from the chip, either in a good way or bad. In this test we're seeing
that the performance gap from a 4000+ (single core) to the 4800+ (dual
core) is going to be MUCH larger than what the number difference will
suggest. However in many other applications the advantage of
dual-core chips will be next to nothing. I guess the theory is that
it will all even out in the end, but I'd still much rather have seen a
more Opteron-style of naming convention.

 

[epaton]

If by "significantly ahead" you mean that the X2 was all of 3.9%
faster than the P4 EE, then yes... Personally I'd call it a tie
between the two dual-core chips, though I suppose AMD can take some
comfort if their 'mainstream' Athlon64 (ie not the FX) can match
Intel's Extremely Expensive Edition.

I am rather disappointed though that AMD is sticking with their
current naming convention with dual-core chips. To me it just really
doesn't do justice to what sort of performance you're going to see
from the chip, either in a good way or bad. In this test we're seeing
that the performance gap from a 4000+ (single core) to the 4800+ (dual
core) is going to be MUCH larger than what the number difference will
suggest. However in many other applications the advantage of
dual-core chips will be next to nothing. I guess the theory is that
it will all even out in the end, but I'd still much rather have seen a
more Opteron-style of naming convention.

i guess a bump of 20% for a nameing scheme using most apps seems about
right, the people who will really benefit from dual cores like coders and
fancy graphics will/should already know about its uses but many people
just buy a pc with higher numbers, explaining the uses/benefits of 2 cores
will result in a blank stare and hell even i would have trouble following
intels naming scheme.

while an industry standard scheme based upon a wide range of benchmarks
would be nice i still figure if you are prepared to look up the benchmarks
yourselves then you can make the best judgement, if not just pay more
money and benefit the rest.

 

[Derek Baker]

If by "significantly ahead" you mean that the X2 was all of 3.9%
faster than the P4 EE, then yes... Personally I'd call it a tie
between the two dual-core chips, though I suppose AMD can take some
comfort if their 'mainstream' Athlon64 (ie not the FX) can match
Intel's Extremely Expensive Edition.

Some comfort? Assuming there there's the same price ratio as between the
current single-core equivalents, I'd say a hell of a lot.

 

[boe]

Are there any games that take advantage of a dual core system?

 

[Ed]

Are there any games that take advantage of a dual core system?

Not that I know of... don't game much.
Ed

 

[Evgenij Barsukov]

Are there any games that take advantage of a dual core system?

Because most of very intensive calculations are done in relatively
small part of the program (lets call it "core") I suspect it is
relatively easy to release a patch that multy-threds only this core.
There were examples in the past, like when 3dNow optimisations came
out, ID Software released path for Quake II that was making use
of this optimizations. So program would basicaly run two different cores
- one for single and another for multy-processor system.

There are of cause also many little "physical simulation" things that
might be distributed through the program. It is also resonable simple
to change all of them at once to multy-threaded version - by changing
optimization settings of the complier or by upgrading version of
underlaying Math library. Most fortran Math libraries (like BLAST) are
being updated in backwards compatible way, so it might be just a matter
of changing one DLL to make a lot of functions multy-threaded.

Corresponding multi-threaded math-packages are widely awailable because
heavy-duty calculations have been run on Multy-processors. Many
math packages are available both in single and multi-threaded editions.

Regars,
Yevgen

 

[Gnu_Raiz]

Because most of very intensive calculations are done in relatively
small part of the program (lets call it "core") I suspect it is
relatively easy to release a patch that multy-threds only this core.
There were examples in the past, like when 3dNow optimisations came
out, ID Software released path for Quake II that was making use
of this optimizations. So program would basicaly run two different cores
- one for single and another for multy-processor system.

There are of cause also many little "physical simulation" things that
might be distributed through the program. It is also resonable simple
to change all of them at once to multy-threaded version - by changing
optimization settings of the complier or by upgrading version of
underlaying Math library. Most fortran Math libraries (like BLAST) are
being updated in backwards compatible way, so it might be just a matter
of changing one DLL to make a lot of functions multy-threaded.

Corresponding multi-threaded math-packages are widely awailable because
heavy-duty calculations have been run on Multy-processors. Many
math packages are available both in single and multi-threaded editions.

Regars,
Yevgen

You bring up a good point, with dual cores the need for a dedicated
physics chip might be reduced, as the second core could do more of the
calculations. Of course this would have to be in software like everything
else so will take some time, but as memory recollects even with a physics
chip you still will need the software. Since gaming on the PC has taken
somewhat of a second stage to consoles, I don't see them spending all the
resources to really make it take off. I hope they do, but I keep hearing
how PC games have been canceled, and I keep seeing more and more console
ports showing up all the time. You can almost count on one hand the PC
games that are made for the PC and use it to its fullest, this is in the
past two years.

Gnu_Raiz

 

[Rob Stow]

Take a look at the game benchmarks at
http://techreport.com/reviews/2005q2/opteron-x75/index.x?pg=14

When comparing the 2.2 GHz chips, the dual-core Opty 175
outperforms both a 2P systems using Opty 248's and totally
outclasses a 1P system using Opty 148.

The power consumption numbers are impressive too.
http://techreport.com/reviews/2005q2/opteron-x75/index.x?pg=16
Under load, an Opty 175 uses about 13 W less than an Opty 148 and
an astounding 126 W less that a pair of Opty 248's.

 

[jack_webb]

Since gaming on the PC has taken
somewhat of a second stage to consoles, I don't see them spending all the
resources to really make it take off. I hope they do, but I keep hearing
how PC games have been canceled, and I keep seeing more and more console
ports showing up all the time. You can almost count on one hand the PC
games that are made for the PC and use it to its fullest, this is in the
past two years.

Gnu_Raiz

I think you could easily blame the PC gaming industry for that.

Video card makers that insist on designing >$200 cards, and game
designers (I'm talking about DOOM3 here) that code extremely bad
software to force users to expensive video cards (Nvidia, that's you).

Corruption tends to kill the goose that lays the golden eggs.

 

[AD.]

Are there any games that take advantage of a dual core system?

The Falcon 4 flight sim was supposed to get a good boost with SMP when it
was in 'full campaign' mode.

In that mode it was modelling an entire war going on around you in real
time. Then again, the 5-10x higher clock speeds today (from when it came
out) probably make that irrelevant.