3DMark 2003

[Scouse]

Fucking CRACKING!!!!!

Just installed it on my laptop:

1.2GHz P3, 64 Meg ATI Mobility Radeon 7500, 256Meg Ram.

I got.............................................






SEVEN 3DMarks!!!!!!!!!!!!!!!!!!!!!



My laptop ROCKS!



(I'll have to wait for my alienware rig to turn up before I can see what it's supposed to do then.... it won't even run the demo )

 

[kan]

Originally posted by Scouse




Does it really come up with that or was it a throwaway comment??!! - I'd be surprised....


If so then it's seriously flawed.

Yup it’s flawed

What durz said covers it really


I think for benchmarking the best tests are the inbuilt ones in games.

If the games you play bench ok then stick with the card you have, if they don’t and it’s noticeable and it bothers you upgrade


Unreal2 seems the best game to stress a system at the moment. Made me upgrade from 256mb of memory to 512 because hard drive was swapping on big levels

 

[Durzel]

<flame>

It's only flawed if you have an obsolete system. i.e. a Pentium 4 and/or an Nvidia graphics card.

</flame>

 

[Sawtooth]

Originally posted by kan
as a benchmarker i think its crap tbh..


saying a celeron 900 with a radeon 9700 is a better gaming machine than a p4 3 gig with a 4600 in it is flawed imho.


plus im gutted my system only got 4660 3dmarks

So did mine but the point is surely that you need a tougher test. Its no good producing a test where a current cutting edge machine tops out. You have to build in a higher requirement to allow for the next CPU/graphics card. I reckon 4500 is pretty much where we are at the mo. Maybe an Alienware rig will will get 5000?

 

[Durzel]

You're saying the fact that people get 1-2fps max on the CPU Test 2 when running P4 2.4 Ghz processors is a too easy test?

Most of the benchmark slaughtered my system, and I've got a 1.5Ghz (ok, thats not cutting edge anymore) and a Rady 9700 Pro.

 

[Munkey-]

hmm, wonder how my

AMD Athlon 600 Mhz
Hercules 4500 prophet 64mb
3hundred and something RAM

will go down

 

[Scouse]

Maybe an Alienware rig will will get 5000?

I'll tell you next week when mine arrives


Durzel - as I understand it, it works thus m8:

The 4 tests are split into:

Test1) DX7 test
Tests 2 & 3) DX8.1 tests
Test 4) DX9 Test


You'll prolly max out your CPU on test 1 because it's the most simple of the tests by far and all the modern graphics cards can handle DX7 calls really efficiently.

For the second two tests you'll do OK with the Nvidia line of cards (especially the TI4600) although theres a problem with bottlenecks with Nvidia cards at the vertex shaders (so I understand)....

For the fourth and final test you'll find that your nvidia cards that aren't the daft looking FX won't work - they're not DX9 cards.....


The 3dmark scores are based largely on Graphics card performance (although CPU does have a bearing when similar graphics cards are used).

If you've got a 3.06Ghz P4 and a Geforce2MX then you're going to find that you can't run most of the tests - and if you can't run them then you score a ZERO for them.

However, if you've got a P3-733 and a Radeon 9700 then you can run all the tests - and get the appropriate scores. - Replace the CPU with your top of the line P4 and you'll get a very good score....


Remember - this product has been designed for benchmarking for the next TWO years - and it's only just come out. It's been designed for DX9 cards because from now on nearly all new card releases are going to be DX9 parts.

Simple eh?



But it does piss me off that I only get 7

 

[kan]

Scouse this is supposed to help people decide if they need to upgrade there systems..

And as it runs like a dog on my pentium4 @ 2.7 with a radeon 9700 pro then i think its fair to say its flawed.

I was getting 1 fps for parts of 1 test, and i think its safe to say that no game on the market would do that to my system.

 

[kan]

just noticed the cow smiley and i must say it owns

 

[Will]

Originally posted by Scouse


I'll tell you next week when mine arrives

I swear this isn't to do with that NTL thread, I'm not picking on you, but I thought the Alienware PCs were a fair bit overpriced for what they are...think thats fair?

 

[Testin da Cable]

I'm dling it atm. curious. for the stat hungry, it's taking me about three minutes

 

[Testin da Cable]

2000 points. it ran like a dog. a dog that was tied by the danglies to a large, unmoving object :/



*repress urgent need to upgrade*

 

[leggy]

Originally posted by Testin da Cable
*repress urgent need to upgrade*

nooooo!!!

doooon't do it ffssssss

NWN runs fine and your pc will own U2....


In otherwords :- I can't afford to upgrade right now so I don't want you to either.

 

[Testin da Cable]

that is indeed so: my poota owns most games I play. still, I conned a mate into buying it off me heh. I'll list my (planned) new components and let you take potshots at em in the hardware forum if I have time today.

 

[Scouse]

Scouse this is supposed to help people decide if they need to upgrade there systems..

No, Kan, it isn't!

It's supposed to show up relative differences in different pieces of hardware. It's been designed so it really punishes todays hardware because in the next two years hardware will come out that breezes through it.

If you base your decisions to upgrade on a benchmark then you're a fool to yourself! Does your PC play your games oK? Yes?? Fine then...

I can get Unreal2 to run acceptably on my laptop - but it gets fuck all on 3DMark2003..... because it's not a DX9 part...

 

[Scouse]

I swear this isn't to do with that NTL thread, I'm not picking on you, but I thought the Alienware PCs were a fair bit overpriced for what they are...think thats fair?

LOL Will

Yes, I do think they're overpriced - but I'm putting in an expense claim with my company for it - and there are loads of tax reasons to do with why I should by one - so I did.



