NV40 Coming @ Comdex Fall / nVIDIA Software Strategies

[David G.]

The fact that NV40 is coming is nothing but good news for us because I really hope it will be a succesfull architecture as I don't want people to pay money for driver cheats derived performance while having monstruos hot , huge cards .

Also , the new card will drive the prices of the 256 Mb cards down and also the rest of the prices which is nothing but good news .

One serious implication of this launch is ATi's net product , the R360 . I really don't think it will be enough to top the NV40 no matter what frequencies are used . Why is this so important ? Because JC said the DOOM III WILL be lanched in 2003 soooo ... if it's not gonna happen soon it's gonna happend later but still in 2003 . nVIDIA already had the NV35 faster than ATi's solution no matter the cheats ... That's a PR loss for ATi .

I think , if no other major improvements are prepared for the R350 , a switch to the 0.13 micron for the whole line of cards and launching a 9900 Radeon with 500/900 Mhz frequencies would be a solution . Sure a 420/800 card would be enough for the NV35 but not for the NV40.

Do you think ATi will launch another answear for the NV40 or will the ensure they'll have a response from now ?

Besides, the official claimed NVIDIA “The Way It Meant To Be Playedâ€￾ strategy to have two ramifications. Firstly, all participating game developers should use only NVIDIA Quadro professional graphics cards when designing games; secondly, many games would have certain features that only NVIDIA GPUs would be able to take advantage of.

Simple quiestion : Don't you think this is a clear anti-trus case ?

Source

 

[3dcgi]

Simple quiestion : Don't you think this is a clear anti-trus case ?

I'm no lawyer, but I don't think a business arrangement like this would bring about monopoly charges. It sucks for non-Nvidia owners, but isn't illegal.

 

[John Reynolds]

The agreement between EA and NV is simply one of EA 'suggesting/strongly urging' to their various developers to use NV boards as their primary development platform. EA is in the business of making money so it's not like we're talking about major compatibility issues for those running non-NV boards. Makes me wonder how many developers at the cutting edge who received 9700 boards last fall are going to heed such suggestions?

Moreover, this comment "secondly, many games would have certain features that only NVIDIA GPUs would be able to take advantage of" is simply the writer's opinion and is, IMO, flat-out wrong.

 

[Arun]

A few things:

1. Loci is also set to release at Comdex 2003, just as the NV40 ( which XBit is not sure of the Comdex launch date, I can personally guarantee it's nVidia's target )
Loci's on 0.13u and its performance target is "2.5x R300". Nice, eh?

2. The NV40 is going to be so darn impressive technologically that Intel, AMD and ATI patents put together couldn't even do something half as impressive.

Okay, so the second point was invented - and it's obviously exagerated. But why am I insisting so much on this point?
Because, based on old CMKRNL posts and new evidence based on the NV3x architecture, it is fairly obvious the NV40 is capable of dynamic allocation between Pixel Shader and Vertex Shader.

The R400 was also slated to have this AFAIK, but since it's canned, it'll have to wait for the R500...


Uttar

 

[Colourless]

I'll believe hype about the NV40 only after it's arrived and it's proven to actually be 'good'.

Wasn't the NV30 supposed to be the best thing ever, that ATI was supposed to have no chance of defeating? In reality, things turned out quite differently. As such, I have no faith at all in what people say about NV40.

Sharing functional units between VS and PS means little other than high triangle counts will slow down your pixel shading abilities. Yes, that is a negetive view of things, but what advantage does sharing really have?

 

[Josiah]

what advantage does sharing really have?

higher performance, higher framerates, maximum usage of hardware resources.

 

[whql]

2. The NV40 is going to be so darn impressive technologically that Intel, AMD and ATI patents put together couldn't even do something half as impressive.

You'd think people would have learnt by now.

 

[nAo]

Sharing functional units between VS and PS means little other than high triangle counts will slow down your pixel shading abilities. Yes, that is a negetive view of things, but what advantage does sharing really have?

IF the architecture is designed in a way that can automatically self-balance, removing units from a pool (say the fragments shading pool) and to devote them to another pool...well I can see a lot of space for performance improvements. Obviously such architecture should be able to quickly react to sudden changes in the data stream.
Efficiency would be sacrificied, but maybe with a very high clock speed for the sea of calculation units this approach could be a win..who knows? :?

ciao,
Marco

 

[Arun]

Sharing functional units between VS and PS means little other than high triangle counts will slow down your pixel shading abilities. Yes, that is a negetive view of things, but what advantage does sharing really have?

