Half-Life 2 XB360 won't be using tiling to achieve A.A.

[Titanio]

*snip*

I thought Brimstone might be conversing with Valve generally about the console versions, not just the 360 version.

Speaking of which, Brimstone - did you ask Valve for any further elaboration on that comment? They mightn't tell you, but you never know.

 

[TheChefO]

(this is one area where Sony can/needs to go for the jugular)

Does this really translate to the living room? How many people will have their consoles setup to a monitor on a desk? Sure they may get a few who don't have the jack to keep up with a high end pc but I don't get the feeling this would change the tides in their favor for the console fps segment.

 

[Jesus2006]

this means they are using the 360 GPU's flexibility to get AA + visuals without using Tiling to an extent it would look even better than if it used Tiling.

Pretty optimistic, but certanly quite impossible on this architecture, especially since the ROPs are inside the EDRAM so you "have to" use it in some way.

But for the resolution, i think there are already games running in lower resolutions, one that does for sure is PGR3, and another one i heard is Oblivion (read on some forum, but got no link). At least Oblivion is the one which has least visual differences between 480p and 720p on my screen (which is 50" )

 

[TheChefO]

if you do figure it out, we'll dub it nAAo360

fixed

 

[Farid]

The only solutions they have, to get better performance than MSAA + Tiling on X360, is to use a smaller backbuffer that fits into the eDRAM without any tiling (1024x600 + 2XMSAA) - that wouldn't give great IQ results, but it would save BW, given that there'd be only 33% less pixels to shade -.
The other solution would be not to use AA at all, and call Motion Blur (classical screenspace vector based effect, not an accumulation buffer effect like offline renderer) and/or DOF "a form of AA". Which would be incorrect, but it wouldn't the first time.

Anyway, why should they be concerned with GPU performance issues for a HL² port?
Are they upgrading the game visually to the point of being potentially GPU bound on X360?

but there are other ways of doing work to get AA aside from tiling whcih is just 1 of the options.

Such as?
How are you supposed to get a real 4XAA (or even 2XAA), MSAA or SSAA, on a 720p rendertarget without using tiling while, at the same time, obtaining better performances than what you'd get if you were using 4XMSAA and tiling on X360?

memexport

MEMEXPORT, what do you want to do with that, with regards to Anti-Aliasing?

I thought tiling was essential to get "free" AA so you could spend on other effects to get visual quality?

The Xenos ROPs, thanks to the eDRAM BW, are capable of sustaining a free 2XMSAA and a low penalty 4XMSAA, but that's only concerning the framebuffer, or part of it, that is in the eDRAM.
Tiling isn't getting you free AA, eDRAM gets you that, tilling permits you to use less eDRAM to render a frame, reducing the transistor cost of the eDRAM. But tiling has obviously a performance cost; if it didn't Ati could had just as well put a very little amount of eDRAM and then increased the number of tiles necessary for rendering a full frame. The performance hit of tiling varies depending on various factor such as the geometry complexity of the scene, etc.

 

[Serenity Painted Death]

Does this really translate to the living room? How many people will have their consoles setup to a monitor on a desk? Sure they may get a few who don't have the jack to keep up with a high end pc but I don't get the feeling this would change the tides in their favor for the console fps segment.

1) It would help. It would make every multiplat FPS instantly superior on PS3.

2) It would be better for PS3 owners. KB/M is vastly superior (though I suppose this gets into the issue of having to separate kb/m and gamepad users... cause the results wouldn't be very amusing)

3) Why would they need their consoles set upt to a monitor on a desk? I'm sure you'd just use a wireless keyboard and mouse at your leisure...

 

[AlNom]

3) Why would they need their consoles set upt to a monitor on a desk? I'm sure you'd just use a wireless keyboard and mouse at your leisure...

Personally, I have issues using mice in the living room. Either I'm hunched over the coffee table, or I'm struggling to get a decent surface for the optical mouse. Add to the latter situation where my hand wouldn't be at a proper level to avoid wrist cramps... Even if I used the arms of the couch, it's not great for aiming.

edit: not saying it shouldn't be an option, but I find it somewhat troublesome for a living room.

 

[TheChefO]

1) It would help. It would make every multiplat FPS instantly superior on PS3.

2) It would be better for PS3 owners. KB/M is vastly superior (though I suppose this gets into the issue of having to separate kb/m and gamepad users... cause the results wouldn't be very amusing)

3) Why would they need their consoles set upt to a monitor on a desk? I'm sure you'd just use a wireless keyboard and mouse at your leisure...

1) for those that have the appropriate setup

2) for those that have the appropriate setup (agreed - faster, more accurate)

3) and place the mouse where exactly? I'm not sure what kind of setup you have but I've delt with the wireless mouse on the couch in many configs and it's a friggin pain which goes against the 'faster, more accurate' advantage. hence - desk.

edit- agreed alstrong

 

[Serenity Painted Death]

On a table? With the keyboard?

On a nice sturdy/flat surface on your couch? (depending on firmness?)

I can think of lots of things, depending. Whatever minor sacrifices in comfort... I, for one, would be willing to make, for vastly superior control (not only in FPS, but also strategy games if they ever catch on).

