Found an Xbox kiosk today... (1 Viewer)

[Gary King]

(cough) Deus Ex.
(cough) System Shock 2.
(cough) Super Mario 64.
(cough) Project Ego.
(cough) Star Wars Galaxies.

 

[AndyVX]

I call it the BS2 because most of the stats were nothing but Bullshit. No Anti Alaising, 75 million polys per second is a joke. It can't do it.

The thing is though, that it can do all these things. AA is possible through software coding, and the 75mpps is also true. Those numbers are the raw processing power of the system, so technically Sony is not lying when giving out these numbers.
I think the best thing to do is let the games be the judge, and not the launch games. Game's that come out 1-2 years afterwords, because that's when developers will have a real handle on the system(s) capabilities(Xbox/Gamecube), and it's taken a little longer for devs to get used to PS2 programming since it's a lot tougher (or so the story goes)
Andrew
------------------
ICQ# - 119869807

 

[Morgan Jolley]

Andy-
Even as tough as it is to program for the PS2, I'm surprised that games like MGS2 and ZOE and The Bouncer look as good as they do. MGS2 had the demo out in May, which was astounding! If they make the games look better because of the learning curve, then boy are we in for a treat!
Gary-
Those games you listed were innovated and creative, but I was talking about visual artistry. ICO looks better than any game I have ever seen in my life because of the artistry behind it. The image maps that were used were high-res and beautiful, the way that colors lit up or blackened the screen was amazing, and the look of the "bad guy" of the game was excellent. Its this type of visual imagery that made the masterpiece that ICO is such a masterpiece.
I was talking about how the developers can make a game look better than what it can be simply through artistry.

 

[Brad Grenz]

So it's more about art direction that art assets. I think I'd agree with that statement.
I think I'd also agree that the Xbox graphics won't ramp up the way the PS2's will. The GPU may be a souped up GF3 in some aspects (the alledged 2nd vertex shading unit), but it's souped down in some ways too. *cough* memory bandwidth *wheeze* And if you ask John Carmack, he'll tell you the Vertex/Pixel shaders aren't as flexible as nVidia would have you believe. The Xbox is already limited by its memory bandwidth. Whether you're talking geometry or texturing or sound processing, there's a fairly low ceiling to work under. In contrast, the PS2 has VRAM with 48 GBps of bandwidth just waiting to be exploited, and a 128 bit CPU with 3 coprocessors waiting to be utilized. There aren't many secrets to be discovered in the Pentium III and you can't do much nVidia didn't anticipate with the vertex/pixel shaders. The Xbox just won't be ramping the way the PS2 can.
If you want to talk about unrealistic performance quotes, MS was claiming 300 million polys per sec., which was bullshit, and 4.8 Gigapixels per sec. which was an outright lie. You run a synthetic benchmark on a Geforce 3 and it'll tell you the GPU can process 4-5 million polys per sec. Multiply that by 2 vertex units in the Xbox GPU and you're looking at 10 million. That's no more then the Gamecube and PS2 are capable of. That number will surely rise which Xbox specific game engines, but it will never exactly destroy the competition's numbers. And practically speaking there won't be a very appreciable visual gape between the Xbox and the other 2 systems when viewed on a television.
------------------
Brad Grenz
Link Removed
the Widgets

 

[Gary King]

There is quite a bit that can be done with the vertex/pixel shaders that wasn't anticipated.
A few weeks ago I hacked out a simplified Rayleigh scattering simulation using the vertex shaders for some particle system stuff I was writing -- that certainly wasn't covered in NVSDK. The vertex shaders are very, very well designed.
There's also a tremendous amount of flexibility in the register combiners, which reduces bandwidth requirements (since using all 9 combiner stages leaves maximum fill at 1/4). The pixel shaders can also be used for quite a bit of effects, too, but you have to be creative (the primary effect is bumpy-shiny).
The memory subsystem in the NV2A is also improved above and beyond what is in the GeForce 3, so it makes better use of its bandwidth than any NV20-generation product (which was roughly 75-80% efficient). Add to all this the fact that rendering is limited to 1024x768 resolution, and there's plenty of bandwidth and graphical flair to go around.
The NV2A rejects pixels at 4GPixels/sec, and when rendering in multisampled mode it can generate up to 4GSamples/sec. The 4GPixel/sec (scaled for clock speed) wasn't really a lie at all, especially since the numbers provided to developers were listed in GSamples/sec, rather than GPixels/sec.
And I get ~24M triangles on my synthetic benchmarks for the GeForce 3. I don't think performance for synthetic Xbox benchmarks has been released yet, but you can rest assured that it is much more than your 10M estimate.

In contrast, the PS2 has VRAM with 48 GBps of bandwidth just waiting to be exploited

I hate doing numerical analyses, but comments like this really deserve some more information...
The 48GBps is only attainable when all 16 pixel units draw a pixel each of the 150M clocks. It's fairly same to assume that over the course of a frame, roughly 40% of the area 'rendered' by the 16 pixel units doesn't correspond to any actual geometry. So of the 48GBps originally available, you're already down to 28.8GBps.
For texturing, each pixel unit reads 128-bits of texture data every 2 clocks. Assuming that only 16 unique texels are needed for every 8x2/2x8 render chunk, this works out to about 10GB/sec of texture data. At 6:1 compression, this would have been 1.666GB, so the PS2 loses another 8.33GB for not having any texture compression, so the theoretical maximum realistic bandwidth is just 20GB/sec.
So even before we consider effects like trilinear filtering, anti-aliasing, or multi-texturing, or the costs of texture uploadsto the eDRAM, etc., you're looking at less than half of the initial 48GB/sec. As triangle size shrinks, this ratio becomes even worse.
If you're going to complain that Microsoft's raw specs were unrealistic, you need to be more objective about other hardware, too. 48GB/sec is nice; however, when it's used as inefficiently as it is in the PS2, you can't really take the number at face value.

 

[Carl Johnson]

Whoa, it feels like I need to grow some brain cells in order to decipher those last couple of posts.

 

[Morgan Jolley]

Wow...I'm REALLY lost.

 

[James D S]

Damn Gary, show a little mercy! Humans want to read this stuff, too.

 

[DaveF]

Forget the geek-speak, I'm still trying to figure out what "ICO" is. "Intergalactic Community Outhouses"?
I don't know why people "justify" their console choice with tech stats -- and then complain about other systems having exaggerated specs. There are two facts about gaming:
- Vendors always exaggerate performance claims.
- Gameplay matters more than graphics
People who want to play games, IMO, don't really care if the hardware of GameBox2 is slightly better/worse than the PXCube -- they just want fun games.

 

[Gary King]

Well, I may have gone overboard...
But next-generation graphics hardware is what I work on day-in/day-out, so it *is* rather important to me.
Summarizing my analysis of the PS2:
It's got a whole lot of raw power; however, the way that power is configured really isn't well-designed for the way it will be used. Bandwidth is important, but the way Sony achieved the PS2's bandwidth was through a couple of assumptions that don't really hold true for 3D graphics.
On the other hand, while certain critical specs for the Xbox (e.g., memory bandwidth) may seem ridiculously low, due to the way memory is needed in graphics systems, it's not as bandwidth-limited as a simple glance would suggest.
quote: Originally posted by Morgan Jolley:
Wow...I'm REALLY lost.[/quote]
Welcome to the world of high-end graphics architecture analysis . I even kept this rather high-level... Hopefully you'll think twice before you try to compare console architectures in the future -- it's a whole lot more than just looking at the numbers and applying some simple one-size-fits-all equation to them.
[Edited last by Gary King on November 07, 2001 at 07:10 PM]