Next Gen Graphic Effects Are Amazing (Xbox 360, PS3)

[_phil_]

But AFAIK parallax map is a kind of displacement map.

that's a pretty long stretch.It's rather a bump (normalmap) with some self-occlusion (creating a little more sense of relief).

 

[Qroach]

titanio,

It supports those framebuffer formats, yes. But that, and supporting HDR effects in those modes easily or at all, are two different things - you need more than just FB format support. I've only ever seen HDR on X360 discussed in terms of FP10 to date - and one would wonder why it is there at all if FP16 or higher was supported, since the eDram would minimise the classic expense of those modes (bandwidth).

Don't make assumptions. have you looked to see if that is the case?

http://news.teamxbox.com/xbox/9487/ATI-Releases-Xbox-360-GPU-Technology-Papers/

 

[Titanio]

titanio,



Don't make assumptions. have you looked to see if that is the case?

http://news.teamxbox.com/xbox/9487/ATI-Releases-Xbox-360-GPU-Technology-Papers/

Where's FP16?

Fixed point is not floating point. It's an integer-based "hack" to avoid using floating point, with all the consequences that has for range and quality etc.

 

[Hardknock]

Some very interesting things here:

We haven't even scratched the surface of what this GPU can do. One thing that interests me the most is the Memexport. What is that about real-time ray tracing??

 

[Joe DeFuria]

What is that about real-time ray tracing??

It says nothing of the sort.

 

[groper]

that's a pretty long stretch.It's rather a bump (normalmap) with some self-occlusion (creating a little more sense of relief).

Actually this is not "a little more sence a relief" than normalM can do but a huge more sence of relief which make the 3D sence of walls , stone roads etc very imprecive.
I dont know what the heck technics Rare does use for PDO and Kameo but this is the first serious PM i can see in a game.
Especially if we compare it with slight , rather unimpresive , PM to Splinter Cell and the 50% failure of PM in FEAR (for bullet holes on wall) .

 

[Hardknock]

It says nothing of the sort.

Well I see Ray Tracing acceleration right underneath Physical Simulation - GPGPU.

What does that represent?

 

[Joe DeFuria]

Well I see Ray Tracing acceleration right underneath Physical Simulation - GPGPU.

What does that represent?

It represents acceleration of certain ray tracing operations. Not "real time ray tracing" which you said before.

I don't mean to be a pain in the ass, but the general populaton throws around "ray tracing" like it's some kind of holy grail for real-time 3d graphics. (A claim that is in and of itself misplaced, but that's another story). You say "real-time ray tracing"...and that's how some strange, wide-eyed impossible expctations start to spread as rumors.

 

[Qroach]

Where's FP16?

They slides are listing "special features/functionaluty" of the xenos GPU. FP 16 is supported as a standard feature in most ATI products already. The fixed point hack is there as a performance trade off if developers want to go that route.

Fixed point is not floating point. It's an integer-based "hack" to avoid using floating point, with all the consequences that has for range and quality etc.

It's nice to have the options to trade off quality versus speed.

 

[!eVo!-X Ant UK]

What is that about real-time ray tracing??

Can you point out were it even say's real time ray tracing ?? because all i can see is ray tracing acceleration. Beside if it were able to do ray tracing in real time dont you think microsoft and ATI would be shouting about about it??

 

[Titanio]

They slides are listing "special features/functionaluty" of the xenos GPU. FP 16 is supported as a standard feature in most ATI products already.

The point is, it's not. Given ATi's history of support for FP formats and blending (or more accurately, the lack of support), I would not assume it is there if they don't mention it. They mention fixed point 16-bit, for god's sake - if they're mentioning that, and did not mention FP16, it should maybe tell us something.

Thanks for the link, though. I've been wondering about it for a while, and it's enough clarification for me that FP16 HDR support isn't there - they just would not have mentioned a fixed point mode like that before FP16, if it was supported. Again, unless someone can confirm otherwise?

It's nice to have the options to trade off quality versus speed.

Including FP10 is an expense - more logic - why include it if full FP16 support is there? The eDram would take care of the biggest HDR-related expense at FP16.

 

[Hardknock]

It says Ray Tracing "acceleration" though. Why would they need acceleration of Ray Tracing if it couldn't perform Ray Tracing in real-time in the first place?

That's why I said real-time ray tracing, but I guess I'm wrong.

 

[Titanio]

It says Ray Tracing "acceleration" though. Why would they need acceleration of Ray Tracing if it couldn't perform Ray Tracing in real-time in the first place?

Since SM3.0 at least, "acceleration" of ray tracing has been discussed. See here:

http://www.gpgpu.org/cgi-bin/blosxom.cgi/Advanced Rendering/Global Illumination/index.html

It doesn't mean it's applicable to games, or as a general method of game rendering. Ray Tracing is also used for more than rendering also.

