Radeon 9800 NP = 9700 PRO.

[iwearnosox]

I know there's been many people asking this question, and by all accounts these are nearly identical on a performance basis, stock clock. The 9800NP is seemingly *very* overclockable, beyond 9800 PRO levels if your card is the 2.8ns memory version. (There is also a 3ns version going around.)

Here's an article.

(Please don't get into semantics over a 1% performance gap, that's identical to me...)

 

[Regs]

The graphs do show that they are equal, however how can I trust the source that these benchmarks come from? firingsquad.com maybe a popular web site to some, but I've never heard of it, so how can I trust it? It would be naive of me to just believe any source out there. How do I know the author of that article isn't just some bozo who dropped out of high school at the age of 21?

Those graphs are only using UT2003. Just by using one game, I would not call that a carefully planned benchmarking experiment. It seems more like, 'hey lets just flop these cards up on a game and see what happens' . The only real differences I see in any of those cards is on the higher screen res. Then the 9800 fairs a little better. What do you expect between a 9700pro and a 9800np with a out dated game?

 

[FluxCap]

Firingsquad has been around for years just fyi but that doesn't mean the benchmark is worth a damn.

 

[Kung Lau]

FWIW, I have a 9700 np and a 9800 np, the 9700 got 13.5k on 3dmark2001se and the 9800np got 15.5k with a 2400xp cpu.
If the 9700 pro hits that neighborhood with that cpu, then it's very close.

 

[Fallengod]

Originally posted by: Regs
The graphs do show that they are equal, however how can I trust the source that these benchmarks come from? firingsquad.com maybe a popular web site to some, but I've never heard of it, so how can I trust it? It would be naive of me to just believe any source out there. How do I know the author of that article isn't just some bozo who dropped out of high school at the age of 21?

Those graphs are only using UT2003. Just by using one game, I would not call that a carefully planned benchmarking experiment. It seems more like, 'hey lets just flop these cards up on a game and see what happens' . The only real differences I see in any of those cards is on the higher screen res. Then the 9800 fairs a little better. What do you expect between a 9700pro and a 9800np with a out dated game?

You have never heard of firing squad? rofl...you must be new. One of the best most professional gamers who knows far more about FPS than anyone else works there. Id trust that site over anandtech or others. Although, im still not saying it could be flawed, but any benchmark done can be flawed or things not taken into account. Benchmarks are never 100% trustworthly.

 

[FluxCap]

Does Thresh still work there?

 

[Deeko]

Yea, you've never heard of firingsquad?

 

[Regs]

You have never heard of firing squad? rofl...you must be new.

Precisely. You really think anybody holds credit for that article except for maybe a few hardware buffs?

 

[iwearnosox]

Originally posted by: Regs

You have never heard of firing squad? rofl...you must be new.

Precisely. You really think anybody holds credit for that article except for maybe a few hardware buffs?

Right, I'd rather get some benchmarks from people here instead.

 

[Pete]

Regs, your posts in this thread alone range from uninformed to nonsensical. FS has been around a while, and their last few video card reviews have been better than most in that they're using custom timedemos.

 

[Ketchup]

Right, with all the speculation concerning nVidia's driver optimizations with the standard timedemos, it's good to see someone out there try to be objective.

 

[classy]

Firingsquad is excellent, period.

 

[Stephan28]

I've been reading ALOT of reviews of the 9800 Pro lately. So much so that they have all become a blur to me but I did kind of like this one in particular.

Linky

 

[squidman]

My boy Mr. Cow is right: 9800np = 9700 Pro. OH MY GOD, did you say it has OpenGL 1.4 as opposed to 1.3 on 9700 Pro? The truth is, 9800 np and 9700 P are so close, its not even funny. But, 9800 np got a significant advantage: its overclocking capabilities (which is mad). 9700 Pro is overclocked 9700 non pro, being sold for 80 dollars more (*thinks of an even plan: to sell oc'd 9500 Pro's as 9700 nps*). 9800 was busted out just as the news of 5900 were heard. We gots to top nVidia! - said ATi. And it did, by slightly altering the r300 core, and selling it for 200 dollars more. Gosh the corporaions are soooooo evil

Pfft. For real, guys u surprise me with how much u care about it all.

 

[Harabecw]

did you all forget about the differences between the R300 and R350's? I think the biggest is support for long shader instructions.

