Hi all,

Just had  a thought about this as I suppose it hasn't been thought about yet nor disussed, so I thought I might as well make a thread about this, as in what type of multi chip design would 3dfx have givven the Rampage and other designs beyond it.

since the Voodoo5 series and everything below it used Scan Line Interleave, Rampage does have a different design so to multi chip it would 3dfx used their Scan Line Interleave design or would they have used a more advanced way like ATi's Tile based Rendering or Nvidia's Split Frame Rendering?

If there is some knowledge about this it may be interesting to discuss :)

It may seem for a few that I am digging in an empty hole, but on there otherside there maybe something to learn about this part of 3dfx's last design.

Uhm, tile based rendering? I think you're a bit confused there, it has nothing to do with multi-chip setups. Tile based rendering is a different way of putting 3D objects on the screen (compared to polygon rendering which is used by all ATi and nVidia gpu's).

When it comes to multi-chip setups, ATi uses split screen rendering as default mode or alternate frame rendering when forced.

ATi uses tile based rendering  aka Super Tiled Rendering ask CJ from tweakers.net he told us this in the forums ;)
http://www.pcper.com/article.php?aid=168&type=expert

Nvidia uses Alternate Frame Rendering & Split Frame Rendering :)

3dfx uses Scan Line Interleave as we all know hehe, but since 3dx SLI would seem rather out dated I wonder what tehc they would of gone for I think Tile Based rendering but we'd never can be sure heh.

wrote on 15.03.09 at 17:34:19:

Hi all, Just had  a thought about this as I suppose it hasn't been thought about yet nor disussed, so I thought I might as well make a thread about this, as in what type of multi chip design would 3dfx have givven the Rampage and other designs beyond it.

One would think that 3dfx would use the standard SLI that they were already using, they after all already owned the process and it worked well on the last several products.

wrote on 15.03.09 at 22:54:23:

ATi uses tile based rendering  aka Super Tiled Rendering ask CJ from tweakers.net he told us this in the forums ;) http://www.pcper.com/article.php?aid=168&type=expert Nvidia uses Alternate Frame Rendering & Split Frame Rendering :) 3dfx uses Scan Line Interleave as we all know hehe, but since 3dx SLI would seem rather out dated I wonder what tehc they would of gone for I think Tile Based rendering but we'd never can be sure heh.

Tile based rendering is something completely different. The PowerVR comes to mind, likewise the Oak Warp 5.  ;)

Supertiling render mode is a Crossfire mode that splits the screen up in small little squares for rendering.  ;)

Similar names, but aren't even close to the same thing.  ;)

I wonder how many  ;)'s I can use before it starts to get annoying lol.

hmm oke but from the ex-Ati engineer I know he told me that ATi also uses Tile based Rendering when CF is used as for that part he should know best for that part heh.

But yeah the main card does the odd tiles where slave card does the even ones or it was the other way around but it does work something like this.

3dfx did the same but with lines instead.

wrote on 16.03.09 at 00:22:45:

hmm oke but from the ex-Ati engineer I know he told me that ATi also uses Tile based Rendering when CF is used as for that part he should know best for that part heh. But yeah the main card does the odd tiles where slave card does the even ones or it was the other way around but it does work something like this. 3dfx did the same but with lines instead.

You must have misunderstood him, because Tile Based Rendering is just that. Not Supertiling.

Don't tell me you have never heard about PowerVR and Tile Based Rendering?

Nope I haven't ;) before 3dfx I was a belovered S3 users  hehe nd before that my dad had a 486 DLC 33 with a Triden 64 heh my real computing days began in Australia in 1987 with a Microbee 500 after this in 1998 my dad came home with a Toshiba 386-40Mhz laptop which was from the Commonwealth Bank heh and in 1992 he bought that Cyrix 486-DLC 33 system and after this I began my own first build it had a sweet Diamond Stealth 2000 Pro with 2MB and was upgradable to 4MB Edo, 3Dfx entered this seup once VoodooGraphics hit the market, oh this system was a Pentium 120Mhz on some SpaceWalker mobo heh.

everyone was talking What is Pentium Pro ginna be like and about this new 3Dfx card that was comming which became the Diamond Monster 3D PCI 4MB which was a worldwide hit, the first V1 was fronm Orichid as far as I can recall though.

Anyways it was 3dfx full throttle since Voodoo5 and till now everything that went around it was either S3, NVIDIA or Matrox heh, ATi was something I learned well with their Radeon 9700 Pro back in 2002 heh.

