Speculation poll: 9800xt vs. nv40(fx6000?)

[Geforce4ti4200]

Hi guys, the poll is self explanitory. How do you think the nv40 will measure up to the 9800xt?

[betterdan]

I bet it will be a little worse than the NV40. I bet people are gonna be surprised at how the 9800XT will hold up to it.

[Danik]

I put owned, although I would have preferred to put down considerably faster. Given the rumored features such as PS 3.0 support, I would think that at least feature-wise it would be a worthy purchase over the 9800 XT. Then again I'm not so sure, since games are just starting to add PS 2.0 support. I'm sure a better generation of cards will come out before heavy use of PS 3.0 is done in mainstream games.

[Lezmaka]

I think the only time they might be in the same ballpark is PS2.0 performance. If the NV40 is close to the 9800XT in overall performance, then I would consider it an even bigger disaster than NV30 was.

[-=DVS=-]

It would be realy funny and bizzare if Radeon 9800XT would outperform new top card from Nvidia , so i doubt it will happen whats interesting is how mutch improved was it did it got extra Vertex Engines added how many pipes added is it 8 or realy 16 low Transistor count makes it very suspicious, how good new AA or AF are , also R420 is suppose to be super supped up Radeon but is it only core and few vertex engines or pipes added also ? so many questions so little facts :(

But only few weeks left :D

[Geforce4ti4200]

if the nv40 gets significently owned by the r420 and isnt much faster than a 9800xt, nvidia will need to sell the nv40 cheaper than ati's r420

[EMunEeE]

I do believe the NV40 will beat the 9800XT across the board. Even though it may be close in some areas such as true DX 9.0/PS 2.0 games. Doom III, the NV40 will crush the 9800XT though. If not, the NV40 will be dead out of the door.

[Andre_hazes]

I believe the NV40 will be (slightly) faster than the R360 in PS 2 intensive games. It will own the R360 in Dx8 and OpelnGL games though.

[zoomy942]

being a hard core nvidia fan (i dont have a good reason) i hope hope hope the nv40 runs quickly... that way it can find a home in my 5900 agp slot!

[blueworm]

If nvidia cant produce a card faster in every way than 9800XT they might as well crawl under a rock and die. Because as somebody has allready said it will become an even bigger failure than NV30.
Since they are going to launch on april13. It makes me think that they have a card that is gonna eat the 9800XT for breakfast...

[Paul]

Seriously, what is the point in this poll?

[Evildeus]

I don't know.

[silence]

this one is weird....if it was r420<>nv40 i would understand the need for poll, but there is no way Nvidia would release nv40 which might be slower then current top ATi.

pointless poll if u ask me.....

[Geforce4ti4200]

wasnt the fx5800 slower than the 9700 pro? It was released way after the 9700 pro as well. Nvidia quickly released the fx5900 which is basically the fx5800 should have been in the first place, 256 bit ram and a core that runs cooler. considering a 9800xt is way faster than the fx5950, the nv40 is gonna be an improvement all right, but is it enough to own a 9800xt in every benchmark or in just some benchmarks? thats the big question

[silence]

wasnt the fx5800 slower than the 9700 pro? It was released way after the 9700 pro as well. Nvidia quickly released the fx5900 which is basically the fx5800 should have been in the first place, 256 bit ram and a core that runs cooler. considering a 9800xt is way faster than the fx5950, the nv40 is gonna be an improvement all right, but is it enough to own a 9800xt in every benchmark or in just some benchmarks? thats the big question


fx 5800 was late, but it IS same generation as r300(9500/9700). changes made between nv30>nv35 can't dispute that fact.

IMO, compering nv40 to 9800xt is just like compering fx5800 to radeon 8500....useless ;)

[zakelwe]

Seriously, what is the point in this poll?

Yours is not to reason why, yours is just to give a big sigh.

Regards

Andy

[Nutty]

Then again I'm not so sure, since games are just starting to add PS 2.0 support. I'm sure a better generation of cards will come out before heavy use of PS 3.0 is done in mainstream games.

The thing is tho, as we're actually seeing, is once a game supports PS2.0, its a minor change to support PS3.0. All you basically have to do, is compile down your high level shaders to PS3.0 instead of 2.0. Quite simple really.

The stumbling block is PS2.0, as its quite different from previous pixel shader versions, and is the 1st to support high level shader compilation. (Unless you count Cg's compilation to PS1.0+ )

[Razor04]

The thing is tho, as we're actually seeing, is once a game supports PS2.0, its a minor change to support PS3.0. All you basically have to do, is compile down your high level shaders to PS3.0 instead of 2.0. Quite simple really.
What is the advantage of doing this though? Is it just being able to say that your game is PS 3.0 and not just PS 2.0? I fail to see where any speed gains could be realized by doing this. Especially when the only IHV supporting PS 3.0 (according to rumors) has a tendency to introduce hardware that doesn't allow the use of those features at a reasonable speed. If this is the case again this generation then re-compiling to PS 3.0 might actually cause a slowdown not a speed increase. Maybe it will be different this time around, but I am not getting my hopes up for useable PS 3.0 in the NV40. I also don't see how recompiling to PS 3.0 would improve quality in any way. It seems to me that recompiling is just for marketing purposes at this point in time and that any differences between a PS 3.0 path and a PS 2.0 path in a game will be very minor.

