Back from the Doom 3 demo at QuakeCon

[Matthew Gallant]

Holy bejeebus.

Any questions?

[Supertanker]

Is it more like Resident Evil, with a handful of slow monsters at a time, or is it more like Doom, with a horde of fast movers?

I know it is pretty, I'm concerned that id has forgotten to include a game with their engine again.

[xahlt]

How was the food, good?

[Matthew Gallant]

The pinkies move really fast, or rather they set themselves and jump all the way across the room at you. Just like in the original. It looks like there'll be pretty intense moments.

They also gave the first public showing of the Revenant, and he has dual shoulder-mounted rocket launchers.

It's not Quake 3 speed, but it's not slow, that's for sure.

[Jason Becker]

Also from QuakeCon

Looks like Carmack announced DOOM III is coming to the XBox.

"Id Software had previously expressed interest in doing an Xbox version of its upcoming first-person shooter, but hadn't confirmed that there would be a console version of the game. Today during his keynote at QuakeCon, id's John Carmack said that the Xbox is the only console platform that id is "completely committed" to releasing Doom III on. Carmack further commented that the Xbox version will have the "full graphics fidelity" of the PC version, which made a major debut at this year's E3, where it won a number of awards."

[Matthew Gallant]

Yep, X-Box, but the editor will be stripped out.

[wumpus]

Wow, that's big news. 64mb is going to make it tough, but I assume with this engine they can chunk up the memory footprint at the expense of more loading segments.

[Brad Grenz]

the Xbox version will have the "full graphics fidelity" of the PC version

I think we can assume this to mean full graphic fidelity of the PC version running on Xbox class hardware. (GF3, PIII 733, low memory and therefore low resolution)

[Matthew Gallant]

I think we can assume this to mean full graphic fidelity of the PC version running on Xbox class hardware. (GF3, PIII 733, low memory and therefore low resolution)

I don't know, but the game is designed around the feature set of a GeForce 1, as he explained in the keynote. I'd say it should run just fine at TV resolutions on X-Box hardware.

[Brad Grenz]

But the game running on a GF1 isn't going to look nearly as good as the game running on post Radeon 9700/NV30 upon release. Which is what "full graphic fidelity" implies.

[Matthew Gallant]

We'll see.

[Ben Sones]

Wow, that's big news. 64mb is going to make it tough, but I assume with this engine they can chunk up the memory footprint at the expense of more loading segments.

They can also reduce texture sizes considerably. I mean, a TV screen has a pretty low screen resolution, so after a certain point, adding more texel resolution to models makes little noticeable difference.

[Jason Becker]

"They can also reduce texture sizes considerably."

Isn't that what S3TC/DXTC is for. DO any devs bother with that stuff or do they not care since memmory on video cards keeps doubling every 1 year now it seems.


"I think we can assume this to mean full graphic fidelity of the PC version running on Xbox class hardware. (GF3, PIII 733, low memory and therefore low resolution)"


Consoles have shown before they can run ports and/or similar games that would require a more powerful PC. The Dreamcast(200mhz cpu, 16MB ram, 8MB video ram) ran ports of UT and Q3A descently. Maybe with some lower detail levels here and there but from my experience in trying UT on a DC it did a pretty good job. A similar level PC couldn't have done that.

[Bub, Andrew]

Aside from the PC, of course, is it an Xbox exclusive?

[Wholly Schmidt]

Bub, from the bits I read at Gamespot it wasn't worded as an "exclusive" in the sense that there were agreements stopping it from being ported elsewhere, it just sounded like Xbox was the only console they actually had plans for at this point. Not sure though.

[Anonymous]

Isn't that what S3TC/DXTC is for. DO any devs bother with that stuff or do they not care since memmory on video cards keeps doubling every 1 year now it seems.

Yes they do, because even though the high-end cards do double in RAM ever year or so, the developers still need to program for a lower common denominator.

Notice that in a lot of games that there is a video option for "compressed textures". In a lot of cases, the compressed textures will also look better than the regular texture, because they were able to squeeze more detail into the compressed texture and not worry about it, because the compressed texture still took up far less memory than the non-compressed texture.