 

[_phil_]

Actually this is not "a little more sence a relief" than normalM can do but a huge more sence of relief which make the 3D sence of walls , stone roads etc very imprecive.
I dont know what the heck technics Rare does use for PDO and Kameo but this is the first serious PM i can see in a game.
Especially if we compare it with slight , rather unimpresive , PM to Splinter Cell and the 50% failure of PM in FEAR (for bullet holes on wall) .

i've got PM in maya
PM is normalmap + heightmap .you cannot "virtually displace" pixel very high.The hack would show its limits quickly.
kameo and pdo have the same PM than rainy toyshop demo from ATI.Ati probably gave them the sample code for it.

 

[ninelven]

Including FP10 is an expense - more logic - why include it if full FP16 support is there? The eDram would take care of the biggest HDR-related expense at FP16.

Relative to what is the question. It was my impression that fp10 took less physical space than fp16 when dealing with the 10MB of edram.

 

[Titanio]

Relative to what is the question. It was my impression that fp10 took less physical space than fp16 when dealing with the 10MB of edram.

Indeed, but for decent usage of the chip, tiling is assumed, that much is clear elsewhere in the design, so I'm not sure why they'd go to the trouble of FP10 here for space concerns if tiling was already being taken for granted. It might make sense in the R520 series, where bandwidth is a concern, but less so in Xenos I'd have thought.

Anyway, my point is that I've yet to see one piece of confirmation of FP16 HDR support. That ATi presentation linked to further up is pretty telling IMO.

 

[ninelven]

so I'm not sure why they'd go to the trouble of FP10 here for space concerns if tiling was already being taken for granted.

Well, I would think more tiles = greater potential performance hit. Of course, I know this was said to be virtually irrelevant... actually:

There is going to be an increase in cost here as the resultant data of some objects in the command queue may intersect multiple tiles, in which case the geometry will be processed for each tile (note that once it is transformed and setup the pixels that fall outside of the current rendering tile can be clipped and no further processing is required), however with the very large size of the tiles this will, for the most part, reduce the number of commands that span multiple tiles and need to be processed more than once. Bear in mind that going from one FSAA depth to the next one up in the same resolution shouldn't affect Xenos too much in terms of sample processing as the ROP's and bandwidth are designed to operate with 4x FSAA all the time, so there is no extra cost in terms of sub sample read / write / blends, although there is a small cost in the shaders where extra colour samples will need to be calculated for pixels that cover geometry edges. So in terms of supporting FSAA the developers really only need to care about whether they wish to utilise this tiling solution or not when deciding what depth of FSAA to use (with consideration to the depth of the buffers they require as well). ATI have been quoted as suggesting that 720p resolutions with 4x FSAA, which would require three tiles, has about 95% of the performance of 2x FSAA.

So it would look like each additional tile has <=5% overhead, say 4% just for kicks. Then with 3 additional tiles (4xAA) that would be a 12-15% performance penalty or 4-5% at (2xAA). As for fp16 vs fx16, I have no idea.

 

[Fafalada]

I don't think it was ever clarified what HDR precisions, beyond FP10, Xenos supports.

It supports other FP formats, but FP10 is the only one that has full functionality (it's not just performance differences).

Of course that's not something you'd exactly want to advertise - suffice to say FP10 is the format you're supposed to be using.

 

[ector]

And if you "play your bits right", you should be able to get some pretty damn neat graphics out of FP10... I'd love to have a full featured format like that going full speed on PC GPU:s. Halving your fillrate to get FP16 or fixed point 16-bit just doesn't seem worth it when you have FP10.

 

[Titanio]

It supports other FP formats, but FP10 is the only one that has full functionality (it's not just performance differences).

Of course that's not something you'd exactly want to advertise - suffice to say FP10 is the format you're supposed to be using.

Thanks for clarifying, that's pretty much what I thought.

I wouldn't expect them to advertise it, but I guess that just highlights the problem with all of our information regarding Xenos. It's basically all ATi advertising, as all our info has been filtered through them. You basically have to pay as much attention to what they don't say as what they do.

Anyway, I'm not a HDR expert, but I don't think it was a wise decision not to have full support for FP16, but it's not surprising given ATi's attitude on that up until the R520 series. The dynamic range offered by FP10 is a tiny fraction of that of FP16 (1024 possible values per RGB component vs 65,536 possible values per component..?), and would require careful management of your data/art to try and hide shortcomings, if it could be done at all. It could potentially also hand PS3 games a default and immediate graphical advantage - probably the most commonly identified distinguishing feature of PS3 stuff to date has been "better lighting". And maybe this is why (?)