Ken Kutaragi:" I can produce the PS3 anytime"

scooby_dooby said:
Well have fun playing semantic games.

That's funny, because here I though you were the one arguing semantics all along with your creative usage of the word customized and the dubious term console GPU...

Funny how Xenos is said to be a console GPU because it differentiates itself from GPUs sold on the PC market by having a unified shader and some eDRAM. That's some very fuzzy logic there. I also find the concious amusing that you are arguing about a console GPU (whatever that is) - last I checked, these parts were dedicated pieces of hardware aimed to solve specific problems. If it's got a unified shader, a more traditional design or has eDRAM or not isn't really the point now, is it?

Xenos is as much a "console GPU" as the graphics-synthesizer wasn't (applying your logic) - they're both simply parts that are in there for certain tasks. Nothing more, nothing less.
 
Dave Baumann said:
You understand that they are differnt pipline structures, yes? G70 does its 4 by using two ALU's per pipeline, Xeno's does its two on a single co-issue ALU. It kinda even out overall!

Ok thanks, I didn't think of that. But is there a reason why RSX has two and Xenos one? It's not that even (74 vs 48).
I'm no expert in this so I'm only saying what I think but maybe I shouldn't, hmm... :p
 
Phil said:
That's funny, because here I though you were the one arguing semantics all along with your creative usage of the word customized and the dubious term console GPU...
It's really not that dubious a term here. The eDRAM and the unified shaders make Xenos a console GPU in the sense that the dedicated and fixed nature of console hardware makes it easier for people to take advantage of it. To release Xenos as a PC part would be largely unproductive, only because it would take so much work for any individual application to support it. So in the sense that it's a departure from traditional PC GPU architecture that requires developers to explicitly take advantage of it, we can say it's most useful in a console. Also, given the amount of eDRAM it has and the typical PC resolutions, the AA logic would pretty much go to waste on a PC. Sure, technically, I suppose a GPU is a GPU is a GPU, but just because all tools can get the job done, doesn't mean that some aren't better suited to certain tasks than others.
weaksauce said:
Ok thanks, I didn't think of that. But is there a reason why RSX has two and Xenos one? It's not that even (74 vs 48).
Comparing theoretical and unachievable performance statistics to each other is generally an unproductive exercise.
 
Phil said:
Xenos is as much a "console GPU" as the graphics-synthesizer wasn't (applying your logic) - they're both simply parts that are in there for certain tasks. Nothing more, nothing less.

Well, the Synthesiser wasn't a GPU was it? I thought it was just some kind of rasterizer...
 
Phil said:
Funny how Xenos is said to be a console GPU because it differentiates itself from GPUs sold on the PC market by having a unified shader and some eDRAM. That's some very fuzzy logic there.

AFAIK Xenos is the only GPU that acts as the memory controller for an entire system. That feature alone makes it unique, as the whole system was designed around that feature.
 
nelg said:
AFAIK Xenos is the only GPU that acts as the memory controller for an entire system. That feature alone makes it unique, as the whole system was designed around that feature.
NV2A was/is acting as the memory controller for the entire xbox system.
 
Sethamin said:
It's really not that dubious a term here. The eDRAM and the unified shaders make Xenos a console GPU in the sense that the dedicated and fixed nature of console hardware makes it easier for people to take advantage of it. To release Xenos as a PC part would be largely unproductive, only because it would take so much work for any individual application to support it. So in the sense that it's a departure from traditional PC GPU architecture that requires developers to explicitly take advantage of it, we can say it's most useful in a console. Also, given the amount of eDRAM it has and the typical PC resolutions, the AA logic would pretty much go to waste on a PC. Sure, technically, I suppose a GPU is a GPU is a GPU, but just because all tools can get the job done, doesn't mean that some aren't better suited to certain tasks than others.

Ignoring eDRAM, we will see unified shaders in PC GPU's soon(ish) won't we? I probably picked up rumours from somewhere like The Inq. but ignoring that, nVidia are rumoured to be on the looking into a unified architecture too aren't they?
 