Texture compression works and it's a useful tool. It's no gimmick.

[Jason Cross]

Isn't that what S3TC/DXTC is for. DO any devs bother with that stuff or do they not care since memmory on video cards keeps doubling every 1 year now it seems.


Virtually all 3D games on the PC use texture compression now, by default. In fact, turning it off in some games (Jedi Knight 2, SoF2) will make some levels have a good 250MB or more of textures at high detail settings, which will thrash on any card.

Sure, 128MB is common, but lately game textures have become a good 2-4 times as detailed and at the same time level designers are using more uique textures so everything doesn't look repeated.

I think the Xbox will have no problem doing Doom3 that looks about as good as the screenshots released, except at TV resolution. As was pointed out here, they'll use exclusively compressed textures and probably no texture larger than 128x128, because anything more detailed than that won't look any better up-close on your TV.

And knowing exactly what the hardware is, they can make some special-case optimizations for Xbox that they can't necessarily do on the PC (like optimize their physics system for a processor with SSE).

[Mike Cathcart]

I think the Xbox will have no problem doing Doom3 that looks about as good as the screenshots released, except at TV resolution. As was pointed out here, they'll use exclusively compressed textures and probably no texture larger than 128x128, because anything more detailed than that won't look any better up-close on your TV.


Except that Microsoft requires progressive scan support in all XBox games, right? I wonder how it will look on those HDTVs.

[xahlt]

Alright, let me ask a question like this then:

If I can get the full graphics fidelity of DOOM 3 (which has alternately been described as "mind blowing", "OMG", and "you're gonna need a new graphics card for this") on a box that is essentially a P3 700, gf3, like Brad says, have other engines for games just been ludicrously underusing the tech? I mean, I'm looking at NOLF2 or Unreal 2 or 2k3, and granted, they are getting released sooner than Doom, but they look like 2 generations behind. Now I assumed it was because the new Doom engine was:

1. coded by probably the best graphics engine people in the business
2. was using next gen hardware that resulted in playable framerates

So, is 2 invalid then, and 1 is many time more true than I would have thought?

I'm just confused how they are doing it. I was sure the engine would be remarkably scaleable (as both the Quake III and Unreal engines were) so the game is always playable, but this full graphic fidelity thing is throwing me for a loop.

[Matthew Gallant]

I'm just confused how they are doing it. I was sure the engine would be remarkably scaleable (as both the Quake III and Unreal engines were) so the game is always playable, but this full graphic fidelity thing is throwing me for a loop.

Well, the X-Box does have a GeForce 3. And anyway, John said at yesterday's speech that the performance bottleneck for Doom 3 is not processor speed or fill rate, but the video memory bandwidth. Which makes sense, because he said that some pixels in a Doom 3 scene are going to have over 20 textures applied to them. Not even the latest cards can do that in one go.

Plus the framerate target is a meager 24 fps.

[Ben Sones]

1. coded by probably the best graphics engine people in the business
2. was using next gen hardware that resulted in playable framerates

So, is 2 invalid then, and 1 is many time more true than I would have thought?

1 is certainly true, and 2 is still valid. The Xbox's PIII-700/GeForce 3 configuration doesn't sound that impressive, but developers can target that as their [i]minimum[/] system requirement. Also the maximum (for the Xbox version), since the hardware is fixed, but that is actually more of a benefit than a drawback because it lets them optimize for a single hardware configuration. I'm guessing DOOM III will look equally good on a PC with the same hardware... at 640x480 and with the texture detail turned all the way down.

[Brad Grenz]

but developers can target that as their [i]minimum[/] system requirement.

I read Carmack was saying a 1 GHz CPU was the minimum spec target for the PC version of Doom III. Sure they can optimize and cut up levels, but the point is saying the Xbox version has "full graphic fidelity with the PC version" implies it will be comperable visually to the game running on a high end PC. This is of course absurd. The Xbox version will be a low res, lower poly, lower texture approximation. It will use most of the same effects, lighting, etc, but it's going to look much better on a tricked out PC.

