NVIDIA's Next-Generation Architecture @ firingsquad (NV3x info)

[goulch]

I mean there needs to be a balance between huge textures + fx's, and triangle counts.

Just look at the latest hot game from blizzard: Warcraft 3. This is the only first PC game in my opinion that successfully try to balance the use of texture and poly counts. Now try to max out the video setting options. I bet even with a combination of P4 + geforce 3 has difficulties of maintaining a mere 60 fps on moderate to heavy count of critters in a given screen.

For those who are satisfied with the GF3+ triangle setup capacity ($$$ card already, + huge 128MB mem) , I bet that most will admit that even from farthest in-game camera position, a WC3 creature seems pointy and patchy. Note: high texture won't help much.

Have anybody seen PS2 game demo of Ratchet and Clank by Insomniac? I would rather go low resolution (ok, individual choice here) with adequate triangles for objects rather than explosive on resolution + huge texture but in a flat wall and ceiling environtmnts .

Again, it is about balance.

 

[Mr.IncrediblyBored]

Originally posted by: ElFenix

Originally posted by: kazeakumaPolygons are not the biggest focus anymore simply because current tech can pump out more than enough. The PS2 is/was a polygon beast but it doesn't make up for it's shortcomings. Most games are blurry as hell due to limited texture space and the PS2 is quite low res compared to PCs. I don't know why you think the PS2 can pump more polys than a pc (remember on a PC the Video card does all the rendering NOT the CPU so your comparison is invalid) because the Xbox which is based on the GF3 can push far more polys AND textures than the PS2 and PC graphics are already leagues ahead. The reason why some PC games lag behind is they are catering for a broad range of people. You can't shut out half your potential market by requiring them to have a $300 video card. That said there are still PC games which are far prettier than any PS2 game out there.

remember that the CPU figures out where the triangles should be then sends vertex data to the graphics card, so yes, the CPU comparison is valid.

A vertex shader can manipulate vertex data without any CPU intervention. Hell, the transform process has been controlled by the GPU since the original Geforce. Uploading vertex data over the agp bus is a relatively simple process. There is little to no calculation done by the CPU.

You can't compare vertex throughput between the EE and a CPU simply because the CPU doesn't play a significant role in the transform process. The comparison is invalid.

 

[Bullhonkie]

Originally posted by: Adul
i bet this will make john carmack happy

Here's a supposed Carmack quote verbatim:

Nvidia is the first of the consumer graphics companies to firmly understand what is going to be happening with the convergence of consumer realtime and professional offline rendering. The architectural decision in the NV30 to allow full floating point precision all the way to the framebuffer and texture fetch, instead of just in internal paths, is a good example of far sighted planning. It has been obvious to me for some time how things are going to come together, but Nvidia has made moves on both the technical and company strategic fronts that are going to accelerate my timetable over my original estimations

My current work on Doom is designed around what was possible on the original Geforce, and reaches an optimal impliementation on the NV30. My next generation of work is designed around what is made possible on the NV30


Posted here by an acquaintance of mine. Sorry if this is a repost.

 

[JackBurton]

Ah true, I had forgotten about that. Nonetheless, PCs are able to push more than enough polys. Who wants to play a game with 100million polys but only 16mb of textures and no fx? Boring as hell if you ask me.

Dude, I don't know what you are talking about, but PC games need A LOT higher polygon count. I don't want a freakin' cube with 8 billion colors on it. Most PC games fake their way through and use techniques to try and mask the obviously low polygon count. Take for instance Max Payne. His head is a freakin' block for God's sake! They just wrapped a colorful wallpaper around it with all the facial features. I don't care if you play the game at 1600x1200, his head still looks like a block...maybe a prettier block at 1600x1200, but still a block. IMO, no game (PC or console) has come close to DOA3 on the XBox (looks wise). The game is simply beautiful!

 

[goulch]

You can't compare vertex throughput between the EE and a CPU simply because the CPU doesn't play a significant role in the transform process. The comparison is invalid.

Ok Ok,

Let us spare Intel's superior P4 for a while and give PC a little help here.

In order to topple the FPU capability of an EE, PC with P4 2.5+ GHz ($$$) need another dedicated specialized "GPU" soldered in AGP card. I have yet to see a GF3+ game scene that amazes me with adequate geometry complexity that EE is capable of with relatively much less frame rate fluctuation.

(Sorry, texture comparison is invalid here while GF3+ is equipped with a relatively whopping 128MB of dedicated video memory).

Edit: I completely with JackBurton.

 

[Adul]

goulch do you even have warcraft 3?
<-- plays with max settings at 1156*864 res on a geforce 4 MX 440. No slow down durong 3v3 online games. And I only have a athlon XP 1500+

 

[GoodRevrnd]

Isn't the point of techniques like displacement mapping to reduce the need for higher poly counts?

 

[Mr.IncrediblyBored]

Originally posted by: GoodRevrnd
Isn't the point of techniques like displacement mapping to reduce the need for higher poly counts?

Not exactly. Displacement mapping is just a way to store low-polygon meshes in memory, and have the graphics chip "tesselate" the old mesh into a new high-polygon mesh based on the values from the displacement map.

Quick explanation:

1) The engine uploads a vertex mesh (a model, rock, wall, table, anything) to the graphics chip
2) The engine uploads a displacement map. The map is treated similarly to a texture map. (i.e. its essentially a 2-dimensional matrix of numeric values).
3) The GPU "lays" the displacement map overtop the vertex mesh. The numeric values in each cell of the matrix will determine the new coordinates of the vertex.
4) The GPU will add new polygons to the mesh as needed to accomodate for the displacement map.

A common example:

A displacement map layed over a flat floor can turn it into a rolling mountain side. The floor can be a simple 4 vertex/2 polygon mesh. The displacement map may end up adding hundreds or thousands of additional vertices.

 

[CrazySaint]

Excuse my ignorance, but what does EE stand for?

 

[BDawg]

Originally posted by: CrazySaint
Excuse my ignorance, but what does EE stand for?

Emotion Engine - It's one of the PS2's processors.

 

[MasterHoss]

I didn't bother to read the article, but my question is... do you guys think the NV30 based cards will feature DDRII when they are released?? Or do you think it's like ATI's new cards just supporting DDRII ???????

 

[Adul]

I bet nvidia will try its best to use DDR-II

 

[Electric Amish]

Haha!!!

They'll do anything to try and bump up their stock price.

amish

 

[GoodRevrnd]

Originally posted by: Electric Amish
Haha!!!

They'll do anything to try and bump up their stock price.

amish

God forbid they live in a cave and release mediocre products!