Xbox 360 running Ruby at Siggraph, Pictures

hasanahmad said:
besides, do you actually think the RSX is more powerful than the R600 slated for Q4 2006?

Well if you take the rumors about the R600, no. Please take this as a grain of salt, to the max, lol. The rumored specs for R600 is this:

65nm
64 Shader pipelines (Vec4+Scalar)
32 TMU's
32 ROPs
128 Shader Operations per Cycle
800MHz Core
102.4 billion shader ops/sec
512GFLOPs for the shaders
2 Billion triangles/sec
25.6 Gpixels/Gtexels/sec
256-bit 512MB 1.8GHz GDDR4 Memory
57.6 GB/sec Bandwidth (at 1.8GHz)
WGF2.0 Unified Shader

http://www.cdrinfo.com/Sections/News/Details.aspx?NewsId=14184

Again, take this as a grain of salt.
 
Thegameman said:
I find a gamespot link.


The 1.5 million is not for the character model the 1.5 million as i see it is just for the face,which would make it very impressive,even at the end this level of detail will not be show,cuz you know how demos are showing unplayable angles,and stuff that you will not see at the end.


After Bandai, Koei showed a real-time demo of its PS3 game, Ni-Oh, which featured a Dynasty Warriors-esque character fighting multiple enemies. Pausing the game and changing the camera angles, Koei chairman Keiko Erikawa zoomed in on the character's face to show its detail, which even showed the pores on the skin. Flanked by an assistant wielding a PS2 controller, Erikawa explained that with the PS2, developers were able to allot about 1,000 polygons to a character's face. With the PS3, Erikawa's team was able to allot up to 1.5 million polygons. Erikawa explained that additional polygons allow for more subdivision surfaces, so faces can have more wrinkles and personality.


http://hardware.gamespot.com/Story-ST-15015-2308-4-4-x

That link and quote is obviously a typo or a flat out LIE.

I'm so sure that we will be seeing character models with multiple MILLIONS of polys / ( 1.5 million, just for the face....right ) in PS3 games. :rolleyes:
 
The jump from SM2 to SM3 (R420 to RSX/R520) is quite minor compared to the jump from SM3 to SM4 (R600). Xenos is beyond SM3, but we don't know yet how close to SM4.

It's the exact opposite IMHO.

SM2.0 = 96 instruction slots
SM2.0_extended = 96 - 512 instruction slots
SM3.0 = 512 - 32768 instruction slots

Both Xenos and NV40 claim (on paper at least) unlimited resources with a bit more than 4k instruction slots (4096 PS30 / 544 VS30 for NV40, while on Xenos those are obviously just shader instruction slots for both).

Considering SM3.0 requires dynamic loops, branching etc. I don't see why it isn't closer to "SM4.0", than it is to anything SM2.0.

Xenos is missing the geometry shader of WGF2.0. Even if it should meet all other requirements, it strips it of WGF2.0 compliance.

Frankly I don't see why all this is so important anyway. Is there any reason why either Microsoft/ATI, Sony/NVIDIA would have to worry about WGF2.0 compliance on a console?

I'll probably get slaughtered for even mentioning it (in this forum *cough*), but if I am to trust the so far developer impressions on the CPUs of the two upcoming consoles, I'd rather expect developers to be extremely careful with what overall content they'll throw at either/or. Or else since IMHO the GPUs in both seem to be the "saving grace", I worry first about what those have to "save" before coming to any comparisons with anything considering the PC, at least until games move out of the single-threading realm.
 
BenQ said:
That link and quote is obviously a typo or a flat out LIE.

I'm so sure that we will be seeing character models with multiple MILLIONS of polys / ( 1.5 million, just for the face....right ) in PS3 games. :rolleyes:


I do say that already that i don't think that in the end games will show a face with 1.5 million polygons,maybe on an FMV after you past a level,or a close up after you win a match or something like that..
 
BenQ said:
That link and quote is obviously a typo or a flat out LIE.

I'm so sure that we will be seeing character models with multiple MILLIONS of polys / ( 1.5 million, just for the face....right ) in PS3 games. :rolleyes:


No, actually, it's quite true, in a manner of speaking.

The 1.5 million poly figure is the number of polygons used to create a normal map of the face, so there really is 1.5 million polygons making up the face, but it's a normal map on a much lower polygon head.
 
Powderkeg said:
No, actually, it's quite true, in a manner of speaking.

The 1.5 million poly figure is the number of polygons used to create a normal map of the face, so there really is 1.5 million polygons making up the face, but it's a normal map on a much lower polygon head.

