ATI - PS3 is Unrefined

Jaws said:
The source is from an ATI technical presentation and also from a MS leaked doc.

No, you've pointed to a thread where someone has done a poor translation of some article that is supposed to be from some presentation. Unless I see it from a trusted source or black and white, I really can't trust it. It just doesn't make much send to do a 4+1 design and make the + 1 only capable of doing adds or muls.

Can you show me Dave's source?

ATI according to dave.

Aaron Spink
speaking for myself inc.
 
expletive said:
Are you being serious?

Even if I was making an educated guess...90% of the time I would be right because I had time on my side. Besides...if you read many of threads on this forum, in terms of raw performance, RSX is clearly ahead. overall, the system [PS3] will most-likely have a little more umph than XBOX360...almost all major devs have admitted this (even John Carmack, who notoriously still prefers the PC, Microsoft, and anything with any relation to Bill Gates) .

Do you not agree?

Or are you arguing just for arguments sake?

Besides, technical power has little to do with how well a console does. Quality content is the biggest differentiating factor.
 
aaronspink said:
No, you've pointed to a thread where someone has done a poor translation of some article that is supposed to be from some presentation. Unless I see it from a trusted source or black and white, I really can't trust it. It just doesn't make much send to do a 4+1 design and make the + 1 only capable of doing adds or muls.

http://www.watch.impress.co.jp/game/docs/20050520/x360_g.htm

aaronspink said:
ATI according to dave.

Not being funny, but I'd like a link please? I've been posting that number on this forum for more than a year...
 
Jaws said:
http://www.watch.impress.co.jp/game/docs/20050520/x360_g.htm

Not being funny, but I'd like a link please? I've been posting that number on this forum for more than a year...
Jaws, all the information I have on Xenos was given to me by ATI - it states that in the article. The article you are linking to above has pictures from the E3 press breifing that ATI did and they didn't actually go into much in the way of specifics there; they certainly didn't detail any FLOPS figures.
 
ROG27 said:
Even if I was making an educated guess...90% of the time I would be right because I had time on my side. Besides...if you read many of threads on this forum, in terms of raw performance, RSX is clearly ahead. overall, the system [PS3] will most-likely have a little more umph than XBOX360...almost all major devs have admitted this (even John Carmack, who notoriously still prefers the PC, Microsoft, and anything with any relation to Bill Gates) .

Do you not agree?

Or are you arguing just for arguments sake?

Besides, technical power has little to do with how well a console does. Quality content is the biggest differentiating factor.
Carmack didn't explicity state or imply that the PS3 was more powerful. He was just saying that he'd prefer a system that was 20% easier to develop for than a system that was 20% more powerful.

That doesn't mean that Xbox 360 is 20% more developer friendly than the PS3 or that the PS3 is 20% more powerful than the Xbox 360. It was just a general statement.

Again, I point to my sig.
 
Dave Baumann said:
Jaws, all the information I have on Xenos was given to me by ATI - it states that in the article. The article you are linking to above has pictures from the E3 press breifing that ATI did and they didn't actually go into much in the way of specifics there; they certainly didn't detail any FLOPS figures.

I've read your article a while ago. Can you point to me where it mentions 240 Gflops, because I don't remember reading it?

The linked article derives it's numbers from the information presented. The leaked doc I have also confirms this.
 
Jawed said:
Sigh, I always forget compression.

Erm, less than that!
I was just going to say! Also isn't Lair something out of the ordinary in terms of texture use? So they're managing some very impressive texture work already in something like...I dunno, 50 MBs? So far there's not much in the way od scenery though.

As for procedural textures, I'm unsure how these will work. Presumably that's what the direct CPU to GPU BW is to be used for but how can the GPU actually use them without them being cached somwhere? The data needs to be there on the texture fetch request. So adding something like Perlin noise doesn't seem straightforward.
 
dukmahsik said:
yeah honestly dave's article is way more authentic

The article Jaws refers to - and you originally posted - is from Watch Impress. I'd hesitate to question its authenticity. Mistakes happen, but given that they even derive the figure I'd wonder..
 
dukmahsik said:
i think we need some more definitive rsx details, why are they holding back so much?

