Xenos hardware tesselator

Lets hope. That is one area I think Sony has excelled--a desktop GPU and tools. PC devs have more time/resources (tools, engines, experience) with modern GPUs and their evolution, and leverging that is a good thing. Creating a business model that targets PC<>Console migration, like MS has, but creating a graphic sub-system that is outside the superset of APIs as well as creating a rendering circumstance that is non-intuitive to the workflow of said developers you are trying to leverage... seems like an odd path to take. Almost like RSX and Xenos should have swapped places.


I think MS was more looking toward the future. If you notice Windows VISTA aswell as Nvidia and ATi are all steering towards Unified Shaders. If that's the case 360 will have the advantage in future PC<>console ports over a more conventional GPU.
 
Basically I think the system is very neat, but if it's hardly ever utilised like it was meant to be, then something somewhere has gone wrong.

Of course, but we're one year after launch. Libs and devs still have a long way to go.

I think Halo 3 and the Forza 2 (PGR3.5 ;)) engines will lead the way.

Maybe they will start using the Halo 3 engine as middleware too...
 
I think it's predicated tiling rendering. (shoot me if I'm wrong :LOL:)

edit: Even though it's a good compromise, I would say that the pros will have to be pretty darn amazing to outset the cons. I mean the system has been out for a year now. The best looking title is a game based on UE3 which doesn't use tiling also given the fact that tons of games are multiplatform out of which possibly none? will use tiling. It also seems like UE3 will become very popular engine for X360 and if there are future updates that will make tiling possible remains to be seen. So I might be sadly mistaken when I say that it looks like about a handfull of first party titles will actually use tiling and even in those titles developers will have to do lot's of work to get it up and running properly. So often when I think about this, I feel like maybe it could have been more usefull if the transistors had been put elsewhere. Basically I think the system is very neat, but if it's hardly ever utilised like it was meant to be, then something somewhere has gone wrong. of course I will get through this by just looking at the games which are very nice indeed:) I don't mind if somebody shoots my line of though to the ground though and explains to me how eDram actually is the best thing since sliced bread.

Too me you'd have to ask, "what would Xenos do without the EDRAM but with a dedicated 128-bit bus?". Because remember, 256 buss es were considered no-no's on these machines.

And to that, we really dont know. Overall though, the 360 is acquitting itself surprisingly well, better than I expected actually, against the competition so far, so it's hard to say.

The EDRAM gets more interesting as time goes on. Because it's notable that EDRAM will drop in cost much more than normal RAM. Someday the Xenos will all be one chip, and easy and cheap to manufacture.

I think 30 MB EDRAM might be a must for all next (next) gen consoles. It will become more doable and attractive as transistors get cheap. Well, assuming a 720P framebuffer is the target. More if it's a higher target..
________
Web shows
 
Last edited by a moderator:
I think 30 MB EDRAM might be a must for all next (next) gen consoles. It will become more doable and attractive as transistors get cheap. Well, assuming a 720P framebuffer is the target. More if it's a higher target..
By next-gen, there'll be the option of GDDR4 or better and 60+ GB/s. The relative benefits of eDRAM are reducing over time.
 
By next-gen, there'll be the option of GDDR4 or better and 60+ GB/s. The relative benefits of eDRAM are reducing over time.


This tells me nothing, if these companies are still scared of >128 bit busses, the divide in bandwidth versus the top PC cards will be much larger in the future, not smaller, as they have moved to 384-bit and 512-bit.

What you can get by ramping GDDR speeds is a pittance by comparison, if they're not going to increase bus width or go to some new technology. Current champ 2 GHZ GDDR4 would only get you slightly above 30 GB/s on 128 bit, while R600 is rumored to push 160 GB/s + (and G80 is already at 86+).

In contrast, two or more process shrinks may make 30MB+ EDRAM a pittance. Although I still wonder if you'll be running out of texture BW, and need that >128 bus anyway...
________
Live Sex
 
Last edited by a moderator:
This tells me nothing, if these companies are still scared of >128 bit busses, the divide in bandwidth versus the top PC cards will be much larger in the future, not smaller, as they have moved to 384-bit and 512-bit.

What you can get by ramping GDDR speeds is a pittance by comparison, if they're not going to increase bus width or go to some new technology. Current champ 2 GHZ GDDR4 would only get you slightly above 30 GB/s on 128 bit, while R600 is rumored to push 160 GB/s + (and G80 is already at 86+).

In contrast, two or more process shrinks may make 30MB+ EDRAM a pittance. Although I still wonder if you'll be running out of texture BW, and need that >128 bus anyway...


If the EDRAM moves onchip what would be the new bandwidth rate?
 
eDRAM and a wide external bus are the two opposite extremes to solving the bandwidth issue, equally inefficient.
 
The problem with eDRAM is that we need something like 100MB of eDRAM before we really solve the framebuffer problem.
 
eDRAM takes away significant amounts of silicon for potential execution logic regardless of whether the amount of eDRAM is "enough".
 
It's a pretty useful feature even if not too flexible, not only to implement displacement mapping, but can be used for other purposes (very fast particles instancing springs to mind), although it's not as flexible as a Cell or a Geometry Shader.
Other than displace map and some utter particles what would more flexiblity allow you to do? Treat the geometry (ie. deform a car after an accident) before tesselating?

If so, you can use Xenon or even memexport to achieve the level of flexibility you want and them send the data for the tesselator? I mean, does it accept that kinda of input, or if that was the case you would have to tessalate with other means?

And about perfomance... being a fixed hardware function one would expect that its kinda fast... Could you give any figure on how this tesselator compares in perfomance when doing displacement map, to lets say, a full featured GS powered Dx10 gpu?

Thanks for the time :p

As far as I can tell, Xenos is able to resolve the individual AA samples in a render target when AA is applied - this means that deferred shading and AA are compatible in Xenos, as is also the case in D3D10.

It is only DX9's inability to allow the GPU to see AA samples individually (when the render target is viewed as a texture) that prevents deferred shading (e.g. in the PC version of GR:AW) from having AA.

Jawed

I remember you pointing that out for Dx10 Gpus, on the gamefest slide thread... When did xenos joined the club too? :p

Does this really means that even with deferred shading xenos can apply AA right? But there's an extra cost for it?
 
Last edited by a moderator:
Back
Top