Official Forceware 56.64 drivers

I already told you what it accomplished and I even took the extremely unusual measure of backing up my statements with evidence. I suggest you re-read my post until understanding sinks in. Or I could save you some time and restate the conclusion right here: "3dMark03 has shown the underlying capabilities of hardware to run pixel shaders."

Hmmm.... that's odd. The only use I've had for 3dMark03 is to see how hardware will handle games of the future. From it, I learned that one of the two major architectures will handle shaders easily without requiring any extreme tweaking or optimizing. And I learned that the other main architecture can only run shaders fast if there's lots of work put into "optimization" efforts.

Trouble is... we're not analyzing the cpu scene here. Whatever's going on there may connect on a tangent with the current subject, but that's it.

With the latest drivers, eh? I can sum up the situation just by saying Application Specific Optimization". Oh, and I almost forgot "Brilinear".
As has been already noted, the most performance intensive shaders (Glow and Depth of Field) were turned off for both parts. I'll note that one of these shaders (glow, iirc) had some issues with it that caused bad inefficiency. One of the two architectures got decent fps anyways but the other didn't. I'll let you guess which card could handle it and which couldn't.

btw, here's another difference which I don't think anyone mentioned yet: Guess which card had ps 2.0 shadows disabled?

It's true that the radeons are often twice the speed of fx's at pixel shading... but there's more then just pixel shading going on. Other factors come into play.

Almost. We're to the point in the debate where you're supposed to say something about how "TR:AoD was a crappy game so it doesn't count!" At least, that's what many people who've share your opinions re. 3dmark03 have done...

Anyhow, I think the points I made above about synthetics in general and 3dMark03 in particular have been supported. 3dMark03 has successfully predicted how hardware will handle dx 9 games.

 

Glow has been removed from the latest patch, so testing it is kinda moot. Unless you're trying to use TR:AoD specifically as a snythetic benchmark.

How many shipping games using PS2.0 don't use PP? How many announced games using PS2.0 don't use PP? AFAIK, every single game released has used PP. If NV4x still has PP I don't expect to see PP any time soon.

Hmm, I missed that earlier. Good catch.

 

While I dont think 3DMARK2003 is the best benchmark.. mainly because according to it, a 1gig+9800XT is faster than a 3gig+9700 Pro, it does help somewhat.

It told us a year ago that the NV3x is a poor PS 2.0 card. Today we see that is true, even with the limited amount of PS 2.0 titles out there. Farcry is a huge game, and the fancy DX9 effects of PS 2.0 cant even be used by a NV3x card unless forced with a 3rd party program.

Code:

Vertex Shaders version 2.0
 Pixel Shaders version 2.0
 Use Hardware Shaders for ATI R300 GPU
 Pixel shaders usage: PS.2.0 and PS.1.1
 Vertex shaders usage: VS.2.0 and VS.1.1

Code:

Vertex Shaders version 2.0
 Pixel Shaders version 2.0
 Use Hardware Shaders for NV3x GPUs
 Pixel shaders usage: PS1.1 only
 Vertex shaders usage: VS1.1 only

Yes thats from the retail game. Whats the point of having a "DX9 card" if you cant have DX9 features in some games? Wish I had known this before I got this 5900.

 

If you are trying to test purely the card and not the CPU, then that's correct - a 9800XT is faster than a 9700 Pro.

 

I understand that, but

I dont think its marketed very well. The score reflects the GPU, but their words do not.

 

Code:

 Vertex Shaders version 2.0
 Pixel Shaders version 2.0
 Use Hardware Shaders for NV3x GPUs
 Pixel shaders usage: PS.2.0 and PS.1.1
 Vertex shaders usage: VS.2.0 and VS.1.1
 Shadow maps type: Mixed Depth/2D maps
 Stencil shadows type: Two sided

Not sure if you are making a "political" statement or just aren't sure what the in-game options do but that's from my 5900nu with all settings at "Very High", I get the same as yours above when using "High" settings.
Funny thing, to play at "Very High" both my 5900nu/56.64 and 9800P /CAT 4.3 have to be lowered from 1280x960 to 1024x768 or the fps is laggy. At the lower res fps is smooth even with AA/AF on.
Now I know that the 9800 is technically better at these types of games but actual exp with this game seems to indicate otherwise.


Mike G

 

Uh, why would I be trying to make a "political" statement? Dont assume.

