What would the "perfect" 3d accelerator be like?

[Livecoma]

There seem to be damn near countless ways of accomplishing the same picture, and they all have their unique pros and cons. Is there a "perfect" way to render 3d graphics on a PC when you take todays limitations away?

That’s it basically. You have 2015's manufacturing processes at your fingertips. You can fit billions of transistors all in one cost effective package. Whats the chip going to do? hehe

[Don't Ask]

In a perfect world, the "perfect" 3d accelerator would be a card without a spazzy hyped up gymic model name and a ridiculous looking logo. This would be a card that could properley implement the features it employs and wouldn't hog down on 2 year old games in it's prime. This card would have features that won't have to wait for dev 2 years down the road and won't be useless by the time they are. The perfect 3d card doesn't come with a price tag over 200 bucks. The perfect card is perfect for everybody.

But back to reality...
..

...ok back to fantasizing

[Nexus]

Raytracing

[Bigus Dickus]

Raytracing would be nice. The answer all depends on the technology available at the time. With sufficient processing power (and this is many, many years away) you could do away with the GPU completely, or simply have a dual CPU system with one processor dedicated to graphics. With sufficient processing power, any effect, any feature, any method of rendering can be emulated by a general purpose processor. It would simply (what an understatement) be a matter of coding drivers to emulate the desired function.

For example, with a suitable driver, any video card should be able to shunt any programmable vertex and pixel shading functions to the CPU. Today this would be insanely slow, due to restrictions on processing power, bus bandwidth, and memory size (not to mention driver developer intelligence :wink:).

Twenty-five years from now... who knows? Perhaps there will be no video cards. A powerful enough CPU simply wouldn't need one. Then again, perhaps developers will choose to stick a GPU format similar to the current one for the purpose of simplifying programming... devlop a high level API that assists in simplifying the programming of some desired effect, instead of coding it manually. Of course, there's no reason why such an API could not be a general "driver plug-in" that breaks down the high level instructions for a general purpose CPU to process.

It's all a matter of who does the programming, and how to streamline the process such that absurd repetitiveness is avoided. The only reason for application specific hardware though, IMO, is lack of processing power.

[Matt]

The perfect 3D card would have 128-tap anistropic filtering, fillrate enough to do 16x FSAA at 1600x1200x32 at acceptable performances. It would be able to bake bread for me, summon lesbians to my door, and scare away Jehovah Witnesses on a whim. Oh, and feed my dogs. :smile:

[CMcK]

The perfect 3d accelerator would, for me, be one which actually delivers the fillrate quoted on the box.

[nAo]

On 2002-02-26 07:31, Matt wrote:
The perfect 3D card would have 128-tap anistropic filtering, fillrate enough to do 16x FSAA at 1600x1200x32 at acceptable performances.
NV30? :smile:

[pascal]

- year 2010
- .06 micron EDRAM process.
- 64 edram.
- 8 pipelines dual TMU with 128 loopback at 1GHz
- sustained 250 millions polygons per second.
- 512MB QDRAM at 2GHz (128 bits)
- fully programable Hardware T&L
- pure OpenGL 2.1
- Perfect for Doom4 with outdoor scenes

[midnight creature]

"summon lesbians to my door"

Being a former partner in a threesome I would warn you away from such an arrangment even admonish you to reconsider.

[Matt]

On 2002-02-26 23:07, midnight creature wrote:
"summon lesbians to my door"

Being a former partner in a threesome I would warn you away from such an arrangment even admonish you to reconsider.


LOL, guess I'll take your word for it. :wink:

nAo:

We can only hope... :smile:

[DemoCoder]

Raytracing is overrated and computationally expensive. It doesn't solve the global illumination problem either.

In most 3D rendered CGI today, scan line rasterizers are used in the vast majority of circumstances. With RenderMan, if you need ray-traced reflections, you can either use a ray-tracer for that particular scene (ExLuna/BMRT/etc), or you can use a custom shader.

