5800 U vs 9600 XT: both 4 pipeline cards with a big gap in performance.

[ronnn]

Originally posted by: shady06

Originally posted by: ronnn
I think the gap in performance is quite small.

between the 5800U and the 9600XT? the 5800U is substantially faster

As compared to what? They are both low end midrange cards.

 

[rbV5]

5700u vs 5800u, 4x1 128bit vs 4x2. 128 bit.

 

[nRollo]

The 5800 was never meant to compete with 9600s, it's meant to compete with the 9700Pro. At most games it does so pretty well up to 12X10 4X8X.

The VGA Charts 3 at THG are interesting to me in that they don't exactly make sense when compared with his other reviews:
5800U beats 9700Pro at every resolution in original review at UT2003 4x8x
On an Athlon 27000, Antalus Flyby, Tom has the 5800U at 76fps and the 9700Pro at 68fps

But check out his next review
On the same computer, the 5800U is still at 76fps, but now the 9700P is up to 86fps?! Nice 21% performance increase for the ATI card, eh?

Then in his next review the 5800U is up to 95fps at Quality settings and the 9700P still at 86
Wowie- huge gains for nVidia this time around! LOL
(still on the same system)

WTF?!?!?!?!?!?!
On a much more powerful system, a P4 3.2, the 5800U has now dropped to 54fps and the 9700P has dropped to 67. (although I suppose Inferno is a much more demanding level than Antalus, and who knows what he put in the timedemo)

 

[Pete]

I think the 5700U is more accurately classified as 2x2, though apparently it can operate as 4x1. So it's more like 5700U:5800U::2x2:4x2.

 

[nRollo]

Originally posted by: ronnn

Originally posted by: shady06

Originally posted by: ronnn
I think the gap in performance is quite small.

between the 5800U and the 9600XT? the 5800U is substantially faster

As compared to what? They are both low end midrange cards.

While it is true the 5800U is a a midrange card now, it still significantly outperforms the 9600XT in over 90% of the games at over 90% of the settings out there. The two are not even in the same class.

Yes you can find a couple games that make use of shader power the 5800U isn't strong on, but for most games a 5800U walks all over a 9600XT. I've never had a card as crippled and slow as a 9600XT, but I know from personal experience a 5800U is as good as a 9700Pro.

 

[SickBeast]

NM. I'm tired of the video forum flame wars.

 

[BFG10K]

Hardly any DX8 games include shaders, and even then, there are only small bits of shaders used for water.

And lighting and vegetation.

That's true. 8x1 is = to 4x2.

Not always.

ATi's performance in CoD is strange, considering nV is usually better in OGL,

It's extremely likely to be shaders causing it. The 9800XT absolutely tools the 5950 in that game.

You can't have two pipelines work on the same pixel.

I was referring to texels. With a base texture + lightmap couldn't one pipe fetch the base texel and the other fetch the lightmap texel and then combine those at the end?

 

[LocutusX]

interesting points, although when I generally think of comparing cards I look at things from the price-point perspective. where I am located (toronto), GeforceFX cards are both scarce and more expensive than ATI cards, so 5800U-based cards end up going for the same price as BBA 9800Pro's -- a no brainer comparison.

most 9600XT's here go for the same price as 5700 (non-U) based cards.


although i understand the point was the hardware similarity between the 5800U and 9600XT...

 

[Pete]

I was referring to texels. With a base texture + lightmap couldn't one pipe fetch the base texel and the other fetch the lightmap texel and then combine those at the end?
I don't think so, but I'm not 100% sure.

 

[Corporate Thug]

Originally posted by: ronnn

Originally posted by: shady06

Originally posted by: ronnn
I think the gap in performance is quite small.

between the 5800U and the 9600XT? the 5800U is substantially faster

As compared to what? They are both low end midrange cards.

to be absolutely clear, the 5800U is substantially faster compared to the 9600XT

 

[LocutusX]

the 5800U was more comparable to a 9700Pro, am I right? So let's say about 5% slower than a 9700P... which means about 20% slower than a 9800P...?

definitely, not even in the same class as a 9600XT.

 

[Blastman]

It?s pretty hard to get a ? ?apples to apples? ? comparison with the FX line to the 9600 -- 9800 with the FX?s running much lower IQ. The FX?s run lower shader paths, Brilinear, 4AA that?s not near as good as rotated-grid 4AA on the ATI cards, etc,etc.

 

[nRollo]

Originally posted by: Blastman
It?s pretty hard to get a ? ?apples to apples? ? comparison with the FX line to the 9600 -- 9800 with the FX?s running much lower IQ. The FX?s run lower shader paths, Brilinear, 4AA that?s not near as good as rotated-grid 4AA on the ATI cards, etc,etc.

Are you talking about them not using 24 bit PS2 on the .01% of games that support it?

