NV30 information.

[Fuz]

What do you guys make of this? I am not sure if 3dchipset.com is a reliable source, but any way.

Taken from here http://www.3dchipset.com/news.shtml

I've received some information regarding the NV30, NV31 and some insight of the NV40 Nvidia is working on. First off though, the NV30 is AGP 8X and it's still a go with 900MHz DDR memory. Also what is now known is that the NV30 is multi chip capable. The number of chips is an odd number. That's all I can say without giving out the exact number. I'll say this though. It's under 8. With regards to FSAA performance the NV30 has a zero (0) performance cost using 4X FSAA. Specifications have been discussed and the NV30 is able to churn out 16 textures/pixel per pass. The NV30 also uses a 4-1 colour compression technique. Nvidia has explained to us/me that the performance we all will see with the NV30 is about double the performance of the Xbox/GF Ti4600. Another feature that is not known to the public is the use of IEE 1394 or known as Firewire will be on NV30 video cards. What this feature is for wasn't discussed thoroughly and all we know is that it has to do with the hardware mpeg2 acceleration built into the chip.

The NV31 has also been discussed with us/me. The NV31 comes out one month after the NV30 and will not suffer the fate of the MX by branding it GeForce 4. You remember that debate about the GF4 MX shouldn't of been called GeForce 4. Well Nvidia has listen and the NV31 line will feature everything the NV30 has but just be lower clocked cards.

Now on to the NV40. Very few details where presented to us on the future chip, but we know that Nvidia is shooting for 600 million polys per second with a 4 Gigapixel fillrate.

So let's recap all of this shall we. What is now known is this:

- Zero (0) performance cost using 4X FSAA
- 4-1 Colour Compression Technique
- 16 Textures/Pixel Per Pass
- 900MHz DDR is still a go
- Multi Chip capable
- Firewire Port

Fuz

 

[Typedef Enum]

Based on the information that has been leaked...the fact that we're talking about a fusion of nVidia/3dfx...and everything in between, the informaiton certainly sounds pretty sound.

As far as the source is concerned, I've found 3DChipset/Solomon to be a pretty good source for reliable information.

 

[DemoCoder]

Sounds "reasonable" unlike some of the other bogus specs floating around. (900Mhz DDR sounds fishy tho, so does multichip) But no other specs are mentioned: DX9? Programmability? (16 textures sounds like DX9 PS2.0) No mention of LMA3 or any super T&L capability. It seems these would be the obvious leaked features and their absense makes the rumor conspicious. NVidia likes to invent new trademark names for stuff and refer to their own improvements with code names. Why don't the rumors talk about SuperColorKompression(tm) or something?

 

[Typedef Enum]

From the sound of it, doesn't it almost imply that NV30 will be employing a 128-bit bus?

Take all the other GF3/4 tech...add another form of compression...tweak the existing GF3/4 tech. another notch, and add significantly improved texture processing...And, the potential capability to multichip...

Factoring in some of the interviews with D Kirk, and maybe...just maybe...they're going to avoid the 256-bit memory int. after all?

 

[DemoCoder]

Yeah, it doesn't sound "revolutionary enough" to justify all the comments coming out of Kirk and Carmack. Either these specs are wrong, incomplete, or we are in for a letdown. Where's the extended backbuffer precision and general purpose programmability that we have been hearing whispers about?

 

[Grall]

Typedef:
NV30 isn't supposed to be an extended/improved GF-like architecture, it's brand new according to Nvidia that is.

Demo:
I wouldn't worry so much, who says the alledged briefing that took place covered all aspects of the NV30? Quite possibly they held back on a number of choice tidbits of info. The information could still be legit, just incomplete.

*G*

 

[Evildeus]

Well, we do not have much information, no? So how can we imply the veracity of the "revolutionary" of NV30? It's more or less what we already know apart from the FSAA*4 no hit in perf.

 

[BoardBonobo]

The multichip solution ties in with something JC mentioned about multichip cards being released within a year (I'll try and find the link) also the free FSAA and other bits tie in with this informal interview which details some of the solutions that 3DFX were working on before they went pop.

It's never going to work on .15u though, it'll need lots of speed and masses of silicon. If they can't get .13u then will they have to release a crippled card?