I'm used to insane multitasking with mouse/kb anyway, so maybe my perspective is a bit skewed, but there is only upside to allowing the option.

 

[TheChefO]

but there is only upside to allowing the option.

agreed - nothing wrong with giving the option:smile:

however regulating gamers with kb/mouse vs gamers without would be an added level of complexity and it would also muddy their message at this point by allowing it on 360 in contrast to ps3 which is attempting to align itself with pc's.

 

[TheChefO]

The only solutions they have, to get better performance than MSAA + Tiling on X360, is to use a smaller backbuffer that fits into the eDRAM without any tiling (1024x600 + 2XMSAA) - that wouldn't give great IQ results, but it would save BW, given that there'd be only 33% less pixels to shade -.
The other solution would be not to use AA at all, and call Motion Blur (classical screenspace vector based effect, not an accumulation buffer effect like offline renderer) and/or DOF "a form of AA". Which would be incorrect, but it wouldn't the first time.

Such as?
How are you supposed to get a real 4XAA (or even 2XAA), MSAA or SSAA, on a 720p rendertarget without using tiling while, at the same time, obtaining better performances than what you'd get if you were using 4XMSAA and tiling on X360?

So by your estimation the only reasonable method would be to render at a lower res with aa to fit into the Edram? (agreed mb or dof != AA)

screenspace vector based effect? would you be so kind as to explain or point me in a direction to find info on this?

 

[ihamoitc2005]

Alu

Xenos has a lot of extra ALUs not being used much

I did not know this my friend, thank you for this information. This information can explain many things I have been wondering.

I wonder why are they not using extra ALU's for shaders? Maybe because unified shader is very new technology? Maybe this is directx api problem? I dont know. I am sure soon they will use full Xenos power!

ps.

My Indian friend tells me that ALU means potato in his language. This can make very bad jokes in GPU forum.

 

[Acert93]

The reason is simply because shaders are still fairly short, the SM3.0 hardware on the market until late Fall 2006 was no too great at dynamic branching, vertex texturing, and other SM3.0 features, and because the typical trend has been 1:1 Texture:Math in GPUs, while Xenos and the X1900 go for a more aggressive 1:3 ratio. Games are just not developed for that yet. That is why you read stuff like Rare getting 1M particles, all running on the GPU, up and running without impacting the framerate.

A number of developers note how you try to minimize bottlenecks and focus on the strengths of a platform. When you drop the ball like MS did and did not deliver hardware with anything closely resembling the featureset of your platforum until 3 months before titles need to go gold... Bleh. Kind of ironic how MS launched early and went with a totally new architecture that was available pretty late in the development cycle and yet Sony waited a year and went with an architectural evolution of a 2004 GPU and put similarly performing / similar feature GPUs in the early devkits to get developers on the ground running.

 

[Megadrive1988]

but I think as time goes on, developers will show Xenos' strengths while RSX will show its limitations, IMO

 

[Bobbler]

but I think as time goes on, developers will show Xenos' strengths while RSX will show its limitations, IMO

What an absurd thing to say.

 

[nAo]

but I think as time goes on, developers will show Xenos' strengths while RSX will show its limitations, IMO

A supposedly efficient unified architecture should show its strengths with far less effort..so I believe it should be the other way around

 

[ihamoitc2005]

?

The reason is simply because shaders are still fairly short, the SM3.0 hardware on the market until late Fall 2006 was no too great at dynamic branching, vertex texturing, and other SM3.0 features, and because the typical trend has been 1:1 Texture:Math in GPUs, while Xenos and the X1900 go for a more aggressive 1:3 ratio. Games are just not developed for that yet.

I am a little bit confused of what you say my friend. Please forgive this but it is a little confusing for me what you say.

Do you say SM3.0 shader is too short to have 100% ALU rate for Xenos ALUs? So for Xenos to have 100% ALU rate then shader must be longer? I do not understand this very well but if you explain I will be happy. Thank you.

Also I also have one more question. Is ALU rate for 1900 and Xenos less than 850 and 7900 for SM3.0 shader game like maybe Oblivion?

 

[Moloch]

I am a little bit confused of what you say my friend. Please forgive this but it is a little confusing for me what you say.

Do you say SM3.0 shader is too short to have 100% ALU rate for Xenos ALUs? So for Xenos to have 100% ALU rate then shader must be longer? I do not understand this very well but if you explain I will be happy. Thank you.

Also I also have one more question. Is ALU rate for 1900 and Xenos less than 850 and 7900 for SM3.0 shader game like maybe Oblivion?

Oblivion is SM2 IIRC
http://www.beyond3d.com/forum/showthread.php?t=29537&highlight=oblivion+shaders

 

[DarkRage]

A supposedly efficient unified architecture should show its strengths with far less effort..so I believe it should be the other way around

I guess that can be said about Cell too. But I remember a different thing said from your own team.

An unified architecture *might* allow for new rendering techniques not available at acceptable performance with traditional architectures.
Something similar with Cell, BTW.

 

[Titanio]

I guess that can be said about Cell too.

*scratches head* In bizzaro world, maybe!

Reading Megadrive1988's comment, funny thing is, in order to maximise the strengths/capabilities of a particular chip you kind of have to run up against its limitations.