Sethamin said:
It's really not that dubious a term here. The eDRAM and the unified shaders make Xenos a console GPU in the sense that the dedicated and fixed nature of console hardware makes it easier for people to take advantage of it. To release Xenos as a PC part would be largely unproductive, only because it would take so much work for any individual application to support it. So in the sense that it's a departure from traditional PC GPU architecture that requires developers to explicitly take advantage of it, we can say it's most useful in a console. Also, given the amount of eDRAM it has and the typical PC resolutions, the AA logic would pretty much go to waste on a PC.

So, in other words, your definition of a "console GPU" is a GPU that isn't used in PCs?

If we are going to look at it from a technical point of view, it'd be futile to even argue that Xenos doesn't share a lot from its PC counterparts. In fact, it seems that ATi will be bringing unified shaders to the PC market as well, so would that make that GPU a console GPU then? Applying your logic and your definition of console GPU it would, yet I don't think we want to go there.

At the end of the day, Xenos is nothing more but a GPU optimized for its environment to its task within the Xbox360 system. In that sense, RSX is nothing more and nothing less.

Sethamin said:
Sure, technically, I suppose a GPU is a GPU is a GPU, but just because all tools can get the job done, doesn't mean that some aren't better suited to certain tasks than others.

But that wasn't really the argument I was making, was it? No one is arguing which GPU is better suited to its tasks, because for one, we don't have specific details on RSX yet and on the other hand, only really touched on the tasks of each specific part and its role within the entire system. Without looking at either more closely, we can't really say which is better suited. At best, it would be an apples / oranges comparision.
 
Phil said:
So, in other words, your definition of a "console GPU" is a GPU that isn't used in PCs?
:LOL: Well that does make sense....
If we are going to look at it from a technical point of view, it'd be futile to even argue that Xenos doesn't share a lot from its PC counterparts. In fact, it seems that ATi will be bringing unified shaders to the PC market as well, so would that make that GPU a console GPU then? Applying your logic and your definition of console GPU it would, yet I don't think we want to go there.

At the end of the day, Xenos is nothing more but a GPU optimized for its environment to its task within the Xbox360 system. In that sense, RSX is nothing more and nothing less.



But that wasn't really the argument I was making, was it? No one is arguing which GPU is better suited to its tasks, because for one, we don't have specific details on RSX yet and on the other hand, only really touched on the tasks of each specific part and its role within the entire system. Without looking at either more closely, we can't really say which is better suited. At best, it would be an apples / oranges comparision.

What i found funny is that the same people slanting Sony for going for a totally custom design with PS2 are the same ones who are now raving about how cool it is that MS now have a "custom" design for themselves... The fact that there's not much "custom" there apart from eDRAM (everything else has been or will be used on PCs quite soon, and the CPU is "just" 3 off the shelf CPUs solded into one), while that Sony made their CPU from the ground up is another matter altogether.

People will always try to find ways to have their "my console is better than yours because..." competitions, and these days it's the "custom-ness" of the console that apparently gives it more "power" and "elegance" than another console they think is less "custom" than theirs.

So bored with this word fights over very silly aspects of the consoles...

Next thing you know it will be "my console is better than yours because it wasn't manufactured in a third world country"
 
I thought in this thread the 'XB360 GPU is custom and PS3 isn't' debate comes from people saying MS don't do innovation. I don't think there's anything here meant as custom is better than off the shelf or not (though admittedly I have lost track of what it's supposed to be about!). The question is whether Xenos is an innovative custom piece or a modifed PC piece, likewise for RSX, and at the end of the day who cares when MS didn't have a hand in designing either so can't really claim to be innovating new GPUs.
 
london-boy said:
People will always try to find ways to have their "my console is better than yours because..." competitions, and these days it's the "custom-ness" of the console that apparently gives it more "power" and "elegance" than another console they think is less "custom" than theirs.
To be fair, I don't think that was the intent, even though some reacted as though it was. A poster remarked that MS didn't really invest in R&D for the Xbox 360. This seems false, regardless of whether ATI borrowed pieces of of its own technology, or IBM tripled up on its cores.