A few, random examples:

1. Faster geometry-only passes, such as Z Passes, because everything is allocated to the VS ( note however that you'd need to make sure outputting isn't the bottleneck if you want the benefit the most from this, so 16 Z output units would come in handy here )

2. Less idling waiting for other type of units to have finished. This would be very apparent IF there were, say, FX12 and FP32 units in the NV40, but I'd guess it's all FP32 ( with maybe still the FP16 register thing ) , so there aren't... Yes, yes, I know, bad example

3. Generally nice performance improvements, because one of the two VS/PS is ALWAYS the bottleneck.

4. Signifiant performance improvements for things like models which were badly LODed and thus got a lot of vertices but few pixels, or for things like the sky using few polygons ( although they often still use a nice portion of polygons, but there's obviously wasted VS power in complex ones such as 3DMark 2003 Game 4's sky )

Just some examples, I'm sure there could be more of them.
And for programmers, this is a lot nicer, because they don't have to think about balancing VS/PS functionality anymore. Not like it's *that* balanced right now often, but eh...

Of course, I realize the NV40 might not live up to this. So I won't insist on how good the NV40 might be. But should I get any info, well, then I'll also keep it for myself j/k, you know I don't do that


Uttar

 

[WaltC]

The agreement between EA and NV is simply one of EA 'suggesting/strongly urging' to their various developers to use NV boards as their primary development platform. EA is in the business of making money so it's not like we're talking about major compatibility issues for those running non-NV boards. Makes me wonder how many developers at the cutting edge who received 9700 boards last fall are going to heed such suggestions?

Moreover, this comment "secondly, many games would have certain features that only NVIDIA GPUs would be able to take advantage of" is simply the writer's opinion and is, IMO, flat-out wrong.

Yes, and I would add that folks need to realize that although nv40 might well be "announced" at Comdex, it won't be "coming" for another few *months* after it is announced. If ATi does another high-end product this year based on R350, it will probably "come" about the time nv40 is "announced," I would think. I would also think that after last year's nv30 announcement at Comdex people would easily understand the timing differences between announcing a product and shipping it in volume.

Also, PR campaigns of this type typically exaggerate, fabricate, and mischaracterize such agreements--I've yet to see any PR on this from EA--I've only seen nVidia's press releases concerning it. It's nVidia claiming unidentified special feature support rather than EA. Check out EA's press releases going back to July 2001 here:

http://investor.ea.com/ireye/ir_site.zhtml?ticker=ERTS&script=400

It was interesting to note that although there are many press releases concerning EA's partnership deals with various companies, and several mentioning development work for GameCube, xBox, Nintendo and Play Station, I didn't see a single one that even mentions nVidia by name.

Obviously, EA does not see whatever agreement it has with nVidia as important enough to the way it does business to merit a press release on the subject. IMO, the deal it has with nVidia is some type of routine software bundling deal that nVidia is setting up for its OEMs, and thus doesn't merit any special mention from EA. It would also seem as though the part of the nVidia PR mentioning the unidentified special feature support in unidentified EA software titles is purely fabrication. nVidia PR is spinning up and fabricating what is probably something very minor in the scheme of things hoping, once again, to create illusions in the minds of its prospective customers. IMO, when I hear EA developers talking about their "special nVidia feature support" in concrete terms then I'll look at this issue a bit differently. Generally it's common practice to release quotes from various software developers to back up comments like these in a press release, and in the nVidia press release I read on the subject there were no such quotes included.

I recall in the final months of 3dfx seeing a raft of similar hyperbolic press releases coming out of the company, and detailing "special" agreements between 3dfx with companies like Epic and talking about "exclusive" product distribution deals with companies in China. IIRC, Epic later made a statement stating that it was working hard on its D3d software engine and that people with 3D cards which didn't support GLIDE would have no problem getting the full benefit of its software. The deal 3dfx announced for China distribution never materialized as far as I know.

The bottom line is to always take press releases of an uncorroborated, sensational type with a big grain of salt--and always consider the source.

 

[LeStoffer]

Of course, I realize the NV40 might not live up to this.

Methinks: Too much DX10 too soon.

 

[KimB]

Yes, and I would add that folks need to realize that although nv40 might well be "announced" at Comdex, it won't be "coming" for another few *months* after it is announced. If ATi does another high-end product this year based on R350, it will probably "come" about the time nv40 is "announced," I would think.

From what I've heard, there's a second iteration of the NV35 due out this fall, followed by the NV40 around Christmas. nVidia shouldn't need the NV40 to compete with an "R360."

