3DMARK05

[liam]

At the weekend I just upped my AMD2100+ 1GB PC2700 ATI Radeon 9800 Pro128Mb
to an AMD3200+ 1GB PC3200 same Graphics Card.
Before I replaced the CPU I downloaded and Ran 3DMark05 and reran it again
after I swapped the CPU.

Before 2397
After 2685

Not really the speed boost I thought I would get.
Some parts of 3dMark05 ran at 1fps....

Got the impression that 3DMark05 is not mean't to be a benchmark for current
equipment.

CS:Source Timedemo was a different story.....

Before 62 fps @ 1024*768
After 120 fps --------------

 

[Sham B]

hi liam

3DMark05 seems to me to be useless for anything except seeing the difference between high end ATI
and nVidia cards. Its become a marketing tool, not a benchmarking tool, especially because theres
very few games that actually use the full capabilities of the new cards (unless you go for silly
resolutions).

IMO, theyve made it too high end to cater for manufacturors, and fogotten about the end users in the
rush. The '04 version seems a far better choice for seeing how your system does for current games

In any case, I use Prime 95 to check for system stability after a CPU/Memory/mobo overclock, and
in-game timedemos or fraps to check for graphics handling/gpu overclocking stability.... that
combination seems to be the best bet overall for my purposes.

I have been able to OC my XP2800 by 10% and my 9800 pro by 15%, both running stable in all current
games by using these more accurate, 'real world performance testing' tools. I could probably go
higher, but there seems no real need - theres nothign that I run/play that makes my system show any
real lag on at the moment at my normal resolutions (1024x768 or 1280x1028, with graphic options set
to 'high' for games, 1600x1280 at 32bit openGL for applications - most processor intensive being
Doom 3 (for games) and SoftImageXSI/Photoshop running at the same time (for applications)), and I
can safely run the whole system in its overclock and with repeated benchmarks overnight with no
errors or noticeable overheating of the components and/or run games at the next highest setting
and/or cut a day off a week long XSI animation render.... which is really what benchmarking is all
about, not (IMO) to get some arbitrary 3DMark numbers.


S

 

[NightSky 421]

At the weekend I just upped my AMD2100+ 1GB PC2700 ATI Radeon 9800
Pro128Mb to an AMD3200+ 1GB PC3200 same Graphics Card.
Before I replaced the CPU I downloaded and Ran 3DMark05 and reran it again
after I swapped the CPU.

Before 2397
After 2685

Not really the speed boost I thought I would get.
Some parts of 3dMark05 ran at 1fps....

Got the impression that 3DMark05 is not mean't to be a benchmark for
current equipment.

CS:Source Timedemo was a different story.....

Before 62 fps @ 1024*768
After 120 fps --------------

I don't have 3DMark05 yet, but based on what I have heard about it, I have
to agree with many of the sentiments that Sham B made in his post. It seems
that many benchmarks overly stress hardware to make the hardware you
currently own look bad, even though it really isn't bad. I also believe
that the scores 3DMark03 and 05 come up with are geared more towards towards
the video card than the CPU, which probably explains why your gain was so
small considering the big leap in processor you made.

For what it's worth, I have a 9800 Pro on a 2.8GHz Pentium 4 (Northwood C)
with 1GB RAM and some parts of the 3DMark03 benchmark which test the video
card run slowly on my system. I'm not referring to the CPU tests, but even
the one part where the space battle is taking place usually sees my frame
rate between 25 and 35 at any given time. Mother Nature can be slow too.

Give hardware a couple of years and I'm sure frame rates will improve in
3DMark05 to the point where it's totally fluid. I'd only be concerned about
real-world performance.

 

[liam]

I don't have 3DMark05 yet, but based on what I have heard about it, I have
to agree with many of the sentiments that Sham B made in his post. It
seems that many benchmarks overly stress hardware to make the hardware you
currently own look bad, even though it really isn't bad. I also believe
that the scores 3DMark03 and 05 come up with are geared more towards
towards the video card than the CPU, which probably explains why your gain
was so small considering the big leap in processor you made.

For what it's worth, I have a 9800 Pro on a 2.8GHz Pentium 4 (Northwood C)
with 1GB RAM and some parts of the 3DMark03 benchmark which test the video
card run slowly on my system. I'm not referring to the CPU tests, but
even the one part where the space battle is taking place usually sees my
frame rate between 25 and 35 at any given time. Mother Nature can be slow
too.

