NVIDIA GeForce 4XX Series Discussion

[Binge]

How would the IQ be better?

There are some close friends who have been to the CES and it's becoming very apparent that the geometry processing of the card has no equal even in the Hemlock dual gpu solution. If they've done a better job with AF, and other methods of post processing then there's going to be more than just ease of geometry rendering with which to be impressed.

 

[mastrdrver]

NDA falls on sunday apparently.

People keep trying to figure out if the videos are of the 448 or 512 variety . I've seen two different people, one on XS and one on B3D, state that they can't demo a 512 core version cause they don't have any.

On XS

huh use your brain guys who is gonna make presentation with there second fastest card ?

and how long are you going to wait for that fastest card since they don't even have one example to test on ? god only knows what will hapen by the time when it come to market

There was a rumor not too long ago that said 448 core version would be the "GTX 380".

 

[TheMailMan78]

There are some close friends who have been to the CES and it's becoming very apparent that the geometry processing of the card has no equal even in the Hemlock dual gpu solution. If they've done a better job with AF, and other methods of post processing then there's going to be more than just ease of geometry rendering with which to be impressed.

Ill take your word for it Binge but I still wanna see it. Ive never notice a damn bit of difference in IQ between the two manufactures.

 

[bobzilla2009]

Indeed, also 99% of all comparisons that lean one way or the other are things where you need a picture testing AF side by side. I've never seen any difference between in game footage tbh, maybe i'm just less obsessed with whether a texture looks ok at 100m and more with just having fun in game ^^

 

[phanbuey]

yeah the marketing director at CES said something about driver level post processing effects to make things look more realistic when he was running the heaven benchie.

danger den needs to come out with a waterblock for this card...

 

[bobzilla2009]

Indeed, but there's no guarantee that it won't cripple performance for most games, after all heaven will have been targeted specifically with the CES drivers. Still, if they can achieve far superior IQ with an acceptable loss on frames, it'll be amazing.

Anyway, the best post processing that can be applied to a game to make it look realistic is the 'pretend to video it, and replay it back', since the youtube videos of heaven nearly always look better than the benchmark itself ^^

 

[kid41212003]

Check this out:
http://forums.techpowerup.com/showthread.php?t=112980
Case certified for Fermi cards, and from the look of it, Fermi won't be longer than GTX280.

 

[bobzilla2009]

It's certified to cool the dx11 nvidia gpu, but uses extra gpu cooling to achieve that? Doesn't sound promising, my hd5870 kept cool in a tiny m-atx case with one cooling fan (before i changed case). Sounds like heat may indeed be a slight problem

 

[kid41212003]

It's a vent.

 

[wahdangun]

yep it's mean you must have an air duct to have certificate from nvdia, so rumor about fermi temperature was right after all,


fermi i's going to be really "hot" card

 

[kid41212003]

The card is freaking hot! I want one NOW!

 

[bobzilla2009]

If the card is so hot it requires a special vent to gain certification from vendors, it can bugger off, and i'll get a second hd5870 unless it beats out 2 hd5870's of course.

 

[HalfAHertz]

The card is freaking hot! I want one NOW!

me 2. I'm not gonna use it tho. I'll just put it up on the wall and stare at it for hours on end, after I finish praying to it that is...

 

[mastrdrver]

G100 loving action!

Bjorn3D Forums

Design Article releases tomorrow 7PM CST with complete Whitepaper info.

New Features, new cache, new Memory setup, and yes it's about 100% performance increase over GTX-2xx so figure single GTX-285 vs 5870 then double the GTX-285 performance.

Then it handles triangles different, triangles on any given frame can number in the hundreds of thousands so that's very important.

It will fold a lot better.

Increased efficiency in several areas.

It's a revolutionary new design oriented toward tessellation (those pesky triangles) and geometric programming. Problem being every wire frame is made up of triangles, tessellation takes those triangle and breaks them down into many smaller triangles. This core is uniquely designed to handle that so geometric and shader heavy games you will see more than the 100% raw power increase.

520USD might handle it. At 2x GTX-285 performance that puts it above GTX-295 performance and it's DX11 ready and designed for that specifically. Current ATI offerings are really good but basically a double the hardware on the same core design to provide more raw power. GF100 is a core design to take advantage of what the industry needs today and for some time in the future.

Read the article tomorrow cause that's about all I can say tonight.

 

[Easo]

Sounds too good to be believable... No, realy!

 

[Steevo]

Bad link. No card. Whitepaper only?


I have a whitepaper too, it is in my printer, unprinted on. It could be awsomesauce.


Triangle handled differently? no WAI! Almost like tesselation but not quite like the tesselation that ATI has had on the previous three generations of cards? You know, the ones that Nvidia took a stand against and told us all that it was worthless. Like that? But better since it is green vapor?

 

[Benetanegia]

Bad link. No card. Whitepaper only?


I have a whitepaper too, it is in my printer, unprinted on. It could be awsomesauce.


Triangle handled differently? no WAI! Almost like tesselation but not quite like the tesselation that ATI has had on the previous three generations of cards? You know, the ones that Nvidia took a stand against and told us all that it was worthless. Like that? But better since it is green vapor?

