Thread 105814132

Why did the CELL processor in the PS3 filter so many programmers and devs?

Closest example was by a Guerilla dev (worked on killzone 2) "Cell CPU is like your mom's plump ass, big beautiful and nobody knew how to use it"

>>105814132 (OP)
>Why did the CELL processor in the PS3 filter so many programmers and devs?
Please, show us how it is done, OP.

Because too many cellcels could figure out how to use mom's plump CELL ass.

>>105814132 (OP)
>Why did the CELL processor in the PS3 filter so many programmers and devs?
let's see:
> 2408 games
looks like it didn't filter anybody. you keep posting this same cringe thread every few weeks. all you do is change the picture and the text slightly. you're worse than that child sex offender that keeps posting the same 2-3 pictures of steve jobs trying to pretend he's not indian. please fuck off and die.

>>105815594
It's not exactly hard to write software that runs on CELL.
The hard part is writing software that utilizes the SPEs to a high degree to get the most out of it.

>>105814132 (OP)
single regular 64bit powerpc core + 8̶ 7 "SPE" special sauce vector co-processors
it was designed at an awkward time, the worse time for this kind of chip really. literally right as unified shader gpus and multi-core cpus were being made. here's cell, instead a more beefed up ps2-like design, originally intended as both the cpu and gpu, due to the advancements in gpus and the difficulty in making the cell (one of the spe's was always disabled due to yields) they were forced to slap an nvidia gpus next to it, which left it as a weird single core cpu that had most of it's potential tied up in these unique vector co-processors but you weren't using them for graphics anymore...
plenty of potential if you dedicate the time to make something uniquely suited for that hardware (i.e. not cross-platform), but wasted otherwise

as for the ps2 comparison, it's "cpu" was a single MIPS core alongside two "SPU" special sauce vector co-processors, and a fixed-function rasteriser for graphics. the ps2 was also quite unique. it's architecture and raw memory speed effectively allowed for programmable shader-like graphics effects in software, in 1999 (final hardware) before gpus with shaders of any kind were a thing. it's immense popularity forced devs to figure it out

>>105815725
-- and by comparison, the xbox 360 shipped with 3 powerpc cores and one of the first unified shader gpus
it was just so much more standard. the ps2 was a product of its' time and really as good as it could have been, the more standard approach when the ps2 was developed is.. well, the dreamcast basically. but for the ps3? they must have had some kind of sunk cost issue with the cell or something because it was pretty obvious looking at pc hardware where things were going, and it was not another ps2

>>105814132 (OP)
Actually the PS2 and PS1 architectures were also very "non conventional" (compared to a PC), devs became lazy when the xbox, which was basically a pc, came out. Programming SPEs was a pain in the ass, it wasn't just a thread to spawn, you needed to (pre)compile programs that would run on them, much like shaders GPUs

>>105815823
the ps1 wasn't anything too weird really. it's hard to define how standard it was considering 3D on a pc wasn't a thing outside of software renderers in 1994, so while the ps1's "GTE" 3d processor wasn't standard, that's because there wasn't a standard. outside of that it was a pretty typical layout, more typical than it's main competitors such as the saturn (dual cpu so early most devs couldn't handle it, graphics processor designed for 2D when 3D was the thing everyone wanted) and n64 (unified memory with a high latency rdram bus which made getting any kind of performance out of it quite the pain in the ass)

>>105814132 (OP)
Lotta multithreading

>>105814132 (OP)
>Why did the CELL processor in the PS3 filter so many programmers and devs?

The processor layout was a strong break with previous processor designs and software ecosystems need a while to catch up to new hardware.
There is no time for that catchup in a console cycle, you kinda end up rawdogging it.
So you have devs work outside of an ecosystem literally figuring everything out themselves instead of relying on some shit that others figured out before.
Ofc you can just say this is a skill issue, and it is, but sony fucked that one up desu.
You don't want to make your console be difficult to develop games for.

There's a reason Xbox and PS now both PCs with an X86 cpu. you can leverage all the x86 ecosystem optimisations for free.

Little to no documentation at the start. They were given the CELL without any information on how to get up and running and were just expected to figure it out.

>>105816043
iirc it was this, documentation was lacking early on but over the PS3's lifespan devs collected/made their own and Sony themselves released more and more.

The PS2 also suffered from that early on with its "Emotion Engine" but that didn't stop the PS2 from having a bajillion games by the end of its lifespan.

There were other issues that buckbroke devs too, mostly having to do with licensing. Sony had a lot of janky rules about getting software onto their platform and a lot of devs simply didn't want to bother. MS, meanwhile, basically just went "Make sure it's not literal malware and that your button prompt icons are the right color and you can do whatever the fuck you want"

>>105816095
>Less than two years before the PS3’s launch, Cerny shifted to working on launch titles. At that time focus was “99% hardware, 1% software”. There was no debugging, no profiling, no graphics driver – it was all “in a primitive state” – and the third parties like EA and Ubisoft were “having an easily more difficult time.”

How many times is the Cell autist going to astroturf these threads?

>>105816158
Ever seen the 2500k guy?

>>105814132 (OP)
The issue wasn't that it was too difficult, the actual issue was that the architecture was too unique, too different from every other system of the era. You could learn to use and master the Cell processor and its SPU but then all the skills you've built up, all the libraries and the tricks you've discovered to optimize the games are useless outside of the PS3 and a couple of IBM mainframes. It's the reason why Gabe Newell infamously called the PS3 "a waste of everyone's time".

>>105816200
The GAPEstation? The GAYstation?

>>105816200
the ps2 was the last time a custom architecture made sense, but the time the ps3 came out, it was no longer acceptable.
prior to the ps2 that was normal, consoles were all unique in usually pretty big ways and you could expect to rewrite large amounts of code if you wanted your game on more than one platform, and in many if not most cases, games were just totally different on other platforms (like literally just similes on different engines made by other dev team). by the 7th gen however, this was just not practical, even the 6th gen it wasn't ideal but they still did it sometimes

>>105814132 (OP)
You’re autistic

My brain damaged computer science teacher in 2013 said the CELL was still the most powerful consumer CPU on the market. I asked him what value a CPU has if nobody knows how to use it.
He said it just takes time and when devs realise its potential and better revisions and documentation with the cell 2 it would be the future of laptops and pcs.
Retard.

>>105816511
No he didn’t

>>105816511
i mean the current meta is ridiculous low speed but 12 core processors so he was kinda right

that said i doubt a teacher would give a fuck about PS3s

>>105816511
he wasn't alone in people saying that, the problem is that such people are comparing the cell to (just) cpus
the cell is a combination of 1 ppc core and 8 essentially floating point units, it's simply unfair to compare it to a regular cpu, because it's more of a hybrid between a cpu and gpu. once you compare it's apparently ridiculous floating point performance to a /gpu/, it stops being so amazing

i haven't looked into how the cell's FP performance compared to modern cpus, but the point is is that it doesn't matter, because in the mean time we've gotten GPGPU and such things are done on the GPU now

>>105815823
at the time of the ps2 and ps1 3d graphics hardware was hardly standardized, you had all kinds of weird shit, from sgi machines to voodoo banshees to sega saturns , by 2004/5 things had changed alot , the programmable pixel and vertex shader was implemented, most of the weird 3d chips had gone extinct and graphics apis had reached industry standardization, the ps3 had basically continued the development path of the ps2 against the rising tide and they fucked up , they were unable to fab the gpu , and so they bolted on an nvidia gpu to a processor that was never intended to be used they way it was. it worked out in the end , but it was a poor choice its kinda like how Sun kept making sparc unix workstations until like 2007, well past the point where the market had disappeared.

>>105816511
this has to be some kind of bot thread right? this exact post was made in the last one

>>105816670
Was that months ago? It was probably me.

>>105816511
it was at the time, the cell is ridiculously powerful, it was such a good deal that the us airforce bought a bunch of ps3s for a compute cluster.

>>105816680
Compared to what though? Surely there's some evidence/benchmarks of workloads the cell was good at.

>>105816791
the cell had stupid floating point performance for it's cost
however, this was mainly due to gpgpu being limited on gpus in 2006, and the fact the ps3 was a /console/ so it's cost was subsidized by game sales. i.e. the console itself was cheaper than it would normally be if it was sold purely on it's own like a gpu is. that and "otheros" allowing third parties to use the ps3 in such ways ootb (which they later removed illegally since it was a selling point)

>>105816843
or in other words, it was hardware uniquely suited for certain scientific and similar workloads, with an feature to allow it to be programmed for such work, sold below what would be typical market value for such hardware...

>>105816885
>>105816843
>I could mine bitcoin on it, therefore it's good
thanks for reminding me that Hentai@Home ran on PS3s though, what a time the 1st crypto boom was

>>105816158
as many times as sony pays him to cultivate the stab in the back myth, in preparation for the launch of PS7 with thirty 300 mhz riscv cores
/v/ermin will eat it up

>>105816901
as a game console, all the otheros uses were beside the point and doesn't by itself mean it was a good games console, it was probably good marketing if nothing else
but yes, that was a real thing that happened

>>105816976
god that era was so fucking stupid, but so full of hope and the fun kind of jank

bring me back bros

fuck man i just realised the ps3 will be 20 next year
never owned a 7th gen console though. remember going to a convention and playing a ps3 (gears of war i think and some combat flight simulator), but that's it.

>>105815666
>The hard part is writing software that utilizes the SPEs to a high degree to get the most out of it.
moving the goal posts to something else makes you look like a fucking spastic. the topic was:
> why did the cell processer filter many programmers and devs?
not
> uuhhhh SPEs and shieeet
shut the fuck up.

>>105814132 (OP)
STOP WITH THIS FUCKING THREAD ALREADY FFS

>>105817368
mods won't do shit about them for some unknown reason. their changes to stop bots didn't stop anything.

>>105817362
But he's absolutely right and you're a retard, anon.

>>105817376
you should see /twg/, some dude spams in there about literal poo for hundreds of posts at a time now

I'm positive the jannies are doing this themselves

>>105817362
Lol.

>>105817385
wasn't the topic. 2408 games, nobody filtered.

>>105817395
>I'm positive the jannies are doing this themselves
i'm sure they are. anons on another website think one of the admins here (confirmed as working for apple, with docs to prove it lmao) is behind a lot of the apple spam here, and the moderator of pol keeps posting threads to keep the board alive. if you ever seen a thread and it's just one post on that id, it's a mod. when things are so bad that you have to artificially generate traffic, and a lot of that traffic is STILL bot traffic, might as well shred the website and call it a day. they removed thread user counter for a reason, just so they could cover up this shit.

>>105814146
I suggest someone lists all the good games which did the needful.

>>105814132 (OP)
Because it was really bad at running regular code.
It's the reinvention of the wheel of shaders and the PS3 needed an actual real CPU to run game code.
Also the split memory that required you to flood the EIB to have more than 256 MB of texture data.

>>105815594
Why do we have so many worthless ritual faggot posters here?
What do they hope to accomplish doing this over and over again?

>>105815888
>considering 3D on a pc wasn't a thing outside of software renderers in 1994
how do you think they made the game

>>105815888
The SCU of the saturn could be used to do 3D math and very well.

However, it's a VLIW hell architecture.

>>105816143
>No debugging, no profiling
How did they get anything done lol

>>105818469
You can bullshit around writing shit to an array and saving it to the disk, assuming the program still didn't crashed.

>>105818361
that's not a PC

>>105814132 (OP)
>Why did the CELL processor in the PS3 filter so many programmers and devs?

Motherfucker the first devs to work on games for the PS3 couldn't even figure out HOW TO TURN THE THING ON due to how shit the documentation was.

wHy DiD tHe Ps3 fIlTeR sO mAnY dEvS

Try giving them a fucking manual worth a damn for once.