However, if Alienware decide to change my specs from 2800+ Athlon XP, 1 Gig RAM, 400Gig hard drive, Radeon 9700 Pro and the 5.1 Klipsch Promedia Speakers, after I have paid then you bet your bottom dollar I'll moan all over this place - NTL stylee

 

[Dimebag]

So when radeon optimised code comes along in the form of DX9 compliancy and ati's cards start beating the pants of nvidias everyone throws a fit.

When p4's are compared to athlons with rambus memory instead of ddr and sse2 optimised code for floating point instructions no one says anything.

At the end of the day the tests might not be strictly "fair" but thats just how it is.

Why do you lot care so much about a 3d benchmarking program anyway? Anyone who bothers about the figures that much is just plain sad.

I have an overclocked athlon 2400+ at 2.4 Ghz. My machine still only scored 845 3d marks because of the gf3 ti 200 in it. Like I care? It still plays anything under the sun fine in 800x600 and even 1024x768.

 

[Scouse]

Exactly Mr. Bag

 

[~YuckFou~]

I don't use 3dmark and I'm not interested in what marks I might get. It seems to me the only reasonable use for it would be to compare your marks before and after some tweaking/over-clocking/fitting new kit to see if any improvement has been made.

 

[Scouse]

Indeedy Mr.Fou

 

[Wij]

Never used 3DMark. Why bother ?

 

[lecter]

For a start it is shit benchmark that isnt representitive of most games.

My idea of a decent benchmark system.
5 Top Games
1. Quake3
2. HalfLife
3. Unreal Tournament 2003
4. Serious Sam 2
5. Jedi Knight 2

each game would output the following info.
max fps/min fps/ave fps/standard deviation
(much like the unix ping tool output)
This would give a sense of stability as well as performance.

The results are then normalised to a average specification machine, so people have a real sense of how many times more performance they get.

 

[Scouse]

5 Top Games
1. Quake3
2. HalfLife
3. Unreal Tournament 2003
4. Serious Sam 2
5. Jedi Knight 2

1) Quake3 - yes - it's hardly a stress on new graphics cards these days - so it's a good CPU test.

2) HalfLife - why?? It runs great on everything and we already have a CPU test.

3) UT2003. Why not - a good DX8.1 benchmark

4) Serious Sam 2 - can't see the point of this really either

5) JK2 - it becomes CPU limited at lowish resolutions so again, I can't really see the point....


We need a good DX9 benchmark game - I presume that it'll end up being Doom3. Which is a bit of a shame because if you read through the John Carmack .plans it seems that he's putting a bit more effort in on the Nvidia side......

 

[Dimebag]

I feel its time to point out, I couldnt help but notice you're all totally gay.

 

[Happy Go Lucky]

Instead on downloading 3dmark I installed FRAPS yesterday, not as easy to use as 3dmark. But I eventually loaded RTCW from the beginning ran around shooting a few guards. Exited and looked at the log file in the FRAPS directory and got an average fps of 57 @1028x768 4x AA. Can't wait to try it on other games.

 

[Dimebag]

I can't wait for the results either! Im so excited for you!!!!!11

 

[Xavier]

Originally posted by Durzel
It's hardly a surprise that nVidia denounce it when they've been routinely shat on in the tests by ATi for the past 6 months.

Thing is, if its using DX primitives, and runs the same on all systems - surely its an empirical benchmark? It's a bit like saying a rolling road isn't an accurate measure of a cars power because one has an analogue display to show the BHP, and the other has digital.

One may be more accurate than the other, but if you compare all systems on the same thing - it's still an "accurate" (in terms of consistent inaccuracy) measurement.

Only a few things to say here really, but they're all pretty important.

The first 3DMark arrived at a time when most games didn't have easily accessible benchmarking options, and those that did were either all of the same engine (unreal/quake or later generations of) - nowadays, a much larger proportion of our games offer some form of metric - so synthetic benchmarks are no longer necessary.

As to the line about them using DX primitives, even with clean DX8.1/DX9 code it's suprisingly easy to write something biased to one platform or another. One good example throughout 3DMark'03's DX8/8.1 shaders. Nearly every DX8.1 game out there today offers pixel shader 1.4 based code first (i.e. Radeon 8500) and falls back to PS1.3 for NVIDIA hardware and PS 1.1 for the remainder. With each step back poly count increases, as does the overall complexity of the scene - it's done this way to make sure each card can run as quickly as possible. In 3DMark'03 however the benchmark only uses the rather ATI-centric PS1.4 and reverts to the uber-compatibility (and complexity) of PS1.1 - totally skipping the paths that NVIDIA hardware would normally use in real world games... and forcing it down the 1.1 path which renders 4x as much in most cases (an increase from 125M to almost 500M).

3DMark2001, for it's credit didn't take such shortcuts, it did things properly - so why can't '03?

If you want the full skinny on the flaws in the benchmark check out the thread I posted here at the time the benchmark was released.

 

[Scouse]

Aye Xavier - but don't forget that until last December Nvidia were one of the design partners of 3Dmark03.

It kinda beggars belief that Nvidia would go along with the design team for so long without pointing this out don'tya think.

It's also suspicious that they pull out a matter of weeks before launch.

And Nvidia cards from now on will support the DX9 stuff - so I think it's really more about 3dmark getting great scores on the current ATI lineup.....

 

[Wij]

Too much info Xav

I'll stick with my GF3...

 

[kan]

Originally posted by Scouse

quote:
--------------------------------------------------------------------------------
Scouse this is supposed to help people decide if they need to upgrade there systems..
--------------------------------------------------------------------------------




No, Kan, it isn't!

ok what is a benchmarking program actually for then? people will run this benchmark to see if there 3d is up to scratch. if it isnt the choice is theres over whether they upgrade or not but if somones that arsed to download a 177mb benchmark i would guess they will "worry" about there score....




i might be talking bollox tho cos im a mess