 

[Prometheus]

Hello I'm new as you can see but I've been visiting this forum for a long time now.Is it me or do those nv30 specs look like rampage specs?
It's sad to see a 2 year old project passed as new.Don't get me wrong I'm neither nvidiot or fanatic.Both companys bring the same products.There hasn't been a new feature in the industry since 3dfx introduced fsaa
Imagine if ati cards had fsaa and nvidia didn't.
Which would you choose?My point is that you can afford to be a bit slower but if you are slower and have the same features then whats the point?

 

[LittlePenny]

interview which details some of the solutions that 3DFX were working on before they went pop.

Please let us never reference an article as fact when the person being interviewed is nothing more than a webmaster of a fansite. Unless the man/woman actually worked at 3dfx everything is just speculative. Plus, Barron already told us most of which was said is incorrect.

Prometheus, these specs are nothing concrete just rumors, please wait until actual products are released before reaching conclusions.

 

[Lessard]

Why does Nvidia absolutely need a 0.13um process for the NV30 if it's a multichip product ?
Doesn't makes sense to me ...
The multichip part is maybe only for the very high end market (workstation)

 

[Gollum]

Why does Nvidia absolutely need a 0.13um process for the NV30 if it's a multichip product ?
Doesn't makes sense to me ...
The multichip part is maybe only for the very high end market (workstation)

Highly likely, a one chip solution would already be fast but expensive, two chips *might* find a niche market for the hardcore crowd (like what the V5 6k was aimed at), anything beyond would certainly be much too expensive for the consumer market...

 

[Doomtrooper]

The multichip solution ties in with something JC mentioned about multichip cards being released within a year (I'll try and find the link) ....

http://slashdot.org/comments.pl?sid=34863&cid=3784210

 

[Edge]

It's 450 MHz DDR memory, cause the memory will not be running at a clockrate of 900 MHz!

 

[Grall]

Gollum:
Why would a multichip card be 'much' too expensive for the mass-market?

Voodoo Graphics/2 was multi-chip. Voodoo5 5500 was multichip, neither product was out of reach of consumers.

Sounds to me you're spouting...well, something.

 

[Pete]

This talk of multichip (Spectre), free FSAA (GigaPixel), and 4-1 compression (FXT) sounds familiar. I've got a good feeling about this. As much as we may lament 3dfx's passing, surely we can't deny a 3dfx-nV pairing will be even better?

 

[Dave Baumann]

I don't think the 4-1 compression would relate to FXT texture compresion (that was, best, 8-1 IIRC) - I would imagine it relates to framebuffer compression.

This has been discussed a few times before already and I think it would be kind of similar to the Z-Buffer compression already seen. It's also an easy explaination for essentially 'free' 4x FSAA. GF3/4 already has fillrate free FSAA becuase of its multiple Z checks per pipe, its main penalty comes in becuase of bandwidth - now, because many of the samples under multisampling use the same colour value (only one texture sample for all subsamples) it would probably be farily easy to gain a frame buffer compression ratio close to 4-1, in which case you've pretty close to just alleviated you bandwidth penalty for MSAA (which is nice!).

Of course, the main drawback to this solution is that to increase the number of samples available means more silicon for more Z check units (although solutions similar to those used for 4xS can achieve this).

 

[Brent]

There hasn't been a new feature in the industry since 3dfx introduced fsaa

hardware programable pixel and vertex shaders on the gpu, that's a pretty big 'feature' heh

granted they still don't have the programability that some developers want, but they are there

we are still at the infancy in 3d graphics and acceleration, exciting times ahead...

 

[Gollum]

Gollum:
Why would a multichip card be 'much' too expensive for the mass-market?

Voodoo Graphics/2 was multi-chip. Voodoo5 5500 was multichip, neither product was out of reach of consumers.

Sounds to me you're spouting...well, something.

You have to be realistic about market realities, the price of a product is not determined simply by the number of chips, but by overall complexity and component costs. I am going by what we think to know about this chip, and everything so far tells us its a hell of a complex monster (~120million transistors). This chip is not coming from the fab cheap, and neither is the rumored 450(900 effective) MHz DDR going to be cheap, so comparing it to a Voodoo5 5500 is really a bit shortsighted.