 

[John Reynolds]

From what I've heard, there's a second iteration of the NV35 due out this fall, followed by the NV40 around Christmas. nVidia shouldn't need the NV40 to compete with an "R360."

Depends on whether or not they want to just compete or to return to their halcyon years of 2000-1H2002 where they absolutely dominated the market. NV management's pride has got to be battered and bruised from these past 8-9 months, and the recent negative 3DMark furor can't be much of a salve.

And new architectures can be quirky beasts for driver teams to get ahold of. The first GF was initially slower in some tests than NV's own TnT2 Ultras. But we'll see. I'd personally prefer to see NV400 ASAP rather than a NV35 refresh this fall, but the two aren't necessarily intertwined.

Edit: Changed NV30 to NV40. #s make me head hurt.

 

[Kristof]

3. Generally nice performance improvements, because one of the two VS/PS is ALWAYS the bottleneck.

More likely to be stalls due to all kinds of interactions relating to memory (off or on chip) and bandwidth issues...

Is it just me or are people getting their hopes up with completely unrealistic specs that are probabaly going to come at some point in the future but not necessarily NV40... I mean lots of people got their hopes up for a primitive processor in NV30, and look what happened to that. Maybe time for a reality check on all the things that are thrown around ? Afterall non of this has been confirmed in any way, not even leaked AFAIK.

 

[WaltC]

From what I've heard, there's a second iteration of the NV35 due out this fall, followed by the NV40 around Christmas. nVidia shouldn't need the NV40 to compete with an "R360."

Hmmm....here it is June and the first iteration of nv35 is not shipping. From what I hear the 5900U sent out around the review circuit won't be shipping in any kind of volume until August at the earliest, with the lower-priced nv35 cards only beginning to ship into the maket in late June or early July. That's the first iteration of nv35.

So how is it you figure nVidia's going to be into the first iteration of nv35 around August, but will be shipping the second iteration of nv35 AND nv40 prior to Christmas 03? Actually, I wouldn't doubt the revisions of the chips may change for nv35 later this year--but I'd expect improvements to be minor as opposed to fundamental. I heard all of last year that nv30 would "definitely" ship by Christmas '02. My hope is that maybe this year we will have a much more sober picture.

nVidia's got to do something to control its runaway marketing. IMO, it's the singlemost reason nVidia is constantly chasing expectations it can never meet. I hope they'll realize it's hurting them more than it's hurting their competition at some point. They knew full well what a dog nv30 was when they launched their "Dawn of cinematic computing" PR blitz at the time of the product announcement--the thing of it is they went ahead and did their PR show anyway. All that dawned was the "failure" of nv30, as the nVidia CEO put it. I wonder what it's going to take before they'll quit chasing their tail in this manner...

 

[nAo]

I wonder what it's going to take before they'll quit chasing their tail in this manner...

Ohh boy, that's easy! Just wait till they'll have a new great product

 

[WaltC]

I wonder what it's going to take before they'll quit chasing their tail in this manner...

Ohh boy, that's easy! Just wait till they'll have a new great product

Heh-Heh...Agreed!...

 

[Doomtrooper]

Is it just me or are people getting their hopes up with completely unrealistic specs that are probabaly going to come at some point in the future but not necessarily NV40... I mean lots of people got their hopes up for a primitive processor in NV30, and look what happened to that. Maybe time for a reality check on all the things that are thrown around ? Afterall non of this has been confirmed in any way, not even leaked AFAIK.

No it is not just you

 

[David G.]

simply the writer's opinion and is, IMO, flat-out wrong.

It better be .

The only time ATi was in a similar exclusive situation was when all the Radeon 2 cards were DX8.1 capable and also were the only ones . But the situation was not that bad for nVIDIA as PS 1.4 was just faster , more performant but not incompatible .

 

[David G.]

A few things:

1. Loci is also set to release at Comdex 2003, just as the NV40 ( which XBit is not sure of the Comdex launch date, I can personally guarantee it's nVidia's target )
Loci's on 0.13u and its performance target is "2.5x R300". Nice, eh?

How do you know that ? 150% faster is unimaginable for my narow mind !

How ... ?! What ...? What would that show ? over 100 FPS in DOOM III w/4xAA and 8xAniso ?

2. The NV40 is going to be so darn impressive technologically that Intel, AMD and ATI patents put together couldn't even do something half as impressive.

Hehe ,

I'm not saying that it's going to be the promised land but if he curent NV35 is as fast as R350 ... and even faster ... and even much faster with cheats . The NV40 will bring at least a 50% performance delta which can't be produced by an 420/800 R350 .