Toshiba, Sony close to 65nm sample production

[Panajev2001a]

Dave,

here it is, your poor man's tiler with NO T&L cost for multi-pass rendering

http://makeashorterlink.com/?S48235ED6

 

[Panajev2001a]

Each PE would likely work on Rasterizing separate Triangles: after-all we expect next-generation to have average Triangle size to fall down to even 1-2 Pixels or lower in some optimized engines and dedicating 4 Pixel Pipelines or more to fill a single Triangle might be wasted.

A good point, particularly for a console - well, non-HDTV at least - where resolutions are low.

Of course, it has the flip side that on larger triangles such an architecture could be inefficient. There is significant cost efficiency and reduction of redundant operations in raising granularity.

You are correct, moving to HDTV will increase Polygons size, but I doubt that all consoles next-generation will have 1080i or 1080p as standard resolution while 480i and 480p will be the likely minimum everyone will cater to.

Also, being a fan for an optimized micro-polygon based approach I cannot really root for larger and larger triangle sizes now, can I ?

 

[Dave Baumann]

here it is, your poor man's tiler with NO T&L cost for multi-pass rendering

Yes, it still is a poor mans tiler. As I said, you'll either use RAM or use geometry performance - in this case its the former as there is a buffer for the geometry. It'll loose out on the advantages of regionalised rendering as well unless there is onchip RAM somewhere that can account for the full frame buffer and Z buffer (with FSAA sizes if FSAA is to be applied).

I'm curious to see that this mentions two processors - given the the PS3 patents you guy have unearthed so far this doesn't seem to fit.

 

[Deadmeat]

...

Just going by what I see, I doubt you'll have multiple thread across each APU

Neither do I. I clearly told you that the PSX3 was going for an MPI-style message passing architecture, not a multithreading. 128 KB exclusive Local Storage that is inaccessible to other APUs is a give-away that PSX3 HAD TO implement some kind of message passing architecture. I wouldn't be too surprised if the first PSX3 dev documentation has lots of API and programming manual from BlueGene/L...

that would be for the system to handle I'd say the threads you exists at the PU level.

Of course every process gets at least one thread, or the process wouldn't run...

 

[zidane1strife]

- how many cycles would it take non-decidated hardware?

Remember, non-dedicated h/w... has in a few thousand seconds(for lowest complexity scenes), AFAIK without acceleration... done what is deemed impossible for the nv30, and the like... these grphx have wowed us, in very big screens...

The latest of these non-dedicated efforts, is comparable in perf to the expected performance of the ps3, only it's stretched across an entire room instead of a console, and likely without the h/w acceleration here and there that has been said is likely for the latter...

As we go back in time, the gphx of these efforts still outshine anything we have today in real-time... but the performance of these machines drops order by order of magnitude...

So at least we know that with orders of magnitude less performance, in a few thousand secs, quite amazing stuff can be done, very clean stuff with superb IQ, IQ so clean that it probably took quite some time to make it so, especially without acceleration... and with extreme lvls of detail and rez, for it is to be blown up, and seen in a very big screen.... all of these can be significantly reduced if the target is to be a far smaller screen, and the audience isn't as critical...

 

[Dio]

I'd argue that a lot of the difference in quality between 'cinematic' renders and what's achievable in vastly faster time on a VPU is mostly down to the amount of time and effort invested to create the scene, and that 'movie shots' have far more control than is available in games (control over position, lighting, background, etc.). Also, there's a hard ceiling for VPU's: nobody cares about anything rendered at 1 frame in 2 seconds, while the movie pros aren't particularly bothered if the rendering time increases a factor of ten.

Remember also that the games guys aren't targetting modern high-spec VPU's yet. There's no desire to, when the baseline required spec on PC is a GF1, and even the peak of the console market is based on relatively ancient technology.

I welcome the likelihood that next-generation consoles will change this, but I fear for the games developers, who will have to spend even more time and effort to get up towards the limits of the hardware.

 

[zidane1strife]

I'd argue that a lot of the difference in quality between 'cinematic' renders and what's achievable in vastly faster time on a VPU is mostly down to the amount of time and effort invested to create the scene, and that 'movie shots' have far more control than is available in games (control over position, lighting, background, etc.).

Let's not forget physics here...

 

[Panajev2001a]

but I fear for the games developers, who will have to spend even more time and effort to get up towards the limits of the hardware.