Cool, well if you do want to know, do a search on wiki on PowerVR, there it explains as well what Tile Based Rendering is.

The NEC/PowerVR PCX1 was actually Voodoo graphics closest competitor.

The PowerVR and Oak were the only ones who used Tile Based Rendering. it's got nothing to do with multichip setups.

Well I have to admit ATi does use Tile Based Rendering as this site does explain:
http://www.theregister.co.uk/2005/05/06/ati_mvp_details/
One particularly interesting aspect of Hexus' revelation is ATI's use of a tile-based rendering scheme. Instead of doing the whole scene as one, the image is partitioned into squares, the better to minimise the bandwidth needed to bat a rendered tile from one card to the other.

More like Supertiling render mode to me, same process as it may seem.

So 3dfx would of used Scan Line Interleave hmm okay they did have everything they needed for multi chip in doors that's a good point Gary, but I do wonder if 3dfx would of made a more advanced way of it like for Mojop for example this would of been 3dfx's DX9 complaint GPU, I suppose to multi chip a GPU that uses advanced shaders Scan Line Interleave maynot be enough to do the job hence the advanced shaders that would of made it harder, as 3dfx was aimed for max speed with their Rampage and above I suppsoe Nvidia's Scalable Link interface may hve been their way out heh, this was the result of Nvidia after they bought over 3dfx tech + assets, a new form of SLI amerged heh.

For Tile Based Rendering: This is still truly not the same thing. Just splitting a Scene into tiles to render it on multiple GPUs doesn't make the GPU a true Tile-based Renderer. For that, the pipeline has to be significantly modified (or let's say: completely redesigned), because it renders top-down instead of bottom-up. This is also, what makes Hidden Surface Removal so much more efficient on a Tile-based Renderer.

However, true TBR also CAN have something to do with multi-chip rendering, as PowerVR has stated years back, that their TBR would do well scaling across multiple chips (each one rendering a set of tiles). Actually they even built such a thing for Arcade Machines. [URL=http://www.allbusiness.com/electronics/computer-equipment-personal-computers/7283758-1.html]See here[/URL].


Quote:

The PowerVR architecture allows content developers to create a single game for a variety of system platforms. Family members include PCX1, an integrated single-chip solution for personal computers and a multi-chip solution for video arcade machines

Both nVidia and ATi (and s3, Matrox, SiS, ...) refused to completely redesign their architectures for true TBR, they just optimized their actual brute-force renderer designs to be more and more efficient.

Note, that TBR is also known as "Deferred Rendering", so that term might help when googleing that up. (nV/ATi use "Immediate Rendering"). Don't confuse that with deferred Shading though!

As for the original question: I guess, when it comes to advanced shaders, that need data from portions of the screen that are not being rendered on the current GPU (e.g. Pixel Shader A running on VSA 1 needs the color of a pixel, which has currently only been drawn on VSA 2...), they might have just done AFR. I'm not sure about the problems in implementation and about bandwith requirements, but handling such shader dependencies over the SLI link seems rather complicated and cumbersome to me, whereas this is not needed in AFR, as the whole framebuffer exists in both GPU's local memory, not just half like it is when doing Scanline Interleaving..

Thanks for your nice explanation GAT :)

So Alternate Frame Rendering hmm year why not, would seem more doable to handle them shaders indeed.

also thanks for describing in perfect detail the difference between Deferred Rendering & Immediate Rendering

But yeah AFR could of been the next step for 3dfx' SLI, I even had an Adavnced version of Scan Line Interleave in mind heh, like Super Tiled Rendering, using odd and even squares via both chips to form a single image, this would also prevent from the cards rendering a seam over de screen.

The main problem I believe results from split framebuffers (Like in supertiling, TBR and Split-Frame Rendering). All those techniques split up the framebuffer to their respective GPUs. For PowerVR only on Arcade Machines, but still.

AFR doesn't split framebuffers, so it should be the safest method for modern shader compatibility I guess.

AFR yeah I suppose so, though Super Tiling isnt bad either it uses odd and even tiles on either VGA to form a single image, without this you'd never get a seam in the middle of your screen, I wonder how this further will progress, I suppsoe there aren't many other ways possible I guess for multi chip, as for 3dfx I think they would of used AFR yeah, as Super Tiling is used by ATi only afaik.