Give hardware a couple of years and I'm sure frame rates will improve in
3DMark05 to the point where it's totally fluid. I'd only be concerned
about real-world performance.

A proper benchmark program would have all the timedemo's from all the
popular games and run them all at varying resolutions / settings.
I don't suppose it would be much for each major game software house to
supply said part of program.

As you say that would be a real world benchmark....

 

[TMack]

At the weekend I just upped my AMD2100+ 1GB PC2700 ATI Radeon 9800 Pro128Mb
to an AMD3200+ 1GB PC3200 same Graphics Card.
Before I replaced the CPU I downloaded and Ran 3DMark05 and reran it again
after I swapped the CPU.

Before 2397
After 2685

Not really the speed boost I thought I would get.
Some parts of 3dMark05 ran at 1fps....

Got the impression that 3DMark05 is not mean't to be a benchmark for current
equipment.

Its NOT meant to be a benchmark for CPUs. The clue is in the title - it is
3D that is being benchmarked rather than cpu performance so it is designed
to be relatively cpu independent. If you are using the same graphics card
you will get a fairly similar score - as you have discovered.

Tony

 

[SteveK]

Its NOT meant to be a benchmark for CPUs. The clue is in the title - it
is 3D that is being benchmarked rather than cpu performance so it is
designed to be relatively cpu independent. If you are using the same
graphics card you will get a fairly similar score - as you have
discovered.

On top of that, 3Dmark05 is the ideal benchmark to make your card looking
really really bad if it's not an ultra high-end card! Piece of trash
benchmark software, probably specifically designed to motivate people
looking at those crazy priced $600+ cards! Futuremark stinks!

 

[HockeyTownUSA]

On top of that, 3Dmark05 is the ideal benchmark to make your card looking
really really bad if it's not an ultra high-end card! Piece of trash
benchmark software, probably specifically designed to motivate people
looking at those crazy priced $600+ cards! Futuremark stinks!

It's geared to be ahead of the curve. It is meant to tax a high-end system
so that it will still be a valid benchmark in a few years. Otherwise if we
all kept using 3DMark2001 we'd all know that we have top-end DirectX 7 video
cards and not DX9c with all the latest shaders and other goodies. My system
has an Athlon 64 3500+ and ATI Radeon X800 XT PE (finally, yay!), and I
scored about 5900. And that would make sense considering the video card
difference. 16 pipes versus 8, huge speed differences in RAM and GPU, etc.
Run 3DMark03 and I'm sure you'll see a significant difference.

 

[SteveK]

It's geared to be ahead of the curve. It is meant to tax a high-end system
so that it will still be a valid benchmark in a few years. Otherwise if we
all kept using 3DMark2001 we'd all know that we have top-end DirectX 7
video cards and not DX9c with all the latest shaders and other goodies.
My system has an Athlon 64 3500+ and ATI Radeon X800 XT PE (finally,
yay!), and I scored about 5900. And that would make sense considering the
video card difference. 16 pipes versus 8, huge speed differences in RAM
and GPU, etc. Run 3DMark03 and I'm sure you'll see a significant
difference.

You telling me DX9c is such a performance hog or are all cards designed
deliberately performance insufficient except expensive high-end hardware?
Nice try, Hockey

 

[TMack]

On top of that, 3Dmark05 is the ideal benchmark to make your card looking
really really bad if it's not an ultra high-end card! Piece of trash
benchmark software, probably specifically designed to motivate people
looking at those crazy priced $600+ cards! Futuremark stinks!

So don't use it! Its only a (FREE!!!) benchmark for goodness sake. The
purpose of a benchmark is to make comparisons, the numbers themselves are
irrelevant. The only thing that matters is whether or not one type of card
scores consistently higher or lower than another. Not surprisingly, the
newest 'high end' cards score higher than earlier models - what else would
you expect FFS????

Tony

 

[SteveK]

So don't use it! Its only a (FREE!!!) benchmark for goodness sake. The
purpose of a benchmark is to make comparisons, the numbers themselves are
irrelevant. The only thing that matters is whether or not one type of
card scores consistently higher or lower than another. Not surprisingly,
the newest 'high end' cards score higher than earlier models - what else
would you expect FFS????

The slideshow effects in some tests are really encouraging even for some
high-end cards..

 

[TMack]

SNIP!

The slideshow effects in some tests are really encouraging even for some
high-end cards..

