nVidia NV20 info...my gawd!! ;)

[Spence]

the nv20 will be good, but its not going to follow those specs too closely especially seeing how there faked

 

[obeseotron]

Holy f'n .... That's hella fast. On top of this chip, which appears to be several times faster than anything 3dfx or ati would be able to put out before well into 2001, nVidia is fast expanding into everything. Did you see the info on NV30? 15 Gigapixels? 600 Million polygons? So much for anything mildly resembling Moore's Law. 18-24 months before NV30 would be NV10, at 15 million polygons and .48 gigapixels. That's 40X and 30X NV10.


nVidia has set it's sights much higher than 3dfx or ATI. They want to be (or beat) Intel. They want to have a chip to do everything that you might wanna do in the future. The Audio/Router chip for the X-Box is just the beginning. If everything goes according to their plan, nVidia would have a graphics line as diverse as Intel or AMD's line of CPUs, complete with a $750 chip at the top. They would have a huge contract with MS and X-Box for tens of millions of chips. They would have their chips in set top and portable devices for all kinds of applications.

 

[ElFenix]

moore's law is transistor count for the cpu doubling ever 18 months. note this doesn't count cache, or we'd have blown by it long ago. nvidia is currently going faster than that, but they weren't the bleeding edge to begin with. if intel can't find it within their cost structure to make a 100 million transistor cpu, nvidia isn't going to either.

 

[Sephiroth_IX]

<< You underestimate the other graphics card makers out there >>



<< DX8 is almost finished >>

YOU KNOW SOMETHING ABOUT g800 YOU BASTARD! TELL ME TELL ME TELL ME TELL ME TELL ME! DAasfdKASFNASDLNKALKBNENALFKDsadslnkaslKTRsdnaDSF!

 

[DaveB3D]

Guys,

I almost hate to tell you this because you guys are so into it, but the trust is that this is nothing but PS. The chip is not done and 4800 mpps got pulled out of somebody's rectum. That is the truth guys. NV20 will be out within the first few months of the next year, as long as no problems come up (and everything is looking ok so far).

Just thought I'd let ya know.

 

[KingDeviant]

DaveB3d, don't worry about upsetting the masses here. What you have said will be conveniently filtered out and never actually processed by their brains. If I'm not mistaken, the NV20 chip hasn't even taped out yet, so don't get your hopes up for a christmas release.

 

[Bignate603]

stupid me, should have waited a freakin month to buy my vid card... grr... would have cost half what is now, could have grabbed the ultra

 

[zippy]

<< YOU KNOW SOMETHING ABOUT g800 YOU BASTARD! TELL ME TELL ME TELL ME TELL ME TELL ME! DAasfdKASFNASDLNKALKBNENALFKDsadslnkaslKTRsdnaDSF! >>



Seph, I read ya loud and clear man...I was thinking the saaaaaaaame thing.

 

[NFS4]

I know as much as you do about the G800 and no more...I am certain about the DX8 issue though

 

[ViRGE]

Don't bother guys, it's a hoax, as the info came from our arch enemies the DPCs. They just want you to waste CPU cyles reading it, then arguing.

 

[Viper Frag]

Nvidia has laid the smack down. I think.

 

[Soccerman]

ahh... I couldn't resist at all..

first of all these ARE rumors.. second of all, the NV20 has Tile rendering which gives their PR department some room to stretch their legs.

wanna know why?

well remember the Kyro? it's wimpy in terms of overall power, but it competes VERY well for the price (I think it's one of THE best Price performance chips out there), becuase it renders with much more finesse then other cards (until the G800, possibly Rampage, and NV20) available for the computer.

now, multiply the available power by about 4X, give it about 3X the bandwidth, and you're beginning to see what the NV20 will be like, in terms of overall real power.

this is how nVidia's PR department gets these 4800Mpixel specs:
Kyro assumed an overdraw of about 4 (4 polygons not seen, but normally drawn on normal video cards). nVidia appears to be doing this also.

so the fillrate required to do that at the SAME speed with something like a GTS, would be about 4800MPIXELS, assuming no MEMORY bottleneck.

so, 4x overdraw. 4800Mpixels effective fillrate. that turns out as 1200Mpixels fillrate. pretty good, I assumed they would clock the NV20 a bit higher then the GTS (probably have a smaller die process).