I did have it set to "Very High", it pops up the first time and sets your video. I played for a few mins to make sure the settings "took". Looked at my log, and thats what it showed. Doubled checked the options, its set to extra high.

Im using the 56.64 drivers, and a BFG 5900NU. If there is another reason why its not using PS 2.0 Id like to know, especially since yours seems to be.

edit, for some reason it took 2 attempts at restarting the game to get the log to show PS 1.1 and PS 2.0 :shrug: Seems to be working fine now.

 

And i suggest you re-read mine...if 3DMark was any good in judging PS 2.0 and future DX9 performance then anATi Radeon XT should score about twice as much as a 5950...more on that in while

You use 3DMark to see how hardware will handle games of the future ? It sounds so easy to put a million holes in the above statement, it's not even funny. First of all, we and they (FM) should point out that it's a DX9 benchmark - and there is still OpenGL (thank god) so i will repeat myself, if 3DMark was any good in predicting "future DX9 game performance" then a 5950 should be 15% slower than a 9800 XT in DX9 games (with aproved drivers etc.). But if we look at Far Cry (PS2.0 in both cards), TR AoD (before they remove glow and with DoF enabled) and HL2 (FP)we see that the 5950 is actually on average 100% slower...so if 3DMark was so good the scores should reflect that.
Capish ?
IMO 3DMark did a very poor job in what you say it's intended to do (although i don't think that's it really - can you spell MARKETING TOOL)

Synthetics are ok most of the time as indicators for "specific situations", but when you start making gross generalisations based on synthetics it can get very ugly.

oh really ?? I guess then all the above examples/future games only use "pixel shading" and no more "stuff" is going on in the background...that's way NV3X is so slow

I now see it's impossible for you to understand me, i guess my grammar is a tad bad, but not that bad. Read my posts again. You act as if i said the NV3X is faster and is penalised in 3DMark by FM not aproving latest NVIDIA drivers when in fact i say the FX sucks in PS2.0 AND 3DMark is a bad benchmark (notice the AND...you act as if i am using OR). If you still fail to see the points i make i am sure you will be able to find a lot of good reads over the net (3DCenter is a good place to start and as is this forum).

Sure whatever

 

When the ps2.0 test is not effected by app specific optimizations it does show the gffxs being half as fast as the radeons (based on 51.75s anti detected results)

When gauging pure dx9 performance you really have to just look at gt4 as gt2/3 are dx8.1. In nature with approved drivers ATI cards are generally atleast 25% faster although you are right in that even that doesnt really sound large enough considering the test uses ps2.0 and vs2.0. Maybe Nvidia is right and the test does use too much ps1.4 instead of ps2.0

 

How many times do I have to POST THIS

2X you say eh ?

5950 Forced to PS 2.0 using 3Danalyze



9700 Pro Ps 2.0



Frame counter is in the right hand corner, and you were saying

 

It would help if it were the same scene.

 

Fixed , it doesn't really matter though.

 

dp

 

Can you do the same thing with an outdoors scene that has a lot more geometry & draw distance involved?

 

And your point is ?
If you ever read the whole post you would see that you just added more evidence ti support my point. Radeon is almost twice as fast as 5950 in heavy PS2.0 apps.

 

Sorry, i ll have to disagree. The PS2.0 test does not count in calculating the final score. Only the four game tests count, and with 52.16 FM aproved drivers a 5950 is on average 15% slower than a 9800 XT.

 

It's telling when the game's installer defaults to PS 1.1 for certain boards. That alone should tell you exactly what developers themselves either think or are recommended to do.

 

Well I see what your saying, but 3d mark 03 couldnt be that biased towards future gaming otherwise it would of disilusioned dx8 owners and be even less relevant to games of the time. 3d mark 03 may not be a good indicator of the full dx9 games that Id guess will appear in 2005 but it will likely do an ok job of indicating performance in primarily dx8/8.1 games with a bit of ps2.0/vs2.0 i.e. many of the games coming out this year, not really sure of many good examples, and saying doom 3 would be silly seeming as thats opengl.

 

Even when it is in black and white...the truth exposed and still arguing. I used to love these type of threads...but not anymore. It is like talking to a pet rock

 

Oh c'mon Doomtrooper, you live for this shit, and with the new big card's coming out soon, you should be priming up for full asshat mode...