Where does it say anything about normal mapping in that quote?

Deriving a normal map from a high poly model and actual using a high poly model are NOT the same thing. EVERY normal map is derived from a high poly model.

That article claims that they are creating character models for their game with 1.5 million polys for the face alone.... a blatant LIE. They even directly compare how many polys they can use with the PS3 to how many they could use with the PS2..... and with the PS2 they obviously weren't talking about normal maping, as the PS2 can't do normal mapping.

I stand by my original statement : That quote is either a typo, or a blatant lie.
 
blakjedi said:
didnt ubisoft use normal mapping on ps2 version of SC:CT?

No. They couldn't use normal mapping so they used another technique that they came up with ( I can't remember the name of it right now ). Which ofcourse didn't look nearly as good as normal mapping does.

The reason the PS2 can't do normal mapping is because it can't do DOT 3 bumpmapping.
 
BenQ said:
EVERY normal map is derived from a high poly model.
Not necessarily.
BenQ said:
That article claims that they are creating character models for their game with 1.5 million polys for the face alone.... a blatant LIE.
It's not a lie it's just marketing BS, it's not that different, I know, but it had to be precised.
Just like when Tim Sweeney said that one character model in the UE3.0 demo had more polygons than a whole level of the original Unreal (up to 20Kpps). He was talking, of course, of the source high polygon model, not the in-game 5-6Kpps + Normal Maps.

Actually in this case, the article writer was also fooled by Koei presentation slides, that didn't explain that the 1.5Mpps face model was used in a APS (Appearance Preserving Simplification) process in order to create a Normal Map for a lower model.
 
BenQ said:
and with the PS2 they obviously weren't talking about normal maping, as the PS2 can't do normal mapping.

Actually PS2 can do PROPER normal mapping, as done in the new upcoming matrix game.
PS2 suprises me once more.
 
scatteh316 said:
Actually PS2 can do PROPER normal mapping, as done in the new upcoming matrix game.
PS2 suprises me once more.

Define "PROPER Bumpmapping"?

When people talk about bumpmapping ( weather they realize it or not ) they are almost always talking about DOT 3 bumpmapping......something the PS2 cannot do.
 
BenQ said:
Define "PROPER Bumpmapping"?

When people talk about bumpmapping ( weather they realize it or not ) they are almost always talking about DOT 3 bumpmapping......something the PS2 cannot do.

its the same normal mpping technique that PC and X-BOX does.

Regardless of where they came from, though, the visuals were extremely beautiful and acted as the catalyst to my eventual excitement. Path of Neo does things on PS2 previously thought impossible. Normal Mapping, for example, was shown to be entirely feasible on Sony's aging machine.

http://ps2.ign.com/articles/611/611395p1.html
 
I know that the search function is jammed and thus doesn't work properly, so I won't use the old "search function is your friend" saying, but let's just say that the topic about Normal Mapping on the PS2 has already ben discussed a few times before around here.
There a paper about that if you're really interested.

Now, back to the topic, the Ruby Demo on the Xenos, and its 1M polygons per frame.
 
*Reads Dave's announcement

That's what I'm saying!

The search function is your friend!

There's already a tons of threads about BM techniques on the PS2 or the DC, just type normal mapping and PS2/DC in the search page and a tons of threads will appear, and you will learn about the now mythical two-pass normal mapping on the PS2...

Now, Ruby, Xenos, Siggraph, Ati booth, 1M polygons per frame, on topic , keep.
 
pso said:
Well if you take the rumors about the R600, no. Please take this as a grain of salt, to the max, lol. The rumored specs for R600 is this:

65nm
64 Shader pipelines (Vec4+Scalar)
32 TMU's
32 ROPs
128 Shader Operations per Cycle
800MHz Core
102.4 billion shader ops/sec
512GFLOPs for the shaders
2 Billion triangles/sec
25.6 Gpixels/Gtexels/sec
256-bit 512MB 1.8GHz GDDR4 Memory
57.6 GB/sec Bandwidth (at 1.8GHz)
WGF2.0 Unified Shader

http://www.cdrinfo.com/Sections/News/Details.aspx?NewsId=14184

Again, take this as a grain of salt.

Yes take those specs with a very big grain of salt, they were an educated guess based on the R500 specs, they have nothing to do with ATI.

Anyway, all of our questions should be answered when R600 is launched and we see how much performance it gets compared to G70/G71 for the amount of hardware its using.
 
Back
Top