Perhaps Sony has been working with the nVidia G80 team in parallel. Work on the G80 just isn't finished yet, delaying the RSX slightly.
 
Jaws said:
I've read your article a while ago. Can you point to me where it mentions 240 Gflops, because I don't remember reading it?
I don't do a FLOPS breakdown (since I find them fairly meaningless), but the ALU's are described a "5D"; ATI gave no indications of any difference in the components.

The linked article derives it's numbers from the information presented. The leaked doc I have also confirms this.
I'm saying that that the pictures in that article do not contain that information - I was both at that presentation and have the full presentation and those figures were not contained in it. Bob held a number of briefings after the presentation, but if you look around the web there were different FLOP figures because he didn't give them correctly to different people.
 
Dave Baumann said:
I don't do a FLOPS breakdown (since I find them fairly meaningless), but the ALU's are described a "5D"; ATI gave no indications of any difference in the components.

Okay, but 216 and 240 would still be 5D though.

Dave Baumann said:
I'm saying that that the pictures in that article do not contain that information - I was both at that presentation and have the full presentation and those figures were not contained in it. Bob held a number of briefings after the presentation, but if you look around the web there were different FLOP figures because he didn't give them correctly to different people.

I recall the different numbers. That doesn't confirm 216 or 240 either way. Watch have taken the scalar as non-MADD capable with a breakdown derivation. Is that an assumption or info they attained from someone there? Also the MS doc still says 216...
 
Jaws said:
Okay, but 216 and 240 would still be 5D though.
Yes, but the implication being they are all as capable.

I recall the different numbers. That doesn't confirm 216 or 240 either way. Watch have taken the scalar as non-MADD capable with a breakdown derivation. Is that an assumption or info they attained from someone there? Also the MS doc still says 216...
I don't know where they have got their information from - as I said, its not been described to me in such a fashion and it wasn't in the presentation they are picturing.

As for leaked MS information the leaked docs I have only have the following caveat on the Vector/Scalar capbilities:

Vector and scalar operations can be coissued, for a total of 96 ALU operations per cycle. One limitation on coissuing is that vector operations with three source operands cannot be coissued with a scalar operation unless the scalar operation uses the third source operand of the vector operation as its input.
I can't find any other references to indicate differences between the Vector and Scalar elements.
 
Dave Baumann said:
Yes, but the implication being they are all as capable.

That would still be an assumption.

Dave Baumann said:
I don't know where they have got their information from - as I said, its not been described to me in such a fashion and it wasn't in the presentation they are picturing.

From the Watch link, I doubt it was presented on any screen. The derivation that watch did was very strange to get 216. And it's non-obvious. As I said, I and others on this board have used 240 since the original leak in 2004. Though that's still an assumption.

Dave Baumann said:
As for leaked MS information the leaked docs I have only have the following caveat on the Vector/Scalar capbilities:


I can't find any other references to indicate differences between the Vector and Scalar elements.

Is this from the original leak of 2004? The one I'm referring to was a few months ago...
 
Jaws said:
Is this from the original leak of 2004? The one I'm referring to was a few months ago...
No. I've just seen where you are getting your figure from. Look in that doc and there is a hyperlink under the table referencing an overview doc - that quote is from that overview doc.
 
Dave Baumann said:
No. I've just seen where you are getting your figure from. Look in that doc and there is a hyperlink under the table referencing an overview doc - that quote is from that overview doc.

"Xbox 360 GPU Overview " ...ah okay, but that's a dead link for me. But yeah, 216 is there, though not easy to find as I originally missed it myself!
 
Alpha_Spartan said:
Carmack didn't explicity state or imply that the PS3 was more powerful. He was just saying that he'd prefer a system that was 20% easier to develop for than a system that was 20% more powerful.

That doesn't mean that Xbox 360 is 20% more developer friendly than the PS3 or that the PS3 is 20% more powerful than the Xbox 360. It was just a general statement.

Again, I point to my sig.

Now, I must ask you, "Are you being serious?" Because if you think he's being purely theoretical there and not refering to PS3 and XBOX360...you are guilty of wishful thinking.

Life is full of things which must be inferred and extrapolated...you don't need to be beat over the face with things to know something is true. This type of thing just breeds cynicism.
 
Back
Top