if I'm correct, with it having 4 pipes, that gives the card a 300mhz clock, with RAM probably just being as fast as possible (RAM bottlenecks ARE lower, becuase when it cuts out polygons, it also cuts out the amount of textures quite a bit).

all of this this makes sense, becuase the &quot;delay&quot; allows nVidia time to get it running on a lower micron process.

that leaves the question of texture units per pipe (I'm guesing 2 to 3, with 4 being a remote possibility). it DOES support DX8 features as well..

can someone like jpprod tell me how much of a hit shading etc, of DX8 gives? I'd like to know, becuase I didn't know that trilinear filtering took up a texture unit..

other then that, I doubt that they included EMBM, becuase they have had plenty of time to do that in the past (so nuts to them!).

finally, I'd like to say this.. GO MATROX! and after that.. GO ATI! release a faster mhz version of the RADEON! and finally, GO 3dfx! get that damned Rampage out there! we need all the competition we can get. a perfect example of good competition is the Voodoo 3 vs the TNT2.. that was soooo sweet, becuase it MASSIVELY reduced the prices in NO TIME!

 

[Sohcan]

ROFL ViRGE

 

[NFS4]

Nice post Soccerman

 

[Viper Frag]

I knew it was too good to be true.

 

[Soccerman]

no, it IS true, the effective speed IS 4800Mpixels, when compared to a normal accelerator, when you are running a scene with an average of 4X overdraw.

you possibly will see 1600X1200X32 @ 100FPS, just look at the &quot;NV16&quot; scores, it gets pretty good FPS at that res (I'll find Anandtech's bench in a minute)..

 

[Soccerman]

here we go... Anand's 640X480 to 1600X1200 benches of the GTS U

analysing this, it appears that the NV16 (what I like to call the GTS Ultra) is still quite a bit bandwidth limited. take a look at this to bench compare performance with the overclocked NV16 to the normal NV16.

let me go into more detail about the Tile based rendering..

the increase in performance is twofold. first, your rasterizer (the pixel pipes) don't have to go through nearly as many cycles to complete the same scene (they don't have to map a texture to surfaces that aren't seen), which is where that insane Fillrate # comes from.

the second part of the performance increase is a result of the first, becuase you don't have to render surfaces that you cannot see, you don't have to use as many textures (unless they are just repeat textures, used a fair amount on the walls in Quake 3 for example), which means, it doesn't have to load nearly as many textures from the RAM.

this can reduce the Bandwidth requirement by a good 1 gig/second or more (I'm estimating, I really don't have benches), depending on the amount of repeat textures, and the amount of overdraw in that scene.

now, looking at the speed of those benches that the NV16 ran, overclocking the core from 250MHz to 285mhz and the RAM from 230mhz to 250mhz yielded a fair improvement in performance, mostly due to I expect, the RAM overclocking.

I'm beginning to think that nVidia will have to impliment something else really new in the next gen processor (NV30) in order to improve things much, becuase Bandwidth is a major problem now.

HOWEVER by that time, the T&amp;L unit will be a much larger factor, which may have a large leap in performance there (hey we BARELY have any monitors that actually RUN 1600X1200, or higher).

I'm beginning to think that after the NV25, we will see a race for much more revolutionary technology. Voxel acceleration, insane polygon counts for T&amp;L, and more little features that can be applied to textures (little tricks like what comes in DX8, things like EMBM, stuff like that).

we will see a plateau in the fillrate race for a while I think is what I'm trying to say. 300mhz DDR SDRAM will be available to video card manufacturers next year, and after that it will be QDR. that will be pretty good, but even WHEN those come out, the fillrate users will easily be able to fill that kind of pipe.

 

[zippy]

These are most likely theoretical numbers that they hope to achieve using tile based rendering once they can get the .15um process down...

However, if nVidia is anticipating these numbers, you can probably imagine the numbers on the Rampage, G800, Radeon2 or whatever, Trident 6MB AGP...scratch that last one, sorry.

BTW, how did DPC get out from under NDA to release these numbers? I'm sure they'd have been taken down and sued already if they were under NDA...

things to ponder...........