 

[Rand]

Originally posted by: Harabecw
did you all forget about the differences between the R300 and R350's? I think the biggest is support for long shader instructions.

I would tend to disagree, I don't see the shader length limitations being at all an impact for quite some time yet. It also seems dubious whether the board genuinely offers the performance necessary to really make much use out of complex shaders of such a length as to necessitate the FBuffer.

In any case, until the F-Buffer is actually exposed via the drivers it matters little. Presently there is no driver support for it under DirectX, and OpenGL support is restricted to OpenGL 2.0
Given the statements of Richard Huddy I havent gotten the impression there is much demand for it from developers presently, so I don't expect to see driver support incorporated any time soon.


In any case I'd tend to go with the 900NP over the 9700Pro in any case.
We've seen the the R350 core has significantly optimized the PixelShader 1.X paths, along with minor gains seen with PS 2.0
With pure PS 1.1 shading being upwars of 30% faster at identical clockspeeds, and PS 1.1 still being the primary target for most developers and the usual fallback from PS2.0 it seems a benefit well worth having.

Beyond that I'd personal suggest that the improvements in FSAA alone are worthwhile.
Ranging from roughly 5% with 2X RGMSAA, up to peaks of almost 40% faster at identical clockspeeds with 6X RGMSAA it's hardly a minor benefit.

 

[Blastman]

Right, with all the speculation concerning nVidia's driver optimizations with the standard timedemos, it's good to see someone out there try to be objective.

I quite agree, I like the steps FS has taken to use custom timedemos.

On the other hand, I haven?t seen any mention of the IQ settings used to compare the cards in their reviews. I mean if Nvidia is hacking IQ and textures to win a few benchmarks and their card does 70fps but looks like crap --- versus the ATI card that does 60 and looks great, what?s the point? The only way to fairly compare the 9800pro and 5900U is with equal IQ settings --- or at least trying to make an attempt to match IQ like what HardOCP has done in their reviews.

My impession is that the 5900U can beat the 9800pro a lot of the time if you use the lowest IQ settings (high-performance). Unfortunately, looking at some of the reviews (like the 3dvelocity one), at the high-performance settings the IQ on the Nvidia cards has been hacked so badly the screen shots (textures) look blurry and IQ is terrible.

3d (5800U) Page 5 ? ........the problem is that unless you really dial up the image quality options what you get is a final rendered screen that just plain sucks I'm afraid to say.

As I mentioned earlier though, the "Aggressive" setting really is too severe to be considered a fair option.

So having seen just how poor image quality is when the GeForce FX is set to "aggressive" I think we're perfectly justified in dropping this setting from all further benchmarks.


hardocp (5900U) The second shot above shows this even more clearly. Look closely at the shot from Serious Sam 2 where I have it circled in white. The 4X AA shot looks slightly blurred compared to the 2X AA shot.

This ukgamer test is a good example of how the 5900U beats the 9800pro in the high-performance setting, but frame rates drop quite a bit once the IQ is turned up to ? ?equalize? ? things and the 9800pro is faster.

 

[Harabecw]

Rand, I should have mentioned...I read in one of Carmack's .plans that some of the instructions (he was doing for D3, I'd imagine) exceeded R300's limit.

If complex shaders are not really practical on today's hardware, consider my point moot. I don't know enough about it.

 

[Rand]

Originally posted by: Harabecw
Rand, I should have mentioned...I read in one of Carmack's .plans that some of the instructions (he was doing for D3, I'd imagine) exceeded R300's limit.

If complex shaders are not really practical on today's hardware, consider my point moot. I don't know enough about it.

Shader complexity and instruction length are not necessarily directly related however, it is possible to make an extremely long but otherwise simplistic shader.
That said, for the most part I really can't imagine the R300 cores limitations to be at all an issue for quite a few years as yet.

 

[dfloyd]

Not sure if its been said but I think Thresh left Firingsquad more than a year ago.

 

[kylebisme]

Originally posted by: Harabecw
Rand, I should have mentioned...I read in one of Carmack's .plans that some of the instructions (he was doing for D3, I'd imagine) exceeded R300's limit.

If complex shaders are not really practical on today's hardware, consider my point moot. I don't know enough about it.

he said he bumped into the r300's limit on some expermental stuff he was playing with, nothing to do with doom3.

 

[Rand]

On another note, for those interested in Doom3 and the like the Stencil buffering fixes alone would probably make the R350 worth it for them.