NVIDIA NV40 GeForce 6800 Ultra Review

They couldn't agree on money.

 

Hope this is not too far off topic but with Nvidia releasing a solid chip - doesn't it make you wonder why Microsoft went with ATI in regards to the Xbox2?


I think it is painfull obvious why they went with ATI over Nvidia. Nvidia did not roll over and play dead when Microsoft wanted them to help foot the bill on a failed xbox.

 

Here is finally a head to head comparasion of the Nvidia 6800Ultra Vs. a Ati Radeon 9800XT HERE!
Call me a Noob will ya ?
Enjoy
*props to HardOCP

 

Seems that on page 17, the Farcry 4x FSAA screenshot doesn't have any anti-aliasing

Wrong screenshot, or did you forget to turn it on Dave?

 

LoL. Now if he only compared the cards using the same resolution and AA/AF enabled.

 

I noticed that too. At the time I thought perhaps some of the diagonal edges matched the degree of rotation on their grid, negating most of the AA.

 

Good to see someone else noticed this. Seems like an awfully big coincidence...

Do you suppose that Nvidia could have included a no-release-date-discussion clause in the NDA? This does seem silly, yet can anyone recall a major product launch where the intial previews didn't at least *mention* some vague timeframe for product availability?

 

On the SS:SE page the graph with AF+AA turned on shows the 5900U much closer to the 6800U than the 5950U ? :?

Is that graph correct ? The table beneath it seems to tell a different story...

Keep up the good work, you ppl @ B3D!

 

I think you probably have your timing a bit confused. It's probably because of the fact that M$ dumped them and went to ATi, and because of the fact that ATi has, unlike nVidia, been shipping DX9-supporting hardware since August '02, that nVidia has finally in '04 brought "a solid chip" to market--well, at least to the preview stage it is currently in...

You have to remember that at the time M$ was laying the groundwork for xBox2 planning, the only thing nVidia was doing was making a lot of noise about how "the future of 3d" didn't include DX9 hardware support, resigning from FutureMark, and all the rest of it. Not to mention that nV30 was a complete bust, etc. If you want to understand the decision a bit better, you have to go back to the time period and revisit all of the rhetoric nVidia was spewing in public at the time. In context, the decision makes perfect sense, not to mention the fact that the nV40U would be totally unsuitable for the xBox2 (the retail versions of the nV40U card costing as much as 2 xBoxes, not to mention the power and cooling requirements, etc.)

Also, ATi demonstrated over and over again in the last 18 months that it could not only design a solid chip, but that it was also able to bring it to market in quantities sufficient to satsify demand, whereas nVidia cancelled nV30, had serious yield issues with nV35 for most of last year, and was only able to produce limited yields for nV38, despite fabbing 38 at IBM. nV40 is far more ambitious than nV38, and while I applaud nVidia for being able to deliver concept prototypes to reviewers, producing this 222M-transistor beast with acceptable yields is a task of an entirely different magnitude. On the whole I have no difficulty whatever in understanding M$'s decision.

 

Maybe we can keep this thread on topic, about Dave's review and the NV40.

 

Actually, the NV40U doesn't seem to be that much bandwidth limited at all in real games (with 4xAA/16xAF at least), which IMHO is surprising.
http://www.3dcenter.de/artikel/nv40_benchmarks/index11.php

 

I don't have a problem with that. However, I felt it was somewhat on topic since Nvidia appears to have released a solid chip. Which raised the question and others. However, I will ask them elsewhere.

My apologies.

 

Oh my, is this the NV40 that has me agreeing with DC wholeheartedly? It must be magic!

What's next, frogs raining from the sky? Cats and dogs living together?

...

A card that resembles (crudely described) nearly 2x R3xx + VS 3.0 (excellently implemented) + PS 3.0 (even with the caveats indicated)...I don't think anyone has a leg to stand on calling the NV40 a bad design. I don't think anyone has a leg to stand on calling it less than damned good. Normalization with penalty removed, even if it is fp16 only (look at the results...does the compiler find it for mdolenc's per pixel lighting test?)? Co-issue that adds a split option beyond 3(and subset)+1, which completes the set of possibilities for execution doubling? A change in registry performance penalty characteristics that makes direct and emphatic sense for shader usage?
A good continuation in the trend of 5900 change versus 5800 and the 5900XT/SE's subsequent pricing, IMO, regardless of what anyone else achieves.

 

Atleast, we will have peace in our time. NV40 = World Peace.

(somehow, I doubt it (peace), not with respect to demalion personally, but I think there is a sh*tstorm on the horizon, I'm just not sure what it is yet, but I have suspicions)

 

Hexus.net, HotHardware.com, and 3dcenter.org all included a little overclocking info. Most achieved quite marginal results.

Well, you're only playing with half the deck at the moment. There's no reason to believe R420 can't beat these numbers.

 

What kind of a review is this? Why no DOOM3, HL2 and UE3 WIP performance stats? Aren't these the games that matter for the NV40?



Seriously though, making a solid decision about the NV40 would have to depend on the answer to "I wonder how it compares to the R420?". On its own, in Tim Sweeney's words :

 

Can't wait to get my GeForce 6800 Ultra. I've waited long enough, time to ditch the 5200 Ultra.. lol.

 

Revisit this at a later date.

 

Good review except I did catch a mistake.

I think you meant "principles" not "principals." Last I checked a "principal" was a person and a "principle" was an idea. Unless of coure you're saying they are using people from one of ATI's design teams.

Otherwise it's a very good review.