 

[Sephiroth_IX]

<< Guys,

I almost hate to tell you this because you guys are so into it, but the trust is that this is nothing but PS. The chip is not done and 4800 mpps got pulled out of somebody's rectum. That is the truth guys. NV20 will be out within the first few months of the next year, as long as no problems come up (and everything is looking ok so far).

Just thought I'd let ya know. >>

Thanks Dave, anything on G800 you plan to feed us hungry fish?

 

[Sephiroth_IX]

Wow, havent had a double post in a WHILE!

 

[RagingGuardian]

Those Dutch bastards must be laughing their heads off at you junkies. I'd believe the NV20 specs when I see em and Anand confirms that their for real.

BlackHawk2 made a good point about companies and their theoretical polygon pushing power. When Nintendo said the GameCube will push only 6-12 milion polys in real world enviroments the Sony fans had a good laugh because Sony said the PS2 can push 75 milion polys. Well the PS2 fanboys should hang out at a site like Anand more often to get some technical overview on their wonder machine. Imagine Ridge Racer 5 only pushes about 4 mill. polys. Where's the other 71 mill?

As for the X-box and the NV25. The video chipset will no doubt be very powerful but by the time it arrives we'll own 2GHz processors and faster video cards.

 

[zippy]

Maybe this has something to do with XBox?

I agree with dave at beyond3d though...and he probably knows a hell of a lot more than I do about this.

 

[DaveB3D]

Soccerman,

NV20 is not a tile based architecture. NVIDIA is not using a tiler and they have no plans on using one any time soon. Rather it is a traditional architecture with occlusion detection. It has been said that in a case of 10x overdraw you'll save like 4x of it, or basically 40%. This is somewhat optimistic though.

&quot;if I'm correct, with it having 4 pipes, that gives the card a 300mhz clock, with RAM probably just being as fast as possible (RAM bottlenecks ARE lower, becuase when it cuts out polygons, it also cuts out the amount of textures quite a bit). &quot;


As I said, 4800 MPPS got pulled out of somebody's rectum. That is not the fill-rate. Even if it were, fill-rate numbers will mean nothing. A large factor is going to be pipeline effiency with DX8.

&quot;all of this this makes sense, becuase the &quot;delay&quot; allows nVidia time to get it running on a lower micron process&quot;


Perhaps, but there is still a definate possiblity that NV20 will be .18. I haven't confirmed yet, but last I knew it was a possibility.


----------------------


G800. I don't know a damn thing. Matrox has been very tight lipped. Well so has NVIDIA in truth, but not as bad as Matrox.

 

[DaveB3D]

Soccerman,

NV20 is not a tile based architecture. NVIDIA is not using a tiler and they have no plans on using one any time soon. Rather it is a traditional architecture with occlusion detection. It has been said that in a case of 10x overdraw you'll save like 4x of it, or basically 40%. This is somewhat optimistic though.

&quot;if I'm correct, with it having 4 pipes, that gives the card a 300mhz clock, with RAM probably just being as fast as possible (RAM bottlenecks ARE lower, becuase when it cuts out polygons, it also cuts out the amount of textures quite a bit). &quot;


As I said, 4800 MPPS got pulled out of somebody's rectum. That is not the fill-rate. Even if it were, fill-rate numbers will mean nothing. A large factor is going to be pipeline effiency with DX8.

&quot;all of this this makes sense, becuase the &quot;delay&quot; allows nVidia time to get it running on a lower micron process&quot;


Perhaps, but there is still a definate possiblity that NV20 will be .18. I haven't confirmed yet, but last I knew it was a possibility.


----------------------


G800. I don't know a damn thing. Matrox has been very tight lipped. Well so has NVIDIA in truth, but not as bad as Matrox.

 

[ElFenix]

me making predictions:
next year everyone will have:
huge real fillrate numbers
some form of hsr/tiling
thusly even larger effective fillrate numbers (though this gets smaller cuz the engines are getting better at less overdraw)
accumulation buffer type fsaa (it really is better)
t&amp;l(&amp;c) units
64mb ddr
really high end consumer vid boards are still going to be $450+ (128 meg ddr, 600mhz!)
the bang for buck cards will follow gfmx example, good price but previous cycle speed, sorry, no v3/tnt2 shootout at <$200.


heheh dave dped... heheh. doh!