Are you talking about the brilinear that the 9600 uses as well, even though ATI never acknowledged it? You know, the one that allows the user to see he transitions between mip bands as they're moving around?
Of course, Anand himself said the FXs have better AF:
<a target=_blank class=ftalternatingbarlinklarge href="http://www.anandtech.com/video/showdoc.html?i=1821&amp;p=14">
http://www.anandtech.com/video/showdoc.html?i=1821&amp;p=14</a>[/L]

The tables have turned and now it's ATI's turn to play catch-up and make their performance mode look better

(I see what you mean, you can't really compare ATI to nVidia, it's worse)
or how about this:
http://www.anandtech.com/video/showdoc.html?i=1931&amp;p=14

The decision on what is acceptable is out of our hands, and we can't really declare a clear winner in the area of image quality. We can say that it appears from the tests we've done that, generally, NVIDIA hardware does more work than ATI

Of course, in the grand scheme of things, there's no real point in comparing a cheap, crippled card like the 9600XT to a real video card like the 5800U. The 5800U would stomp it's 4X1 architecture flat in all but the 3-4 DX9 games out there. I don't think any hardcore gamers are buying 9600XTs, it would be too frustrating!
The 5800U was a usable alternative to the 9700Pro at launch, as good as it got.
The 9600s are just bargain cards.

 

[Blastman]

Anandtech graphic card reviews are a joke. They couldn?t do a proper IQ comparison if the IQ truck ran over them.

A good example of this apples/apples IQ comparison would be the nordic ?mid end shootout review.

Against an 5900XT the 9600XT ?

wins 1
ties 5 ???(within 5% either way)
losses 5

So the 5900XT scores a clear victory. But?

The 5900XT is running Brilinear -- probably a 5% gain. The 9600XT adaptive-Tri is much better.
The 5900XT is running a much lower 4AA than the 9600XT -- probably a 10% gain.
The 5900XT is running lower shaders in Halo.

If you add 15% to the 9600XT scores those ties turn into wins, and one can consider Halo a win, the overall complexion of the review changes substantially. For the 9600XT we now have ?

wins 6
ties 1
losses 4

An overall victory over the 5900XT if IQ was matched.

 

[BFG10K]

I don't think so, but I'm not 100% sure.

Yeah I'm not sure either. I suspect the answer depends on what loopback can and can't achieve.

 

[Cerb]

If IQ is matched w/ AA, the 9600 gets a run-away win.
The AF is suprising, actually. Even 4x AF seems to come out and grab me eye. The reviews talking of NV'a AF being better are right!
The AA, too, though. Rotated grid looks SO much better. Luckily shaders and such are finally making AA less of a concern. Since I got Gunmetal w/ my card, I decided to play it...and to be honest, 2x AA was enough. I am impressed (with the visuals--the game itself is crap). I'll see about Aquamark and such if I can (tired, need sleep, etc. right now).

I don't know about a 5800U, but for a 5900XT, you get good performance in older games, and in the worst case for newer ones that are in the works, you get 9600 XT performance, based on current numbers. So if you don't have $300+ and aren't pleased w/ ATI, it's a solid buy.
If you have the cash, or have a GF4 right now, though...wait for the vanilla and GT 6800.

 

[nRollo]

Originally posted by: Blastman
Anandtech graphic card reviews are a joke. They couldn?t do a proper IQ comparison if the IQ truck ran over them.

A good example of this apples/apples IQ comparison would be the nordic ?mid end shootout review.

Against an 5900XT the 9600XT ?

wins 1
ties 5 ???(within 5% either way)
losses 5

So the 5900XT scores a clear victory. But?

The 5900XT is running Brilinear -- probably a 5% gain. The 9600XT adaptive-Tri is much better.
The 5900XT is running a much lower 4AA than the 9600XT -- probably a 10% gain.
The 5900XT is running lower shaders in Halo.

If you add 15% to the 9600XT scores those ties turn into wins, and one can consider Halo a win, the overall complexion of the review changes substantially. For the 9600XT we now have ?

wins 6
ties 1
losses 4

An overall victory over the 5900XT if IQ was matched.

Hmmm. I've never thought Anand's reviews were a joke; this is one of the places I trust for good information on hardware.

As for your "well I think we have to add this subjective percentage to all the 9600 numbers because I think that would be right" logic.....errrr....yeah.

The fact of the matter is you'd have to be pretty much nuts to prefer a 9600 over a 5800/5900 anything.

It's like saying "There are 3-4 games the 9600 can tie the performance of those other cards on, and it will run all the other games in the world much slower. The 9600 is the card for me!"

I've had 5800s, and 9700P/9800P, used them all for many months. These "huge" differences in IQ you talk about simply don't exist. Remember, that is between the real ATI cards that don't have forced, hacked cheater drivers like the 9600XTs.

You're entitled to prefer a 9600XT, but for me, I could never use a card that slow. (even if it does tie some faster cards at DX9 PS2 that neither of them is capable of running at a decent framerate- no one with a 9600XT is going to be running HL2 in PS2 unless they want to watch those bugs do their impression of break dancing)