I know what you mean, but you make it seem like they are slave labour

Well... sometimes... it might not be far from the truth

 

[Dio]

The realities of the market will drive many more independent developers to the wall - only the big developers / publishers will have the resources to find the big hits and absorb the ones that don't sell.

Unless some solution is found...

 

[Super Grafx]

Remember also that the games guys aren't targetting modern high-spec VPU's yet. There's no desire to, when the baseline required spec on PC is a GF1, and even the peak of the console market is based on relatively ancient technology.

indeed. PS2 is 1997-1999 technology. GameCube is 1998-2000 technology and Xbox is 2000-2001 technology.

 

[cthellis42]

The realities of the market will drive many more independent developers to the wall - only the big developers / publishers will have the resources to find the big hits and absorb the ones that don't sell.

Unless some solution is found...

Unlikely. The reality in most markets, too. Short of some stand-out efforts that come through, the consumer pushes for "bigger and better" as much as possible--and that means fewer and fewer can provide it. Developers have to keep becoming bigger and better as well.

 

[nobie]

There's been some success on the PC with "bargain bin" titles like Deer Hunter, and some say that shareware will make a comeback soon. In Japan there seem to be alot of games produced by smaller developers at a lower budget, but unfortuneately most of these games never get translated and released in the west.

 

[Dio]

Unlikely. The reality in most markets, too. Short of some stand-out efforts that come through, the consumer pushes for "bigger and better" as much as possible--and that means fewer and fewer can provide it. Developers have to keep becoming bigger and better as well.

What the industry is looking for is the equivalent of independent cinema. It's a major problem that there isn't a low-cost choice for making games; I mean, there isn't even a middle ground.

Movies get made with budgets up to $200m; a typical film might have a $20m-$40m budget; good, 'cheap' films can be made for $2m or even less.

There isn't this option in the games industry - the budget for a game can't be much less than $1m, and can't be much more than $10m (and even then only the Blizzards of this world can afford to spend that much). Realistically game budgets are in the $1-3m range, which means there's massively less scaling than there is in the film industry.

 

[Fafalada]

It's a major problem that there isn't a low-cost choice for making games; I mean, there isn't even a middle ground.
There isn't this option in the games industry - the budget for a game can't be much less than $1m, and can't be much more than $10m (and even then only the Blizzards of this world can afford to spend that much).

That's actually not quite true right now.
At the moment, the lowcost choice is GBA (technically PSOne games are still made now and then too but it's not exactly comparable volumes so I'm not mentioning it) - and profitable games for it are made with budgets as low as 100k$.
Next year there will also be about a hundred of NGage metoo's which all fit the same lowcost bracket as GBA games do right now (unlike GBA and NGage they will have some native 3d capabilities but we're still talking PSOne levels of detail at best).

With arrival of PSP level machines it's true even that market segment will eventually grow more expensive, but it'll still be a lowcost alternative to PS3/XBox2 etc.

 

[PatrickL]

You really hope to see hundreds of Ngage games. For what i know it seems that sales are very low in Europe.

 

[PC-Engine]

GBA games sell for $15-$35. Nobody is going to buy N-Gage games at that price except those people who'll buy UMD movies they already own on DVD just so they can watch those movies on-the-go.

 

[London Geezer]

GBA games sell for $15-$35. Nobody is going to buy N-Gage games at that price except those people who'll buy UMD movies they already own on DVD just so they can watch those movies on-the-go.

What if (serious question, i really don't know) u could record movies on blank UMDs to then watch the on the go? I would certainly consider buying that...

 

[notAFanB]

GBA games sell for $15-$35. Nobody is going to buy N-Gage games at that price except those people who'll buy UMD movies they already own on DVD just so they can watch those movies on-the-go.

What if (serious question, i really don't know) u could record movies on blank UMDs to then watch the on the go? I would certainly consider buying that...

well I'm sold...

seriously tho this isn't even remotely likely as long as they think they can milk revenue from it.

 

[Dio]

I wasn't really thinking GBA or mobile when I wrote my comment - but yes, that is an out for many small developers.

 

[Fafalada]

You really hope to see hundreds of Ngage games.

I was talking about phone-game machine wannabees - or if you prefer, NGage "clones", not games.
NGage was only the start of the flood of this kind of hardware.