I'll take a card that can run OpenGL2.0/Stanford's Real Time Shading Language anyday over a card that accelerates ray-tracing. I think radiosity acceleration would be far more useful too.




<font size=-1>[ This Message was edited by: DemoCoder on 2002-02-27 01:46 ]</font>

[multigl]

amen democoder amen

[3dcgi]

On 2002-02-26 07:31, Matt wrote:
The perfect 3D card would have 128-tap anistropic filtering, fillrate enough to do 16x FSAA at 1600x1200x32 at acceptable performances.


The perfect 3d card won't need antialiasing. It will render resolutions high enough that we can't see the aliasing. Of course we'll need better monitors to go along with this rendering power.

[pascal]

I forgot

Sthocastic AA like superscene or betterhttp://www.3dlabs.com/product/technology/superscene_antialiasing.htm :cool:

[Dave B(TotalVR)]

"The perfect 3d accelerator would, for me, be one which actually delivers the fillrate quoted on the box."

So KYRO II then?:P

Anyway, I wanna see a 4 piped dual TMU (for single cycle anisotropic) tiler running at 300 Mhz, dual 300 Mhz 128 bit DDR memory bus going to texture RAM and framebuffer/tile setup RAM. A 2 chip board which efficiently shares tiles between the two procs. That gives 2.4 Gigapixels/s fillrate and a very high efficiency. On top of that DX9 support and the phattest TnL unit. Free MSAA and the nice nifty tiler only features like hardware translucency sorting and generalised modifier volumes.

The technology for this aint far off and it would be an absolute beast. In fact, it could probably be done today (if the DX9 specs were out, are they?)

[Musivarius]

The perfect 3D accelerator would not only render very nice teapots, it would make a very nice cup of tea too. :smile:

[CMcK]

On 2002-02-28 14:17, Dave B(TotalVR) wrote:
"The perfect 3d accelerator would, for me, be one which actually delivers the fillrate quoted on the box."

So KYRO II then?:P


I'm already a Kyro II owner.

A perfect 3d card would have free hiqh quality FSAA, free aniso, higher colour depth, hardware support for loads of lights, proper shadows. And perhaps an onboard DSP so developers can create their own effects, like procedural textures, without having to use the CPU.

[KingRoLo]

All of the above with the core running @ :grin: 1Ghz :grin:

RoLo

[Doomtrooper]

Useable features with a price of $299, not a penny more.

[Judas]

Someone forgot to mention 64bit internal true color :wink:

[Exposed]

I'm surprised no one stated the obvious...the perfect accelerator is one that renders without that annoying artificial fog to hide the back clipping plane. It's everywhere....whether it be role playing games or flight simulators, almost anything rendered in 3D. And usually its only purpose is for performance considerations. It's like a cockroach....it must be squashed :wink:

Only then will we have a perfect 3d world :smile:

[zidane1strife]

The perfect 3d accelerator would have to be one which features infinite bandwith, processing ability, resolution, effects, memory, fillrate, can render in infinite ways(like holo, materialing stuff, directly transmitting to the mind, etc...) etc... It would also be easy to program, and with no bottlenecks... It would be a VERY VERY long time before they make one...heheh

[Livecoma]

I wonder when we are going to see &lt;.13 graphic chips. Intel is supposedly introducing their .09 setup early next year IIRC, but how long will it take TSMC to get there?

Muhauhauah how much embedded memory could you fit on a .09 chip? :grin:

BTW I find it really interesting that Intel does not seriously enter the graphics market. They can obviously afford it, and they have a bazillion plants with AFAIK the most advanced fabrication process available...

I guess there will probally never be a perfect accelerator, but atleast we can count on them getting better every generation heheh...

[pascal]

After thinking a lot:

The 3D card will have ist own optimized, powerfull and flexible game engine with lots of usefull tools and great software support.

We will have games ready to use all its power after 6 months with the card :grin:

[Dave B(TotalVR)]

What would the "perfect" 3d accelertor be like?

That's easy, MINE! :lol: