Cell article: 40x faster than EE, Cell arch to last 10 yrs

[ERP]

I'm bored so .....

The setup limit is 2cy/tri on xbox, I don't remember off the top of my head exactly how many iterators you can use in this case, but I believe it includes shading and possibly one texture.

In completly artificial cases you can always hit the setup limit, because you can just index in cache vertices and do no transform work.

Assuming minimal useful vertex (Perspective correction and transform) work and no cache hits, then you can transform in 3cy/vertex. You can actually do more than this in 3cy because there are unused execution units and a stall in this shader.

So the threoretical maximum is 116.5 million
The vertex transform rate more like 77 million although this totally ignores the post transform cache.

You can actually get pretty close to the first number in a real application if you render rectangular patches with very fine tessalation, although the data has to be perfectly laid out.

 

[Phil]

Qroach,

I don't think you've been paying attention to what I wrote above. Again:

- 66 million/pps number is clearly a stated performance figure of VU1 and excludes VU0. <check>

- it was never a number with any ambiguous "in-game" scenario in mind, why pretend it was? <check>

- I don't put much weight into what *you* think Xbox could achieve or not, as I'm talking about what was publicised by marketing at the time <check>

- Xbox was marketed with polygon numbers using inflated "effective drawing" fillrate numbers which it never had (~4 Gpixels instead of ~1). <check>

There's really no point in taking this any further. Simple facts. Quit being so damn defensive.


PS: VU0 isn't crippled by anychance, how do you get that idea? It isn't as complex as VU1, but why would it have to? It doesn't have a direct path to the GIF and can be used in many ways in addition to VU1.

 

[ERP]

If I remember correctly the 66 million number the GS setup limit.

If you include perpective division VU1 can't reach that limit on it's own I believe.

 

[Qroach]

66 million/pps number is clearly a stated performance figure of VU1 and excludes VU0. <check>

Ok, for one thing I said I "wouldn't be suprised" if it was included. Not that it indeed was <check>

- it was never a number with any ambiguous "in-game" scenario in mind, why pretend it was? <check>

Who's pretending it was, i said the numbers were theoretical. <check>. Obviously not in-game.

- I don't put much weight into what *you* think Xbox could achieve or not, as I'm talking about what was publicised by marketing at the time <check>

it really shows how impartial you are phil. You can believe whatever you want, ok? no need to get so uptight. So You're telling me the 66 million polygon figure you said "sony" publicised were actually NOT publicised by sony?<check what you wrote>

- Xbox was marketed with polygon numbers using inflated "effective drawing" fillrate numbers which it never had (~4 Gpixels instead of ~1). <check>

That's not correct <check> don't confuse the effective fillrate numbers published with the polygon performance numbers published.

There's really no point in taking this any further. Simple facts. Quit being so damn defensive.

I'm being defensive? no, I'm afraid you're being defensive <check> Does this mean ERP is also being defensive? I'm just curious what you're basing the "defensive" idea on.

VU0 isn't crippled by anychance, how do you get that idea? It isn't as complex as VU1, but why would it have to? It doesn't have a direct path to the GIF and can be used in many ways in addition to VU1.

That's what I meant by crippled. It could have been more useful from what PS2 programmers have told me, but you can trust your judgement on that if you want.

 

[Phil]

This talk is getting awefully redundant so I won't be replying to any more off topic talk about something that wasn't really a core argument but more just a side-in comment.

I was merely pointing out that Sony *could* have posted a much higher theoretical polygon number derrived by both VUs dedicated to geometry which would be equal to the number Microsoft is derriving by stating that their system can do 1 TFLOP! It's all subjective of course, that's just my take on it. In the end, the 66 figure is as valid as the 116.5 one for Xbox (a peak figure), which it never was marketed as to being anything else.

 

[BOOMEXPLODE]

It's true, MS is atleast as bad as Sony with their hype. Of course PS2 had ridiculous hype, but so did X-Box, and Now X-Box 360 does 1 teraflop? If hype were an olympic event, Sony and MS would be neck and neck.

 

[randycat99]

Aside from this bit of side-discussion, I think the point that was made a bit earlier that if there is anyone to complain that Sony bakes their specs, they should consider that MS bakes them even more...and I think that is pretty much on-target (not to mention the "glass houses" advice). It was true in the latest generation, and appears to be true with the next generation (with this "Xenon will be in excess of 1 TFLOP" statement).

