Quake4: Nvidia6800 vs ATI Xenon

Assuming this isn't an API issue. Is there anything fundamentally different about Xenos that would require retooling? The USA will do automatic load balancing to keep shader utilization high, no?

Then what about the edram does that require special coding to get 100 efficiency?
I understand that ideally a 100% rewrite is the best solution, but still I'd like to get a feel for how well bare-bones PC->Xbox ports will perform and how much effort is needed to get them up to speed.
 
I'm tellin you guys we're finally seeing what these console CPUs are doing to developers. These CPUs are not easy to get going fast.
 
ihamoitc2005 said:
but what does CPU utilization have to do with reducing graphics assets in quake4?

Because shadow calculation is done on the CPU, not in the GPU, and if they can't figure out how to make decent code for the CPU, they have to make it do less work -> downgrade models.
 
Shadow calculation

Apoc said:
Because shadow calculation is done on the CPU, not in the GPU, and if they can't figure out how to make decent code for the CPU, they have to make it do less work -> downgrade models.

But like collision model, shadow calculation can use a simpler model than drawn character model, in fact, usually is.
 
polygon model

TrungGap said:
I guess you haven't played doom3.


Well Doom3 drawn model is not so high polygon count is it? Even then there is no guarantee same as shadow model. Most games is not the same.

Point is simply that CPU shadow calculation does not have to be tied to drawn model polygon count, just as collision model does not have to be same. Therefore, CPU performance no impact on drawn polygon model, only on shadow and collision model.

Therefore question is why drawn model polygon count reduced? Could it be that X2 port using CPU for skinning as well? Regarding GPU, what can be cause for such low GPU performance? Can OpenGL incompatibility explain 25% GPU performance? I do not have answer to these so thats why I ask.
 
Well the point is that they stated they planned to have the game running the exact same content, while also running at least 30 fps by the time the game comes out. Early on in the dev cycle for a game, the game does not run as fast as it will towards the end. In fact, the final part of making the game (unless I'm mistaken) is optimizing it for the hardware. I'd say they're most likely either working on that now, or going to be working on it soon. Quake 4 for the PC has been in development much longer and the engine was already running at a decent speed on PC with Doom 3, therefore it should take much less time to get the engine running at a good speed on the PC with Quake 4 assets. With the X360, they first have to get the engine up and running with a different graphics API on an in-order CPU with 3 cores. It's going to take some time to get the code running at higher speeds. So like I said, considering the port is in earlier stages than the PC counterpart, combined with different hardware and a different API, and also considering that ports don't exactly push the hardware, the fact that the game is only running around 30 fps and using simpler models (for now) is not a big deal and is definately subject to change.
 
ihamoitc2005 said:
Well Doom3 drawn model is not so high polygon count is it? Even then there is no guarantee same as shadow model. Most games is not the same.

Point is simply that CPU shadow calculation does not have to be tied to drawn model polygon count, just as collision model does not have to be same. Therefore, CPU performance no impact on drawn polygon model, only on shadow and collision model.

Therefore question is why drawn model polygon count reduced? Could it be that X2 port using CPU for skinning as well? Regarding GPU, what can be cause for such low GPU performance? Can OpenGL incompatibility explain 25% GPU performance? I do not have answer to these so thats why I ask.

No. Carmack made the point (many times) that he wanted the shadow to be the same detail as the model's edges. That's why the models were so low-polys, because if they were more "rounded", then the shadows would have to be more complex, which would spiral down the performance.
Personally i think there is more than enough juice to have very good shadows on next gen machines, and Carmack's approach is just wrong for these architectures.
 
GPU performance

Mordecaii said:
It's going to take some time to get the code running at higher speeds. So like I said, considering the port is in earlier stages than the PC counterpart, combined with different hardware and a different API, and also considering that ports don't exactly push the hardware, the fact that the game is only running around 30 fps and using simpler models (for now) is not a big deal and is definately subject to change.

But if Xenon code can only get 50% performance of x86 code then 30fps vs 60fps makes sense. But that does not explain GPU performance and art degrading. With 30fps, GPU should have so much extra time that not only should art be not degraded, it should be greatly improved no? So question remains why they only use 25% expected GPU performance.
 
One question I do have... from what I understand, you can modify OpenGL to add new "techniques" and have them in your game, whereas with DirectX it's controlled by Microsoft... Is it possible that there are some features with the engine that aren't translating as well as they would have hoped to DirectX and therefore they are having to spend extra time working on that?

Edit: ihamoitc2005- You ignored the part of my post stating that game performance was basically guaranteed to improve as they continue to work on the game so why are you so hung up on the performance of the game at this moment in time? There isn't a magic button you can press to port code from PC to the X360 and have it run perfectly, especially when the game started off as OpenGL on the PC.
 
Last edited by a moderator:
Gpu

Mordecaii said:
Edit: ihamoitc2005- You ignored the part of my post stating that game performance was basically guaranteed to improve as they continue to work on the game so why are you so hung up on the performance of the game at this moment in time? There isn't a magic button you can press to port code from PC to the X360 and have it run perfectly, especially when the game started off as OpenGL on the PC.

But they say they are targeting 30fps with no improvment in art and in fact some decrease. I never said portin is easy and as i posted earlier both ATI and Microsoft are anti opengl, but targetting 25% expected performance is very poor no?
 
25%

london-boy said:
25% of what?

Of expected performance. To simplify for ease of communication, if Xenos = 2 x 6800, then perfect port should be able to do quake 4 at 120fps with PC version graphics content or 60fps with 2x better graphics content or 30fps with 4x better graphics content. If CPU and not GPU causes frame-rate to be 30fps, then why graphics content not just not 4x better but actually slightly worse.
 
Trolling??

Apoc said:
Is anyone more thinking that ihamoitc2005 is trolling this thread?

What is meant by trolling? I started this thread and I am looking for answer to questions that explain GPU performance.
 
ihamoitc2005 said:
What is meant by trolling? I started this thread and I am looking for answer to questions that explain GPU performance.

After 4 pages did you notice anything? Your questions were answered as best we could given the information we have. You keep going on and on about it.
When we have more info, it will be posted here.
 
ihamoitc2005 said:
What is meant by trolling? I started this thread and I am looking for answer to questions that explain GPU performance.

It means they don't like what you are saying and want you to stop.

Keep in mind that this board is essentially a Microsoft/Windows x86 pc graphics site and the vast majority of people here come from that world. You aren't going to get a serious discussion of performance problems with the 360 hardware here.
 
i think he is Ken Kutaragi... the mother of all playstation ******s of the world

HEIL SONY! :D
 
Back
Top