The comparison to Sony came about as if to argue that since Sony invested more, the PS3 must be more valuable to Sony. It was argued that Sony invested more in Cell because it could use it in more markets, therefore it was not a PS3-only investment, unlike MS Xbox 360-investment. Regardless, even if Sony did dump more R&D dollars into PS3, it logically does not mean that Xbox 360 was starved for R&D.

.Sis

PS My console is better than yours.
 
Shifty Geezer said:
I thought in this thread the 'XB360 GPU is custom and PS3 isn't' debate comes from people saying MS don't do innovation. I don't think there's anything here meant as custom is better than off the shelf or not (though admittedly I have lost track of what it's supposed to be about!). The question is whether Xenos is an innovative custom piece or a modifed PC piece, likewise for RSX, and at the end of the day who cares when MS didn't have a hand in designing either so can't really claim to be innovating new GPUs.

Somewhere on this board, I read that MS did have a VERY large hand in designing Xenos but searching for "xenos" "design" "MS" turns up thousands of posts...
 
blakjedi said:
Somewhere on this board, I read that MS did have a VERY large hand in designing Xenos but searching for "xenos" "design" "MS" turns up thousands of posts...
That's what I recall--though it probably was more around guidance than anything. But I thought MS had a few engineers on staff from that old Web TV or Utimate TV or whatever it was called. In fact, aren't they using that scaler techn in the Xbox 360?
 
and the CPU is "just" 3 off the shelf CPUs solded into one

Those CPUs were built to order. The front end is the same as or at least very similar* to the Cell's PPE but the vector units and I/O system is completely custom. They're certainly not OTS parts as you can't get those cores anywhere else (the G5 is a completely different processor).

The PPE's vector units are also custom as they are not standard VMX (you may note there's very little public info on them).


*The PPE's front end may have changed a little, I found a discrepancy in the docs, though it could be a mistake. They have had an extra year to work on it so I wouldn't be the least bit surprised if there were some changes.
 
london-boy said:
But mine stands up for a looong time and keeps on going. Bit noisy though.
As long as you don't stand it up while your using it--that could damage whatever you have in there.



:oops:
 
Sis said:
That's what I recall--though it probably was more around guidance than anything. But I thought MS had a few engineers on staff from that old Web TV or Utimate TV or whatever it was called. In fact, aren't they using that scaler techn in the Xbox 360?

MS has several ATG (or Advanced Technology Groups) around the world who supposedly had a hand in helping design (or maybe just set the specifications for) Xenos and Xenon. For example I think I remember reading in a print article that the Shanghai ATG had been working Procedural Synthesis algorithms for the last five years or so. And that they specific required the dot product instruction in the Xenon's pipeline for that task. I think the cache-lock stream to Xenos capability is definitely an offshot of that capability built into Xenos.
 
Phil said:
So, in other words, your definition of a "console GPU" is a GPU that isn't used in PCs?
<shrugs> Sure, why not? You know of any other places a GPU is used? I did also note (implicitly) that the size of the eDRAM is suited to resolutions that a console outputs.

Phil said:
But that wasn't really the argument I was making, was it? No one is arguing which GPU is better suited to its tasks, because for one, we don't have specific details on RSX yet and on the other hand, only really touched on the tasks of each specific part and its role within the entire system. Without looking at either more closely, we can't really say which is better suited. At best, it would be an apples / oranges comparision.
I wasn't really addressing your argument. I was only addressing your point that you can't call Xenos a console GPU. I agree with you wholeheartedly that we can't say anything about a Xenos to RSX comparison, because we don't know much of anything about RSX at all. Even when we do, it's still going to be an apples to oranges comparison, although I'm sure that won't stop anyone. Which brings me to my next point...

weaksauce said:
You think RSX will be less than 65% efficent?
I have no idea, and neither do you. Even if we did, it's still a meaningless comparison. They're different architectures, and they'll be good at different workloads. It's very difficult to ever definitively say that "X is slower, Y is faster" in the computer world, because everything is a tradeoff to optimize the common case. The less often a man makes declarative statements, the less apt he is to look foolish in retrospect.
 
Back
Top