Xenos - RSX - What was left out?

[Titanio]

Well, MS wasnt exaclty lying with the GP stuff....Sure they may have less on Beta than on Alpha and Have much more FP on beta than on ALpha, but they never said that they Had more GP on beta than on Alpha.

There has to be a tradeoff, ultimally, Xenon still has like 3x more GP than the PS3, and thats where they are coming from.

The point is, if their GP isn't even that good...who defines where GP ends and begins? It seems to be a very relative thing. SPEs are more GP than VUs for example. They just seem to be counting their PPEs and counting Sony's, and going "oh look, we have 3x as many". Well hey, look! Sony has 7 SPEs and MS has none - they have infinitely more!

I think not counting the SPEs when your cores aren't that general purpose anyway is silly - I don't think they can use the PPE or X360 cores as a lower boundary on what constitutes a GP core - who defines where that is? Relative to a G5 or P4 people could say the X360 cores aren't very GP.

In the end, making this a point of contrast ultimately seems very silly when they admit themselves privately that it wasn't their priority. They were simply looking for any way to talk up their numbers relative to Sony post-E3, and whilst it wasn't even very convincing at the time, it's even less so now.

 

[therealskywolf]

^ There must be a reason why SOny was only counting FP no?

Oh i see.... Look i'm kinda sick of this Hardware vs Hardware. I'll let games be the judge.

 

[Pozer]

Well, MS wasnt exaclty lying with the GP stuff....Sure they may have less on Beta than on Alpha and Have much more FP on beta than on ALpha, but they never said that they Had more GP on beta than on Alpha.

There has to be a tradeoff, ultimally, Xenon still has like 3x more GP than the PS3, and thats where they are coming from.

The point is, if their GP isn't even that good...who defines where GP ends and begins? It seems to be a very relative thing. SPEs are more GP than VUs for example. They just seem to be counting their PPEs and counting Sony's, and going "oh look, we have 3x as many". Well hey, look! Sony has 7 SPEs and MS has none - they have infinitely more!

I think not counting the SPEs when your cores aren't that general purpose anyway is silly - I don't think they can use the PPE or X360 cores as a lower boundary on what constitutes a GP core - who defines where that is? Relative to a G5 or P4 people could say the X360 cores aren't very GP.

In the end, making this a point of contrast ultimately seems very silly when they admit themselves privately that it wasn't their priority. They were simply looking for any way to talk up their numbers relative to Sony post-E3, and whilst it wasn't even very convincing at the time, it's even less so now.

In a perfect world I'd agree with you 100%, however considering some of the mudslinging KK has done in the last 3 months it's kinda hard to consider Sony a PR victim.

Both sides seem to be making overly false claims of superiority but some of the KK remarks have been very low class.

 

[Titanio]

^ There must be a reason why SOny was only counting FP no?

The thing is, there are standard metrics for counting (paper) floating point performance, and it is an oft-used one for comparing system power. We can see how Sony arrived at their figures, and they counted everything MS has to offer in that regard.

Microsoft's comparison is based on no such metric. How do we count "general processing" performance? MS has taken it upon themselves to count only PPEs and no SPEs. Who are they to do this? Sony counted everything MS was offering wrt floating point performance, but the reverse cannot be said of MS's comparison wrt "general processing", IMO.

I just think it's a very fitting end to that comparison - that the one thing they were crowing about publically is something which there system isn't exactly a shining star at either. It's like an average student highlighting the short comings of a less than average student, whilst ignoring his exceptional abilities in other areas.

 

[Farid]

There has to be a tradeoff, ultimally, Xenon still has like 3x more GP than the PS3, and thats where they are coming from.

Not exactly, MS PR consider that the Cell CPU has only one core, the PPE, the SPEs are simply related as "somethings" (the exact term) in their marketing plan.

Since the PPE is a PPC core, and seeing that the the XeCPU has 3 PPC cores, MS consider logically (Marketing-logic, before someone objects) that the XeCPU has 3x time the GP power the Cell has.

Of course when you include the SPEs, the XeCPU lose this GP advantage.

 

[therealskywolf]

There has to be a tradeoff, ultimally, Xenon still has like 3x more GP than the PS3, and thats where they are coming from.

Not exactly, MS PR consider that the Cell CPU has only one core, the PPE, the SPEs are simply related as "somethings" (the exact term) in their marketing plan.

Since the PPE is a PPC core, and seeing that the the XeCPU has 3 PPC cores, MS consider logically (Marketing-logic, before someone objects) that the XeCPU has 3x time the GP power the Cell has.

Of course when you include the SPEs, the XeCPU lose this GP advantage.

So now SPEs are good at GP?

And we cant forget that we cant count PPE + Spes at the sametime, since PPE will be busy (at least a part of it) feeding the Spes right? ITs not like they work on their own like in Xenon.

 

[Titanio]

So now SPEs are good at GP?

Oh, something tells me being "good" at GP isn't a criteria for inclusion as far as MS is concerned

It's not about being good. It's about whether they'll never find a use for "general purpose" processing on them.

And we cant forget that we cant count PPE + Spes at the sametime, since PPE will be busy (at least a part of it) feeding the Spes right? ITs not like they work on their own like in Xenon.

It depends what model you use. The SPEs can work away with little to no PPE involvement. The PPE can and is being used for game code.