[Case]

Someone earlier suggested the Xbox GPU is a "GeForce3". Actually, it's more comparable to a GeForce4, with dual vertex shader units. GeForce4 is an updated version fo the Nv2a.

Also, given the thin API layer and the shared memory architecture, the Xbox performs more like a 1GHz PIII PC with a GeForce4.

As ever,

Loyd Case

[Brad Grenz]

Also, given the thin API layer and the shared memory architecture, the Xbox performs more like a 1GHz PIII PC with a GeForce4.

Yes, I hear having significantly less memory bandwidth to go around makes the Xbox faster. I think I read that at xbox.com in a very professional looking html table.

[wumpus]

Debatable, since as someone pointed out to me on these boards a while ago, the "P3" inside the Xbox only has 128kb of L2 cache instead of the 256kb of a real P3. It's more like a Celeron.

I'd still tentatively argue that the Xbox should perform somewhat better than a "real" PC with 64mb, a P3-733 and a GeForce 4 ti 4200, mostly because of the faster memory interface (200mhz DDR) and the aforementioned thin APIs. Hopefully the console guys are writing more to the metal anyway.

[wumpus]

Yes, I hear having significantly less memory bandwidth to go around makes the Xbox faster. I think I read that at xbox.com in a very professional looking html table.
You shouldn't debate hardware with Loyd Case. But if you insist, point me to any Pentium 3 PC with 200mhz DDR memory.

[Brad Grenz]

You shouldn't debate hardware with Loyd Case. But if you insist, point me to any Pentium 3 PC with 200mhz DDR memory.

It's not just the system memory it's also the video ram. Besides, the PIII's front side bus limits it's ability to access system memory to something like 1.06 GBps. Even an SDRAM based PIII (1.06 GBps) with a GF4 4600 (10.4 GBps) has nearly twice the combined video/system memory bandwidth of an Xbox. Oh, and that PIII would have 325 mhz DDR in the box. Sure, I'll concede an Xbox running an Xbox game would be effectively faster than the same level hardware in a PC that had to load a full copy of Win XP first, but a slim OS/API can't manufacture capabilities.

[Case]

The PC architecture is hobbled by the AGP port, which maxes out at roughly 1GB/sec. The Xbox doesn't have that limitation. The built-in memory controller in the NV2A chip acts more like an SGI shared memory system than a standard PC, so the memory utilization is much more efficient.

Also, more of the workload *can* be offloaded to the GPU, though in practice, it's hard to say how it's actually going to be parsed. I believe, however, that the Xbox system will be good enough to run Doom3 at NTSC resolutions with adequate speed.

What's more interesting to me is what graphics API is Carmack going to use. The Xbox doesn't have OpenGL built-in, so it will have to be brought along somehow -- or the renderer will need to understand DirectX Graphics.

As ever,

Loyd Case

Oh, and that PIII would have 325 mhz DDR in the box. Sure, I'll concede an Xbox running an Xbox game would be effectively faster than the same level hardware in a PC that had to load a full copy of Win XP first, but a slim OS/API can't manufacture capabilities.

[Brad Grenz]

The PC architecture is hobbled by the AGP port, which maxes out at roughly 1GB/sec. The Xbox doesn't have that limitation. The built-in memory controller in the NV2A chip acts more like an SGI shared memory system than a standard PC, so the memory utilization is much more efficient.

I remember reading that the integrated graphics of the first nForce ran at approximately a 6X AGP port speed. I've never been real clear on how that works as you'd think having the graphics core on the north bridge would cut AGP out of the equasion alltogether. But my current impression is that there, for legacy reasons or whatever, there must be some sort of AGP tunnel in the nForce (and by extension, the NV2X). If we're talking about something around 6X AGP speed it's still only like a quarter of the available memory bandwidth in an Xbox. I don't deny that Xbox has the advantage of not needing to upload textures across the AGP bus, but pointing that fact out is something usually done, in my experience, to deflect questions about the Xbox's hamstrung memory bandwidth. It's nice to just be able to change a pointer instead of copying a texture from system to video memory. But that's kinda besides the point, a seperate issue. When you're talking about memory bandwidth for a GPU you're not so much concerned with how fast a texture can be moved to local memory, but how fast that texture can be read in order to render a scene. I've actually debated with people who had no clue that textures needed to be read by the GPU in order to be applied to polygons in a scene. And this is where bandwidth gets gobbled up. It's also where the Xbox's shared memory comes up short compared to a dedicated solution like the GF4 or an embedded, dedicated solution like the PS2's Graphics Synthesiser and Gamecube Flipper.

Sure, the Xbox will be able to run a version of Doom III running at 640x480 at 24 frames/sec. It will probably even be one of the better looking Xbox games, but the PC version will be using high poly models, more complicated maps, larger textures and will run at higher resolutions. I'm just really annoyed by the phrase "full graphic fidelity". Talk about your carefully crafted PR-speak. It implies the Xbox version will look every bit as good as the PC version, but avoids having any meaning whatsoever. It had the desired effect, every Xbox fan site out there is trumpeting the fact that the Xbox version will look as good as the PC version. The quickstep they're doing about whether Doom III will be console exclusive to the Xbox is grating as well. They've never said it's Xbox only, they say they're commited to Xbox right now, that the Xbox is the only one they're certain will be capable of handling a port. Xbox fan sites get to say the Xbox is the only system powerful enough for Doom III while id and Activision leave the possibilities open for PS2 or Gamecube ports.

[Jason Becker]

"or an embedded, dedicated solution like the PS2's Graphics Synthesiser and Gamecube Flipper. "


The PS2 also only has 4MB of video memmorry which severley hamstrings it too. From things I've read devs would much rather have the Xbox's unified memorry over the PS2's setup.


"It implies the Xbox version will look every bit as good as the PC version,"

And what percentage of gamers have the super tricked out system that you are talking about? In reality seeing the systems that most people have(even people who are into games), it will probably look pretty close excpet for the resolution diffrence for most people.

[Case]

It's interesting to note that the Xbox memory controller is nothing like the Nforce. In some conversations with some of the NV2A architects, they were delighted that "... NV2A didn't have any of the PC baggage". It's really a lightweight memory controller, and it runs at 233MHz. The 64MB of RAM behaves more like local frame buffer. In some ways, the PIII is more the coprocessor here.

However, the 733MHz PIII may be somewhat limiting in other areas. For example, the Doom3 engine is capable of per-polygon collision detection -- which is handled by the CPU. I don't think the vertex engine in the NV2A is flexible enough to offload this. If the triangle cound is high, then whoever handles the Xbox port may have to work around this.

Finally, there's the question of "visual fidelity". Given that Doom3 will have multiple rendering backends, it's possible that the Xbox will have a backend different from the GF4, for example. Already there's been discussion that those with GeForce2's will see much lower triangle counts. So "full visual fidelity" is an incomplete statement. It's more like "full visual fidelity like a PC running a Geforce???".

As ever,

Loyd Case

[Erik Andersson]

Some quotes from John Carmack (from slashdot):


The X-Box GPU is more of a GF4 than a GF3, but a modern PC is generally much higher end than an X-Box.

However, you can usually count on getting twice the performance out of an absolutely fixed platform if you put a little work into it. There are lots of tradeoffs that need to balance between the different cards on a general purpose platform -- things that I don't do with vertex programs because it would make the older cards even slower, avoiding special casing that would be too difficult to test across all platforms (and driver revs), and double buffering of vertex data to abstract across VAR and vertex objects, for instance. We might cut the "core tick" of Doom from 60hz to 30hz on X-Box if we need the extra performance, because it has no chance of holding 60hz, but the PC version will eventually scale to that with the faster CPUs and graphics cards.



The generic back end does not use vertex programs, or provide specular highlights, so the custom back ends provide both performance and quality improvements.

There are some borderline cases that may or may not get custom coding -- Radeon R100, Matrox Parhelia, and SiS Xabre are all currently using the default path, but could benefit from additional custom coding. I will only consider that when they have absolutely rock solid quality on the default path, and if it looks like they have enough performance headroom to bother with the specular passes.