Indeed, they are 'encouraging' - they are encouraging graphics card
manufacturers to make cards that can display them at decent frame rates.

Tony

 

[HockeyTownUSA]

You telling me DX9c is such a performance hog or are all cards designed
deliberately performance insufficient except expensive high-end hardware?
Nice try, Hockey

Sorry. Guess you misunderstood what I was trying to say. Basically 3DMark05,
from what I understand, uses all the latest shaders, real-time shadows (i.e.
multiple light sources), among many other things I don't understand. A
top-end DirectX 9 card should be able to handle this better than an older
card that either doesn't even support these features or has a much slower
architecture (combination memory, GPU, pipelines, etc). I can understand
complaining if a game runs as a slide show with a top end machinel, but a
benchmark is just that, a benchmark of one's system and/or components. It's
like someone with a Pentium 60 with a Voodoo expansion card complaining that
3DMark03 runs like a slide show.

 

[SteveK]

---- Original Message ----
From: "HockeyTownUSA" <[email protected]>
Newsgroups: alt.comp.periphs.videocards.ati
Sent: Wednesday, December 22, 2004 11:27 AM
Subject: Re: 3DMARK05

....

Sorry. Guess you misunderstood what I was trying to say. Basically
3DMark05, from what I understand, uses all the latest shaders, real-time
shadows (i.e. multiple light sources), among many other things I don't
understand. A top-end DirectX 9 card should be able to handle this better
than an older card that either doesn't even support these features or has
a much slower architecture (combination memory, GPU, pipelines, etc).

yes, deliberately selling crap designs (cards) combined with junk drivers,
that's why only the very fastest cards can make the show (at all), but
people paying big bux for all the middle-end crap? Fair trading? I don't
think so..when watching all those DX8/9 based 3DMark slideshows and no it's
not on a P60.

 

[TMack]

---- Original Message ----
From: "HockeyTownUSA" <[email protected]>
Newsgroups: alt.comp.periphs.videocards.ati
Sent: Wednesday, December 22, 2004 11:27 AM
Subject: Re: 3DMARK05

...

yes, deliberately selling crap designs (cards) combined with junk drivers,
that's why only the very fastest cards can make the show (at all), but
people paying big bux for all the middle-end crap? Fair trading? I don't
think so..when watching all those DX8/9 based 3DMark slideshows and no it's
not on a P60.

Most people buy 3D cards to play games. If the cards that they buy can play
the games that they want to play then what does it matter whether or not
they get decent framerates in some parts of 3DMark05? Its a BENCHMARK not a
game. If you buy an average family saloon car to drive to work, do your
shopping, visit your relatives etc and it does all of this perfectly well it
would be ludicrous to suggest that you hadn't got a fair deal because it
won't do 0-60mph (benchmark) in under 5 seconds (top of the range
performance). It would also be ludicrous to say that the benchmark was crap
because your car was much slower than top-end models.

Tony

 

[Aki Peltola]

Most people buy 3D cards to play games. If the cards that
they buy can play the games that they want to play then what
does it matter whether or not they get decent framerates in
some parts of 3DMark05?

Believe or not, but there are people who are buying new 3D
cards just to get good score in 3DMark, no matter if their
games are already playing nicely on maximun details etc.
They just want to possess the latest tech stuff.

I'm not saying that I would fully understand that, but I guess
it's pretty same deal as with tuning a car... And a good thing
for us low-end consumers too to have pioneers like that
stomping down the prices

 

[HockeyTownUSA]

Believe or not, but there are people who are buying new 3D
cards just to get good score in 3DMark, no matter if their
games are already playing nicely on maximun details etc.
They just want to possess the latest tech stuff.

I'm not saying that I would fully understand that, but I guess
it's pretty same deal as with tuning a car... And a good thing
for us low-end consumers too to have pioneers like that
stomping down the prices

They're called "early adopters", at least that's waht they called it in
college marketing or business courses. I figured I am kind of one of them
with certain things. I manily wanted an upgrade to play Doom 3 and Half-Life
2. Well got an X800 Pro because that's all I could find as XT PE's are
pretty much harder to find than a wookie's underwear. Well finally got my
X800 XT PE just several weeks ago (after a big Best Buy fiasco), and now
realized that Doom 3 and HL2 played just fine. Should have just kept the Pro
and saved a few bucks. Oh well, it's one of those things...