The Game Technology discussion thread *Read first post before posting*

The interesting thing about KZ2 SPU tech in my opinion is that they do all postprocessing on quarter resolution frames rendered to main mem by RSX (which also triggers SPU pp via interrupts), and they are computationally bound instead of BW. The resulting lowres postprocessed frames are alpha blended in a smart way so that unaffected pixels are still from HD frame as opposed to lowres one.

They also have some vague numbers regarding the percentage cost of RSX vs percentage cost of SPU for the tasks they moved to SPUs.

Apparently more optimized version of their tech is already in EDGE now.

Overall though, while they seem to be doing nice stuff, I cannot help but feel that Insomniac's shader approach to SPU programming in general (not specifically gfx aid) is more "interesting" than GG or even Naughty Dog (Ice Team).
 
Did anyone else read the slides from Tri-Ace about their shader and post processing tech? Their engine allows the artists to design shaders for specific objects but they end up with tens of thousands of them. I think that may be one of the reasons they drop the resolution so low during combat, it would be impossible to optimize that many different pixel shaders and it would be hard to weed out every one that causes a severe performance degradation.

On the Killzone 2 Slide above, one of the things that impresses me is how much white space is left over! Still lots to be done keeping all those SPUs busy.
 
This thread doesn't get much love :(
I guess that crytech and larrabee are stealing the show.
I would have loved to read some devs reactions especially in regard to the new improved 360 devkits.
I remember reading complain about it (ie lack of RAM).

I too would love to know what type of performance improvements might be expected from the new 360 SDK with 1GB RAM.
 
On the Killzone 2 Slide above, one of the things that impresses me is how much white space is left over! Still lots to be done keeping all those SPUs busy.

Both Insomniac and Guerilla Games mentioned that devs should not try to optimize too much. At this stage, there are enough (fast) cores in Cell to improve performance as long as the algorithm is scalable.

betan said:
Overall though, while they seem to be doing nice stuff, I cannot help but feel that Insomniac's shader approach to SPU programming in general (not specifically gfx aid) is more "interesting" than GG or even Naughty Dog (Ice Team).

Yes, the SPU Shader helps Insomniac to get around the limited local store, and code sharing + loading problems. It's a bigger win than I initially understood it to be.
 
...Why would they double the memory of the SDK kits if the 360 units themselves lack that kind of memory?
gamasutra said:
The company also, on the fully professional development side, has announced an entirely new version of its XDK development hardware, which contains two times the RAM of the currently released version. This additional RAM will be used for debugging and optimization tools separate to the game's assets and code, and will allow developers to manage their game development process more efficiently. The upcoming XDK software update which will ship alongside these units will continue to support existing XDK development units.
It's unclear to me if this will allow for better profiling tool or if it's just a matter of gaining time.
 
...Why would they double the memory of the SDK kits if the 360 units themselves lack that kind of memory?

Now developers can load all stuff from their game into the first 512MB memory without reserving some space for debugging/managing tools.
 
...Why would they double the memory of the SDK kits if the 360 units themselves lack that kind of memory?

Actually it's standard practice for devkits and has been like forever. I even doubt that this is actually new for the 360 devkit - unless it's 2GB rather than 1GB it's very likely to always have been that way. You want to be able to develop your game with debugging processes and tools running, and use more memory giving you time and flexibility to optimise other parts of your workflow before you've optimised memory use, etc.

As for the white spaces in the SPE usage graph, the Killzone devs have stated before in an interview that they aimed for 60% utilisation in order to ensure a stable framerate.
 
Now developers can load all stuff from their game into the first 512MB memory without reserving some space for debugging/managing tools.

Indeed... There was an IGN blog post by a BioWare developer mentioning how backwards it was to develop content for the end-user memory space because of the original developing configuration. I'm sure I've posted it before. It was a long time ago.

Actually it's standard practice for devkits and has been like forever. I even doubt that this is actually new for the 360 devkit - unless it's 2GB rather than 1GB it's very likely to always have been that way.

