***OFFICIAL*** FX5700/FX5950 Thread

[GTaudiophile]

HardOCP's FX5950 PREview. (Kyle does a GREAT job with IQ comparisons.)

FX5700 Review @ Homelan. (Thanks, Virge.)

FX5950 & FX5700 Review @ GamersDepot

FX5700 Review @ FiringSquad

FX5700 & FX5950 Review @ Tomshardware

Anand's FX5700 Review

---

DET 52.16 Drivers for Win2K/XP for those interested.

 

[GTaudiophile]

Kyle talking about the DET 52.16 drivers:

What this means is that at their default Application Preference settings the 9800XT is doing better Trilinear filtering than the 5950Ultra is with the 52.16 driver. This is the default settings folks, and no matter what you change the AF level to it will keep on using this level of filtering, even if you select otherwise in any D3D game.

 

[ViRGE]

Yay, I can talk. Our 5700U preview is up; it's not much, but it'll do.

 

[GTaudiophile]

Originally posted by: ViRGE
Yay, I can talk. Our 5700U preview is up; it's not much, but it'll do.

Thanks.

 

[GTaudiophile]

Kyle has some good comments on IQ.

Good eye, Kyle:

FX5950 flashlight.
9800XT flashlight.
See any difference?

It appears that the GFFX 5950Ultra?s 4XAA matches best with the Radeon 9800XT?s 2XAA. And 4XAA on the Radeon 9800XT just blows the rest away. NVIDIA still has a way to go before they catch ATI in terms of AA IQ at 4XAA.

AA Compared

 

[GTaudiophile]

GamersDepot: The 9600XT bests FX5700 in TR:AoD.

 

[Schadenfroh]

Originally posted by: GTaudiophile
Kyle has some good comments on IQ.

Good eye, Kyle:

FX5950 flashlight.
9800XT flashlight.
See any difference?

It appears that the GFFX 5950Ultra?s 4XAA matches best with the Radeon 9800XT?s 2XAA. And 4XAA on the Radeon 9800XT just blows the rest away. NVIDIA still has a way to go before they catch ATI in terms of AA IQ at 4XAA.

AA Compared

they still suck compared to 3dfx voodoo5 FSAA

 

[GTaudiophile]

Anand's Review Online

 

[GTaudiophile]

After testing the GeForce FX 5700 Ultra, we have been very pleasantly surprised by NVIDIA. We mentioned last week that the 5700's new architecture might help to close the gap. In fact, NVIDIA has flipped the tables on ATI in the midrange segment and takes the performance crown with a late round TKO. It was a hard fought battle with many ties, but in the games where the NV36 based card took the performance lead, it lead with the style of an ultra-high end card.

I have to strongly disagree with you here, Anand. The 9600XT wins a total of 12 benches compared to the FX5700's 16. Both cards crash in one bench, the 9600XT in Homeworld 2 and the FX5700 in TR:AoD. I can't call that a TKO. Lead with style? In the "X2: The Threat Performance 4xAA/8xAF" bench, the FX5700 wins by .4 FPS. In the "Unreal Tournament 2003 Performance 4xAA/8xAF" bench, the FX5700 wins by 2.4 FPS. Lead in style?

Leading in style is what the 9600XT does to the FX5700 in ShaderMark 2.0 @ FiringSquad!

Lastly, let's talk about value. The FX5700 costs $199 WITH a $20 rebate, meaning it really costs $219. The 9600XT has an MSRP of $199 and includes a coupon for a game worth $50! Not hard to do that math.

 

[LukFilm]

Anand's review has something screwy with the graphs on page 11. And is it REALLY that hard to run the article through a spell check? "Bearly" and "impresive" isn't very impressive on a reputable site like this one.

 

[GTaudiophile]

Dave Baumann of B3D comments on the FX5700's ShaderMark 2.0 performance.

 

[BenSkywalker]

Kyle has some good comments on IQ.

Good eye, Kyle:

FX5950 flashlight.
9800XT flashlight.
See any difference?

Actually, Kyle needs his eyes checked. Open up the screenshot where they comment on the missing lights for both the 5900 and the R9800. One board is running AF, one isn't(Gearbox's official stance is that AF doesn't work in Halo ATM). It isn't exactly hard to miss either. Also, he fails to mention in MaxPayne2 and Halo at least that the R9800 is only running trilinear on the first mip level and straight bilinear on the rest when AF is enabled. One thing to have trouble notice a filtering implementation that is blending mips together, but to miss bilinear? At least he didn't miss the failure of proper lighting with the flash light, that's refreshing. I'm a bit confused as to why he didn't run the Halo bench? Or, at the very least, try and play the game through the same way. Even the nV cards have framerate spikes and drops in the wrong spots. Well, at least it doesn't come off as biased as the GD comments, they take issue with the texture LOD but ignore the shader LOD as it favors the 'wrong' company in their comparison.

 

[Regs]

For Mid Range performance you can't argue that it does very well, however Nvidia is making you pay for that little edge in performance. For someone who loves strictly Nvidia products and wants the best performance on their system and needs, it's definitely good news.

I find it terrific though, that Nvidia finally opened up to how their cards work and what we could expect from them in the future. I was getting sick and tired of trying to read bologna from Nvidia tech- wanna - be's and their half-assed technological assumptions on this message board.

The information given by Anand shows just how much R@D Nvidia does on the backend of their hardware. Development of new and very intellectual real-time compilers, drivers, and register allocation could hopefully open up the door for what is to come in the future of games.