[Woodelf]

I find it hard to believe that Nvidia isn't aiming above ATI's current. I no longer judge performance by clock/feature speed speculations on untested part's (P4 vs athlon/5700 vs 9700).

[The Hip]

The NV40 will pretty much dominate the 9800XT and probably also be both faster and more feature rich than the R420/423.

From what I understand - the R42X series is built on the tried and tested R3XX foundation and featurewise only offer limited "extras", whereas the NV40 is a new architecture, with more pipelines and a whole truckload of "extras".

There's only so many times you can extend and refine a tried and tested foundation - and come out a winner.

[GlowStick]

What is the advantage of doing this though? Is it just being able to say that your game is PS 3.0 and not just PS 2.0? I fail to see where any speed gains could be realized by doing this. Especially when the only IHV supporting PS 3.0 (according to rumors) has a tendency to introduce hardware that doesn't allow the use of those features at a reasonable speed. If this is the case again this generation then re-compiling to PS 3.0 might actually cause a slowdown not a speed increase. Maybe it will be different this time around, but I am not getting my hopes up for useable PS 3.0 in the NV40. I also don't see how recompiling to PS 3.0 would improve quality in any way. It seems to me that recompiling is just for marketing purposes at this point in time and that any differences between a PS 3.0 path and a PS 2.0 path in a game will be very minor.
Every generation of PS allows you to do 'cooler' effects and the same effects faster.

For example, in HL2 there are PS1.3 shaders that can be ran in one pass via PS2.0's new instructions. as gabe stated at ati shaderdays. So since shader code for PS2.0 is written in HLSL, it should be almost like compiling a program.

[Richthofen]

What is the advantage of doing this though? Is it just being able to say that your game is PS 3.0 and not just PS 2.0? I fail to see where any speed gains could be realized by doing this.


the speed is realized because you safe loopbacks and so on.
You can prosess the shader in 1 run while with PS2.0 you might need a second run - of course depending on the shader.

Further on it is a marketing feature which will sell products and which is easy to implement in games.

Finally there is nothing like a slow PS3.0 product. The point is a PS2.0 run in 3.0 mode won't be slower so you can only gain something.
On the other hand a 3.0 shader will be slow in 2.0 mode or can't be run in that mode.
Because the competition is lacking in the PS3.0 department there is no possibility to compare Nvidia's PS3.0 speed with long and complex shaders because they are the only company so far supporting that feature. With about 69% DX9 market share and a pretty easy way to implement that feature in real software the chance that it is a nice advantage is pretty high.

[Danik]

The NV40 will pretty much dominate the 9800XT and probably also be both faster and more feature rich than the R420/423.

From what I understand - the R42X series is built on the tried and tested R3XX foundation and featurewise only offer limited "extras", whereas the NV40 is a new architecture, with more pipelines and a whole truckload of "extras".

There's only so many times you can extend and refine a tried and tested foundation - and come out a winner.

After reading some things in regards to GDC I'm more inclined to agree that the NV40 will compete very well, if not surpassing the R420 in performance. Popular opinion is that the R420 will not support PS3.0, while the NV40 has been confirmed. I'm also more inclined to believe that the NV40 is truly 16x1, and not 8 virtual pipes + 8 true pipes. But then again who knows. Some sites claim their sources say 8, while others say 16. I think performance will be good based on GDC reports. Carmack obviously has reason to say he likes the NV40 because of TWIMTBP and Doom 3, but the reports of the performance in Unreal 3 seem to be a good sign. I'd imagine there was at least heavy PS2.0 going on and more likey PS3.0 since the game is slated for 2006(could be wrong on this date). The card is apparently no slouch at fill rate as well, since it was rendering high polygon characters

Edited because I was wrong in the amount of polygons.

[Lezmaka]

The game doesn't have 6 million polygon characters, at least not as rendered. It's similar to DOOM III. The original character models may have millions of polygons, but the final model that is rendered only has a few thousand, and each frame averages about 100,000 polygons. Even if NV40 has 16 true pipelines (or R420 for that matter), there's no way it can render 6 million polygons and have a usable framerate.

[Danik]

The game doesn't have 6 million polygon characters, at least not as rendered. It's similar to DOOM III. The original character models may have millions of polygons, but the final model that is rendered only has a few thousand, and each frame averages about 100,000 polygons. Even if NV40 has 16 true pipelines (or R420 for that matter), there's no way it can render 6 million polygons and have a usable framerate.

Thanks for correcting me, but even if it is 6 million prerendered isn't that something to look forward to? If I'm not mistaken the statement in the article I'm referring to ended with a "!".