Well, accept it or not, it was pointless. The HD5750 can't keep up with it, the HD5770 can't keep up with it, the HD5850 can barely keep up with*. Hence HD2xxx/3xxx couldn't keep at all, and neither the HD4xxx. For me that means it was absolutely pointless until now and with reservations: we are talking about Dirt2 here, a game that barely uses tesselation or any other DX11 feature... Not that those cards manage Unigine benchmark better.

http://www.pcgameshardware.com/aid,...chmarks-Top-article-of-October-2009/Practice/
http://www.hardocp.com/article/2009/11/06/unigine_heaven_benchmark_dx11_tessellation/

Tesselation has always been a fantastic feature that couldn't be implemented yet in any advantageous way. It's useful to expand the level of detail beyond what he had, with a relatively low performance hit (about 40%); but it has never been useful to create the same detail at improved performance or a small improvement at no cost, which was what low performance (by todays standards) cards would need.

* Not even the HD5870 qualifies for some people. I'm talking about the ones that "need" at least 60 fps

 

[a_ump]

never thought of it that way. so very true. all this hype, i keep in mind the HD 2900XT's hype and following flop thou just incase fermi does the same

 

[mastrdrver]

Bad link. No card. Whitepaper only?

Looks like someone said too much too early. The link was the quote.

 

[HalfAHertz]

@Benetanegia

I have to dissagree. You are only half right. The best thing about tesselation, and the one you forgot to mention is that it's scalable. I woudn't be surprised if future game titles have an option to adjust the level of tesselation. That would bring the best of both worlds - improved graphics and acceptable performance. Correct?

I wager we should give game devs some time to familiarize themselves with the new features, before that hapens tho.

 

[Benetanegia]

@Benetanegia

I have to dissagree. You are only half right. The best thing about tesselation, and the one you forgot to mention is that it's scalable. I woudn't be surprised if future game titles have an option to adjust the level of tesselation. That would bring the best of both worlds - improved graphics and acceptable performance. Correct?

I wager we should give game devs some time to familiarize themselves with the new features, before that hapens tho.

I wouldn't say the tesselation present in Dirt2 is excessive at all. On the contrary is quite simple and only used in few areas. Neither is excessive at all in Heaven demo. It's just the bare minimum if you want to use it for anything more than smoothing things out just a bit (in fact I find it a little bit lacking in order to use it as a displacement mesh, I always took the Heaven benchmark as a first step towards what can be done). And IMHO unless you are using it for all we see in th Heaven benchmark it's even more pointless, because you could achieve the same results with more geometry and better normal mapping (even nowadays 512x512 or 1024x1024 normal maps is usually what you get). For example, adding some 6 sided (12 poly) spikes to the dragon wouldn't increase geometry complexity too much at all and with the proper normal mapping you would achieve almost the same you do with tesselation. Despite that, the beauty of tesselation is that can be used everywhere without the need for the artist to think where to put more effort (and performance), but in order to achieve that you need some minimum level of tesselation. In all honesty, you can't use less than that. IMHO!

 

[HalfAHertz]

I agree, there are some alternatives for achieving the same effect, like cleverly placed bump mapping. But as you said it yourself, it requires more effort for an inferrior result.

All I'm saying is that we should give it moretime. It's still a "new" technology in a sence that game devs have just started using it, even tho Ati has been promoting it for years. I'm pretty sure there's much much more to follow.

Who knows, maybe after a year or so games may become so dependant on tesselation that not having capable hardware would lead to inferrior gameplay. That would surely force a spur in graphics sales...

 

[phanbuey]

never thought of it that way. so very true. all this hype, i keep in mind the HD 2900XT's hype and following flop thou just incase fermi does the same

yEP... Im hoping its good, but ive seen this cartoon before...

 

[Benetanegia]

I agree, there are some alternatives for achieving the same effect, like cleverly placed bump mapping. But as you said it yourself, it requires more effort for an inferrior result.

All I'm saying is that we should give it moretime. It's still a "new" technology in a sence that game devs have just started using it, even tho Ati has been promoting it for years. I'm pretty sure there's much much more to follow.

I'm not saying that the feature is pointless, or that it shouldn't be used and pushed forward. I know it very well and I use it almost everyday when I'm modeling (offline, in 3dmax), but it was not really applicable until now. Not in a way that was worth the effort of implementing all the underlaying system in the game engine, specially considering that you would have to create the normal version for those cards that just can't use it anyway (no support, low performance...). An environment that was supposed to use tesselation, and which relies on tesselation in order to create any detail, would just suck without it. Just like Heaven demo sucks with tesselation dissabled. So you have to add the details, want it or not, and tesselation would always have to go on top of that detail, making the scene around 3x more complex than without, hence producing a performance drop.

And about giving time... I'm not a believer of that anymore, we've used that "excuse" too many times with too many things in general (DX9 in te past, DX10...). They've had the time with Dirt2, it's been an AMD game since long ago and delayed the game even further to release it DX11. IMO there's very little more they can do performance wise, and again, IMO they are not using it too much either. The Radeons are supposed to be able to create 850 billion tesselated vertices per second. There's not that much on Dirt2, that I can assure. What most people apparently overlook is that although tesselation is a far faster way of creating geometry within the GPU, you still have to process, lit and texture that new geometry and it has an almost comparable performance drop to adding so much detail in the first place. You skip the geometry step but you can't skip the rest of the pipe...

 

[crazyeyesreaper]

games will have options to scale tessellation example AvP will allow you to set tessellation at off low medium and high so as to provide better scalability