The 1GB may or may not be "recent", as in 2009/this month, but the original SDKs at least were still using 512MB. :eek:
 
Thanks for posting. The slides are easy to understand.

Good to know the moving level problem can be handled easily by the SPUs. I also like the idea that the SPUs and PPU can be toggled on the fly to run the same algorithm (for debugging purposes). Wonder if there's a time where they need all the 5 SPUs and the PPU to churn the same set of data in the same frame.

The presentation does not cover AI and physics (if present) specifically though. They are probably subsumed under the generic "Simulation" block.
 
Yeah, those slides also state, that they use the SPUs for procedural textures, which sounds interesting.

EDIT: I just ask myself, why they only leverage 5 SPUs... same for Killzone more or less, as they used it only for sound (afaik) and unzipping stuff, the rest of the time, it idles.
 
I suddenly wish the PS3 could display PDFs ... This is the second time I've wanted to read a pdf from the PS3 webbrowser.
 
Yeah, those slides also state, that they use the SPUs for procedural textures, which sounds interesting.

EDIT: I just ask myself, why they only leverage 5 SPUs... same for Killzone more or less, as they used it only for sound (afaik) and unzipping stuff, the rest of the time, it idles.

It looks like the devs have left quite a bit of processing power on the table too.

I'm guessing in GoW3, the 5th SPU is designated to handle all ad hoc requests that may disrupt workflow. It can probably chip in to help out important tasks from time to time; but don't count on its timeliness (e.g., handling fire and forget activities).
 
Yeah, those slides also state, that they use the SPUs for procedural textures, which sounds interesting.

EDIT: I just ask myself, why they only leverage 5 SPUs... same for Killzone more or less, as they used it only for sound (afaik) and unzipping stuff, the rest of the time, it idles.
Do you think they could be using the extra SPU almost solely for procedural textures?

I suddenly wish the PS3 could display PDFs ... This is the second time I've wanted to read a pdf from the PS3 webbrowser.
Same here.
 
Do you think they could be using the extra SPU almost solely for procedural textures?


Same here.

No...

what the PDF says is, that they use the SPUs for several things, including procedural textures (which not many games used before, iirc), and they also state several times, that they use 5 SPUs for doing their stuff, although they have 6 available for the game.

In the Killzone PDF, the chart showing the SPU usage, the 6th SPU basically was idling all the time, except for 2 minor things (sound and unzipping), which took less than 5% time in the chart. So they basically still have 1/6 plus the rest of the idle time of the other SPUs to leverage code. Not too bad for improvements actually.

Or do they not use the 6th SPU, because of some TRC set up by Sony (we did hear something a long time ago, that the OS can basically take away one SPU at short notice for OS stuff... maybe to circumvent any stalls, they just don't use it at all or so).
 
What do you guys think about this...

Christian Allen, an Ex-Bungie Dev, was nice enough to sit down and talk with our very own HipHopGamer at GDC ‘09 about the issues that matter the most to you, the gamer.

Q. After working on the 360 hardware can killzone 2 be done on that platform?

Oh yeah, Killzone 2 could totally be done on 360 if Guerilla wanted to. The tech is there, if the team and the publishers were motivated, they could do it, just like Halo could come to PS3 if MS didn’t own it. It really comes down to console exclusivity. The tech differences between the PS3 and the 360 aren’t all that much; it’s really about the delivery of content.

Q. Is Working on the PS3 as difficult as some may say?

From my experience tangentially with PS3, it was more difficult than 360 in the past, but from what I have heard this has greatly improved recently. My experience has definitely pointed me to the 360, as that is what I’ve worked on and that is what I know; but I wouldn’t shy away from a PS3 game now.

http://www.hiphopgamershow.com/?p=6299
 
Posted on April 1st and Christian Allen isn't an ex-Bungie developer, as he is still at Bungie heading up a super secret project. Going with that being an April Fools.

Edit: Nevermind, after watching the video interview, I guess he does no longer work at Bungie.
 
Last edited by a moderator:
Back
Top