 

[PrinceXizor]

1. Kyle has had a beef with Nvidia IQ for awhile, so its not surprising that he doesn't like the 5700 Ultra.
2. I think, on the whole, the AT review was excellent, about the only quibble is the conclusion (as GTaudiophile pointed out). How come no one has the guts to call a draw a draw?
3. In comparing the AT vs. HardOCP reviews, remember, Kyle was including IQ issues in his conclusions, AT specifically stated that the IQ portion (a significant factor) has been set aside (other than glaring anomalies) for a future article. So it appears 5700Ultra vs. 9600XT is a dead heat without considering IQ issues.

Well...time to read some other reviews!

P-X

Extremetech review http://www.extremetech.com/article2/0,3973,1361641,00.asp
Nordichardware review http://www.nordichardware.com/reviews/graphiccard/2003/FXMania/
Techreport review http://www.tech-report.com/onearticle.x/5799

 

[Tab]

I say the are pretty even for the most part.

 

[GTaudiophile]

Originally posted by: Tabb
I say the are pretty even for the most part.

I agree, but I'd still like to see the XT with faster memory. I mean, the AIW 9600 PRO uses 350Mhz (700Mhz) RAM, so why must the XT use 300Mhz (600Mhz)?

 

[hurrikaane]

Even but with IQ going to ATi. So why pick Nvidia. Especially since it's retail price is not $200.

 

[MDE]

Originally posted by: BenSkywalker

Kyle has some good comments on IQ.

Good eye, Kyle:

FX5950 flashlight.
9800XT flashlight.
See any difference?

Actually, Kyle needs his eyes checked. Open up the screenshot where they comment on the missing lights for both the 5900 and the R9800. One board is running AF, one isn't(Gearbox's official stance is that AF doesn't work in Halo ATM). It isn't exactly hard to miss either. Also, he fails to mention in MaxPayne2 and Halo at least that the R9800 is only running trilinear on the first mip level and straight bilinear on the rest when AF is enabled. One thing to have trouble notice a filtering implementation that is blending mips together, but to miss bilinear? At least he didn't miss the failure of proper lighting with the flash light, that's refreshing. I'm a bit confused as to why he didn't run the Halo bench? Or, at the very least, try and play the game through the same way. Even the nV cards have framerate spikes and drops in the wrong spots. Well, at least it doesn't come off as biased as the GD comments, they take issue with the texture LOD but ignore the shader LOD as it favors the 'wrong' company in their comparison.

AF does work in Halo, but AA doesn't work.

 

[Markfw]

Originally posted by: GTaudiophile

Originally posted by: Tabb
I say the are pretty even for the most part.

I agree, but I'd still like to see the XT with faster memory. I mean, the AIW 9600 PRO uses 350Mhz (700Mhz) RAM, so why must the XT use 300Mhz (600Mhz)?

Even, but with FS and AA it seems the 9600XT seems to win the mid segment. I think the 9700pro is still the best bang for the buck.

 

[Lyfer]

$220 before a mail in rebate? I'm not waiting 6-8 months (which is usually how long they take) for a mail-in rebate. Refurb 9700 Pros are $220, and for $30 more I can get a 9800NP.


The only good I see in the 5700FX ultra if it brings the 9600 Pro price down to $120-$130.

Btw is the 5700 NON ultra overclockable to Ultra speeds? And how much is the Nonultra?

 

[Nebor]

I don't see how you guys can get so excited about midrange cards....

 

[1966]

Originally posted by: Nebor
I don't see how you guys can get so excited about midrange cards....

Not everybody is rich and can afford to buy high end cards.

 

[Regs]

Originally posted by: 1966

Originally posted by: Nebor
I don't see how you guys can get so excited about midrange cards....

Not everybody is rich and can afford to buy high end cards.

They even recommend you to buy a mid range video card now since by the time the more resource hogging games come out, the High end cards of today will be cheaper or the high end cards to tomorrow will run the games better.

However, that's another recommendation I disagree with (Edit) Derek. Why wait 12 months for the max performance? If you have the cash and desire, just buy the high end card. However if you're not a hardware junkie like most of us are, admittedly or not, med-range cards will be perfect for many peoples needs.


I do not think any of those med-range cards deserve the "crown" . I just think the 5700u is a step on the right track for Nvidia. I'm a little uncomfortable that Nvidia literally has to have a sit down with game developers to show how their new technology works, but what's new?! I could answer my own question there, but you get the point.

It was a highly optimistic review by (Edit) Derek that was less carefully worded unlike the other reviews. But sometimes a little optimism is warranted apart from all the other grim news they have reported.

 

[GTaudiophile]

Just to prove my point again that Anand's conclusion is seriously flawed, someone over at Beyond3D tabulated these numbers based on Anand's review:

NON AA
---------
5700 wins 10 times
9600 XT wins 6

Where the 5700 won, it won on average by 15%
Where the 9600 won, it won on average by 17%


WITH AA / ANISO
---------
5700 wins 6 times
9600 wins 6 times

Where the 5700 won, it won on average by 23%
Where the 9600 won, it won on average by 54%

 

[Markfw]

Originally posted by: GTaudiophile
Just to prove my point again that Anand's conclusion is seriously flawed, someone over at Beyond3D tabulated these numbers based on Anand's review:

NON AA
---------
5700 wins 10 times
9600 XT wins 6

Where the 5700 won, it won on average by 15%
Where the 9600 won, it won on average by 17%


WITH AA / ANISO
---------
5700 wins 6 times
9600 wins 6 times

Where the 5700 won, it won on average by 23%
Where the 9600 won, it won on average by 54%

I agree