So what we have left, is the "1 TFLOP PS3" argument cannot be substantiated in the way that the "critics" would wish (i.e., person xyz from Sony said abc in press event 123).

In an equal sense, the goal of 1 TFLOP is certainly within reach using Cell architecture. So there is really very little left to argue on that front (unless you are obsessed with the "who claimed PS3 would be 1 TFLOP" subject).

PS3 will come when the time has come. It will be great. There's really not much to complain about unless you are bent on finding something to complain about... (waaaa-waaaaa, PS3 won't be 6.2 TFLOPs, for example) C'mon, people! Let's get a grip, already?

 

[Qroach]

Boomexplode, exactly.

 

[Qroach]

I think the point that was made a bit earlier that if there is anyone to complain that Sony bakes their specs, they should consider that MS bakes them even more...

Jeez, whats with the one sided argument? MS is just as bad Sony and the reverse. This incessive need to try and prove how your favourite vendor is in someway "more honest" is ridiculous. This was teh same when people kept trying to prove how thier favourite video card vendero was the "most honest"

PS3 will come when the time has come. It will be great. There's really not much to complain about unless you are bent on finding something to complain about... (waaaa-waaaaa, PS3 won't be 6.2 TFLOPs, for example) C'mon, people! Let's get a grip, already?

If you change what you wrote only a little...

Xbox 360 will come when the time has come. It will be great. There's really not much to complain about unless you are bent on finding something to complain about... (waaaa-waaaaa, Xbox 360 won't be 6.2 TFLOPs, for example) C'mon, people! Let's get a grip, already?

apply this same logic to xbox 360. why complain if it can't reach a terflop or 5 trillion gigaflops if it's still amazing?

 

[jvd]

Man talk about stiring the roost with saying only one thing .


Look it was said by sony that the cell will be 1tflop . Its vague but sony is talking about cell being in the ps3 . Its hype by sony to get fan boys going (it worked here) and then later on they could claim they only said cell as whatever will hit that target .

Its the same with ms saying the x360 will have over 1tflop performance.

Its marketing speak. I was pointing out that sony does it too . You guys take to much stuff to heart esp when it comes to sony .

 

[randycat99]

You don't see anywhere in my post that Xenon will be a failure for not really delivering >1 TFLOP, do you? However, it definitely is worth mention that MS actually did make that claim, when it is even clear to us that the numbers cannot remotely add up to support that claim.

Hence, there is only one side to this story. Sony has made claims that are taken out of context, and some see fit to fault them for this. MS just makes outright incorrect claims in a bid to take Sony's hype+1- no context, just plain, baldfaced misleading. Hence, those who complain incessantly that Sony is full of hype and lies, need only sit down and realize that their star MS does it worse- not "as bad", not "less than, but no further than", worse. Hence, basing an argument from this [the must beat back Sony's lies crusade] is faulty from the start. Those people should take note.

Why should this not be turned around for PS3 fans, as well? Put simply, you don't see them posting a new topic a day of how XB2 will be a failure because xyz, don't believe the hype, blah-blah-blah... That's why. So your questioning me why I didn't write my comment in reverse is bogus. You got fingers- you write it, if you want to hear it so bad. Take note how there is no element in that statement that makes a dig on the "other" console. It was only there to say that PS3 will come when it comes and it will deliver serious performance, regardless of what claim has been made by who knows who... That is the bottomline. That's it, nothing more. So don't attempt to read any further meanings into it.

 

[randycat99]

Man talk about stiring the roost with saying only one thing .


Look it was said by sony that the cell will be 1tflop .

Actually, you made the claim that this was said for PS3. It seems necessary to reiterate that no one has been able to substantiate your claim. Nor have you been particularly initiated to substantiate it yourself (30 pgs or not). So your statement turns out to be utterly unfounded insofar. You just hope that it will "stick".

(Does this ring a bell at all wrt a certain CBS newscaster that wanted a story to be true so bad, he just said it was true, and hoped no one would go back to verify the source...or lack of one)

 

[jvd]

Actually no . Sony said that cell would be 1 tflop. They said cell will be the cpu in the ps3 .

IF you want to dispute facts go ahead. Good luck .

But since this is turning into a flame war quickly with sony fans argueing against every other fan on the boards this is getting locked