The VSA100 was designed to be a easily scalable chip, even without the delays a single chip would not have been able to compete. The Voodoo 4 (1 VSA100 chip) was a value card, the design only got competitive in the entusiast market through multichip solutions and intentionally so IMHO (1 chip = value, 2 chips = enthusiast, 4 chips = hardcore). NV30 however is obviously already complex yet powerfull as a single chip solution, which is also entirely in Nvidia's tradition, and if it turns out to be scalable its not gonna be the same stratergy as the VSA100 used (more like 1 chip = enthusisast, 2 chips = hardcore, anything more = high-end/workstations). VSA100 also had the advantage that it could use slower but cheaper SDRAM than a GF2, which depended on the significantly more expensive (back then anyway) DDR RAM for bandwidth.

Lets sum it up, the V5 was a affordable multichip solution because the VSA100 chips were relatively cheap to produce and RAM was both cheaper and more available than the cutting edge technology Nvidia depended upon. According to all the rumors though, NV30 is complex and probably expensive as a single chip already, plus it apparently still depends on rare top-speed DDR for bandwidth. Considering that, I don't see how e.g a 4 chip NV30 card could be affordable for consumers. Of course you could probably buy such a card, it just won't be marketed as a consumer product and be friggin' expensive...

 

[Joe DeFuria]

A bit OT...but I just love analogies...so....

we are still at the infancy in 3d graphics and acceleration, exciting times ahead...

I'd say we are more in the early adolescent phase of 3D graphics.

Infancy: Doesn't do much...just lies around, poops a lot, and requires much attention just to keep alive and functioning at a minimal level. Has little resemblance to its ultimate potential, but definitely a life form. I'd say this characterizes the nVidia NV-1, and early S3 and 3D Labs consumer (GiGi) parts as well.

Toddler: A bit more capable. Can actually get around and do stuff, you can sort-of communicate with it....sometime it underatands and follows your commands, sometimes it doesn't. Has clear and severe limitations, but you can actually start to see the potential of what the future will bring. Characterizes the Rendition Verite, 3dfx Voodoo Graphics, nVidia Riva 128, ATI Rage Pro type chips.

Kid: Delivers a level of functionality, usefulness, and reliability that is a clear step above toddlers. Can do many things that more "advanced" people can do, only slower -- using a simplistic and primitive way of thinking. Characterizes the Voodoo2-Voodoo5, TNT - GeForce2, ATI Rage 128 - Radeon type chips. (Through DX7 functionality).

Pre-Adolescent. Things are starting to show signs of change. The child starts to show more advanced ways of thinking and acting, but isn't quite capable. Often, when the child tries to be an "adult", it ends up generally falling well short of typical adult expectations. This characterizes DX8 hardware....GeForce3/4, ATI Radeon 8500.

Adolescent: Clearly, this batch of folks thinks that they know-it-all, and are as "mature" as fully-fledged adults. There is a mix of under and over achievers in this group. They claim they can do it all, with varioius degrees of success in different areas. They will listen and claim to understand toeverything you say, but not necessarily follow your instructions the way you expect or want. Sometimes they will exceed your expectations of their capability, and other times they will disappoint. This will characterize early DX9 hardware like NV30, P10, and R-300.

Early Adult With the adolscent "kinks" worked out, plus a dose of experience and humility, these people actually have a relatively good grasp of the "real" challenges in life, and will set a good course for tackling them. Some things are still awkward, but the basics are pretty much mastered, and a solid foundation for the future is laid. This should characterize DX10 or 11, and OpenGL 2.0+ hardware, along with the arrival of "acceptable" high level shader languages.

Adult: This is a very long period in which there are few "revolutionary" changes from this point forward. Most everything is built upon the evolution, extension and maturation of the early adult stage. Greater stability and efficiency as the person gets older. DX 11 or 12 and beyond. Sometime during late adulthood, a new infant will be born.....The adults pretty much ingore it.

Senior Citizen Now, these people have lived a long, healthy, and productive life, and are pretty much set in their ways. That other infant is starting to hit the pre-adolescent stage. Stubbornly, adults shrug him off as just another immature brat. The adolescent is making grandiose claims of a revolution and a "new age", to obsolete the Seniors. The seniors try and resist the revolution, but deep in their heart, they know that their days are numbered....they just won't admit it.

Death The cynical, stubborn and highly "resistant to change" graphics chips will put up a valiant fight, but in the end, death of the graphics sub-system as we know it today is inevitable. That adolecent kid has entered early adulthood. What is it? Probably some uniform programming subsystem where discreet CPUs / GPUs don't exist, but are a single, unified architecture....