The NV20 back end has more work in it than any other, with two different cases for the lighting interaction, but on the X-Box I would probably create additional special cases to optimize some of the other possible permutations.

[Jason Becker]

That second quote seems like the classic case of developing for the PC and the headaches it can can involve, with all the diffrent card performace/feature levels.

That last line in the second quote sounds like he could be involved with any port to the Xbox, or that he's already been messing with it some.

[Brad Grenz]

The PS2 also only has 4MB of video memmorry which severley hamstrings it too. From things I've read devs would much rather have the Xbox's unified memorry over the PS2's setup.

I think most would actually just rather have more VRAM in the PS2.

It's interesting to note that the Xbox memory controller is nothing like the Nforce. In some conversations with some of the NV2A architects, they were delighted that "... NV2A didn't have any of the PC baggage". It's really a lightweight memory controller, and it runs at 233MHz. The 64MB of RAM behaves more like local frame buffer.

I would expect the Xbox memory controller is like the crossbar memory controller in a GF4. But then I expect the nForce memory controller is based on nVidia's GPU memory controllers. They can't have hacked out too much since it still needs to manage external memory requests from the PIII, MCP, etc. I'll buy that they cut out lots of legacy stuff like PCI and AGP, though. I wasn't sure about it.

However, the 733MHz PIII may be somewhat limiting in other areas. For example, the Doom3 engine is capable of per-polygon collision detection -- which is handled by the CPU. I don't think the vertex engine in the NV2A is flexible enough to offload this.

That's another thing people are frequently unable to grasp. And it's an area where the PS2's design philosophy has an advantage. The "A PIII (read:celeron) 733 should be enough" line has always stuck in my craw. There's lots of really intensive computations you might want to do in a game that a vertex shader can never do. The emotion engine was engineered to be incredibly flexible in this regard. Sony traded away pure triangle numbers in favor of versatility. Vertex shaders are more flexible than a hard coded DX 7 T&L unit, but you still can't use their juice to run physics simulations, AI routines or encode a DTS bitstream in realtime.

Already there's been discussion that those with GeForce2's will see much lower triangle counts. So "full visual fidelity" is an incomplete statement. It's more like "full visual fidelity like a PC running a Geforce???".

And that was my original point. It's PR fluff full of implied meaning but devoid of actual information.

That last line in the second quote sounds like he could be involved with any port to the Xbox, or that he's already been messing with it some.

He's probably dabbling, but historically id dishes off ports. It sounded like he was eager to move on to his next engine generation with DX9 as a starting point.

[Jason Becker]

"I think most would actually just rather have more VRAM in the PS2. "


err so? I'm sure they would like 256MB ram with 500GB/sec of bandwidth if they could. Thats not reality though. I'm talking about what is availible today.

"The "A PIII (read:celeron) 733 should be enough"

The XBox CPU isn't a Celeron there's more diffrence than the 128K of cache.

[Brad Grenz]

err so? I'm sure they would like 256MB ram with 500GB/sec of bandwidth if they could. Thats not reality though. I'm talking about what is availible today.

But they also tend to like the 30 million system installed base. Everything's a compromise, and who's to argue if MS offers to fully fund your development?

The XBox CPU isn't a Celeron there's more difference than the 128K of cache.

Actually, the only other difference is that the L2 is 8 or 16-wat associative as opposed to 2 or 4 in the PIII derived celeron. I think the Xbox CPU is technically a mobile PIII which has which has the lower L2 but with the same association.

[Jason Becker]

"But they also tend to like the 30 million system installed base. Everything's a compromise, and who's to argue if MS offers to fully fund your development?"

I don't have a clue as to what your saying. Funding development? The point was about the limitations that the PS2 has itself, and you go off into hypotheticals and I don't know what...

[Case]

There is one other difference -- the FSB clock is 133MHz, as it is with the desktop PIII.

As ever,

Loyd Case


Actually, the only other difference is that the L2 is 8 or 16-wat associative as opposed to 2 or 4 in the PIII derived celeron. I think the Xbox CPU is technically a mobile PIII which has which has the lower L2 but with the same association.