 

[Sis]

I just think it's a very fitting end to that comparison - that the one thing they were crowing about publically is something which there system isn't exactly a shining star at either. It's like an average student highlighting the short comings of a less than average student, whilst ignoring his exceptional abilities in other areas.

No, it's like a sprinter saying "I'm a faster runner than my competitor" and someone saying, "You think you're a good runner? Your marathon times are terrible!"

.Sis

 

[Farid]

So now SPEs are good at GP?

Yes, they're good at GP work, they're just not as easy as a PPC core to work with.

And we cant forget that we cant count PPE + Spes at the sametime, since PPE will be busy (at least a part of it) feeding the Spes right?

This is not mandatory.

ITs not like they work on their own like in Xenon.

SPEs can work on their own.

 

[Titanio]

No, it's like a sprinter saying "I'm a faster runner than my competitor" and someone saying, "You think you're a good runner? Your marathon times are terrible!"

.Sis

That analogy is a little less clearly tied to what we're describing..might like to map it back for me Who's the sprinter? Who's the marathon runner?

 

[Sis]

No, it's like a sprinter saying "I'm a faster runner than my competitor" and someone saying, "You think you're a good runner? Your marathon times are terrible!"

.Sis

That analogy is a little less clearly tied to what we're describing..might like to map it back for me Who's the sprinter? Who's the marathon runner?

Eh, it made sense in my own head.

If the XeCPU is a sprinter, and marathon runners are standard off the shell chips, and running is general purpose processing, then, uh... I hate analogies!

Your quote:

I just think it's a very fitting end to that comparison - that the one thing they were crowing about publically is something which there system isn't exactly a shining star at either.

implies that you expect the XeCPU to have GP performance capabilities as that of an off the shelf component from Intel or AMD. I was merely trying to suggest that perhaps that's not a fair comparison, since the chips have different purposes, but it doens't detract from the XeCPUs GP performance.

.Sis

 

[Brimstone]

So now SPEs are good at GP?

Yes, they're good at GP work, they're just not as easy as a PPC core to work with.

The SPE's just keep getting better and better the more we understand them. Computer Science is being re-written.

 

[Johnny Awesome]

Well, from what I'm seeing it looks like CELL is about twice as powerful as Xenon for realworld game development once this generation gets cooking. 8 hardware elements vs. 4 (3 *1.33 - boast from the extra hardware threads) for Xenon.

I'm a big X360 fan, but I think this is how it will all shake out. It's not surprising that Sony's CPU will be twice as powerful. It has a few more transistors and is coming out 6 months later.

90% of the games won't be noticeably different though, I imagine.

 

[Titanio]

Eh, it made sense in my own head.

If the XeCPU is a sprinter, and marathon runners are standard off the shell chips, and running is general purpose processing, then, uh... I hate analogies!

Your quote:

I just think it's a very fitting end to that comparison - that the one thing they were crowing about publically is something which there system isn't exactly a shining star at either.

implies that you expect the XeCPU to have GP performance capabilities as that of an off the shelf component from Intel or AMD. I was merely trying to suggest that perhaps that's not a fair comparison, since the chips have different purposes, but it doens't detract from the XeCPUs GP performance.

.Sis

I didn't mean to suggest its GP capabilities were on a par with off-the-shelf desktop chips - from what we're hearing it doesn't. I'm saying it's funny they made such a point over that vs Cell given that their own performance in that area is less than stellar.

 

[PC-Engine]

There has to be a tradeoff, ultimally, Xenon still has like 3x more GP than the PS3, and thats where they are coming from.

Not exactly, MS PR consider that the Cell CPU has only one core, the PPE, the SPEs are simply related as "somethings" (the exact term) in their marketing plan.

Since the PPE is a PPC core, and seeing that the the XeCPU has 3 PPC cores, MS consider logically (Marketing-logic, before someone objects) that the XeCPU has 3x time the GP power the Cell has.

Of course when you include the SPEs, the XeCPU lose this GP advantage.

Then you have the 80% general purpose game code which Xenon seems to be better at.

 

[3roxor]

Then you have the 80% general purpose game code which Xenon seems to be better at.

I see you fell for that damage control microsoft did

 

[Titanio]

Then you have the 80% general purpose game code which Xenon seems to be better at.

You believe that?

Even if we want to run with that - 80% of what? Lines of code? Execution time?

You could have have a program where 80% of the lines of code is doing one thing, but 80% of the execution time is spent doing another..

And if you think Xenon has an advantage there - how large is that advantage? As large as Cell's advantages in other areas?

Regardless, devs would seem to disagree. Or Tim Sweeney at least - he seems to think the things the SPEs aren't "immediately" useful for only take a small fraction of execution time anyway. Doesn't exactly gel with MS's argument very well..

 

[version]

you speaking prototype cell processor, but in ps3 will be second generation cell cpu with full 2 thread... xenon has 3....

 

[Fafalada]

There must be a reason why SOny was only counting FP no?

Well the reason would be that an actual Metric exists for FP(reliable or not), while there's no such thing for the so called "GP power".
Which makes it all the better target for damage control PR exploits.

 

[PC-Engine]

you speaking prototype cell processor, but in ps3 will be second generation cell cpu with full 2 thread... xenon has 3....

What are you talking about?

Anyway if one reads the leak carefully, it talks about the performance of each core. Alpha only has 2 while final has 3. I'll let the rest of you figure out what that means.

It also talks about performance per clock. It's pretty obvious that higher clock designs will be able to do less work per clock per core.