"Here's the black box. Now figure it out dudes."

Fuck off

>>105818731
>couldn't even figure out HOW TO TURN THE THING ON
I can see how such mental giants would be filtered even by something as simple as the cell.

>>105818680
oh sorry it's a woRkStatIoN

>>105818759
It's not simple or complex, its bad.
massive pipeline, no branch prediction, no out of order execution.
I seem emulation people equate an SPE to the Wii CPU when running emulation code, even with the massive clock speed difference.
In game code is less severe, but shit like data compression, script execution, AI, collision tests and all the cell fucking SUCKS.
And again, if you want more than 256MB of texture, you have to swamp the Cell bus with texture data being copied to the VRAM, which limits what you can do with it's CPUs.
A very common optimization on the RPCS3 is just to ignore most Cell code, as it's just post processing shit on the picture.

>PS3 was a nightmare to program for
>everything still ran better back then than any current Unreal Engine 5 game
what the fuck happened?

>>105819076
It didn't.
Frame rates were all over the place in that generation, with for example the cry engine capped at 31 FPS on consoles (not 30).
It's hilarious to see the digitalfountry people having an aneurysm looking at it.

>>105819076
one of the biggest complaints about PS3 ports was low framerates wtf are you talking about

GTA4 ran at literally ~20 fps

>>105814132 (OP)
Sony could have invested in a G80 instead and gotten a real Xbox crusher.

>>105819613
The original PS3 plan didn't involved a GPU at all.
It was just 4 Cells doing software rendering.

>>105819613
almost certainly too early for that, the ps3 was supposed to launch in 2005 like the 360, and was delayed a year, only ati had developed a working unified shader uarch at that time.

>>105814132 (OP)
would something like this with daisy chained RISC chips take off?

>>105822267
All CPUs are RISC nowadays.
But what you describe probably happens in GPUs.

>>105817368
>STOP TALKING ABOUT TECHNOLOGY ON MY TECHNOLOGY BOARD

>>105818830
is your shift key broken or something?

>>105822358
Always has been, even the 8086 was a RISC core that emulated x86.

why did sony turn their back on the tried and true mips architecture

>>105822959
By the late 90s, MIPS was focusing on the embedded market.

The Cell processor was the physical incarnation of the memeflops bullshit.
100% designed to create the biggest flop number rather than being designed to be any good at anything in particular.
It would be a lot better if two of the SPEs were 4 stage long cores dedicated to do branching shit.
You would be able to use em to for example reduce the data that actually gets sent to be processed and do shit like AI and data decompression.
The memeflop number would decrease, but the actual performance would go way up.

>>105819076
no

>>105822525
We've had this same thread, same discussions, same information, same conclusions, for dozens of threads over the last year or two.

>>105823361
You've just described half the threads on the damned board at any given time.

>>105822525
>>105823363
>butt hurt dense lad wants to see reposts after reposts after reposts!
Kek!

I hope you realize that reposting the same threads/posts over and over again is considered as spamming!

>>105823604
I'm not the one making them, I just enjoy discussion (even if repetitive) of actual technology on the technology board

>>105816976
a fellow navy college ps3 cluster owner

>>105819035
>The local store does not operate like a conventional CPU cache since it is neither transparent to software nor does it contain hardware structures that predict which data to load.

Don't other issues pop up like spe's doing duplicate work because they're unaware of the other's local store if you're not careful?
So what do you have to do on this thing?
Predict when you need to do I/O preemptively to use the async stores?
Be aware of scheduling and execution order for each SPE when you have stuff queued?
Write something to do the caching for you since there's no hardware level?
DMA requests manually for non-local storage?

>>105823997
Yes

>instead of admitting that I fell for sony marketing i'm going to pretend that this 20 year old cpu is the most powerful chip in existence but nobody on the planet can possibly figure out how to use it
What mental illness causes this?

>>105824101
It was a good idea at a bad time. GPGPU alternative right as GPGPU took off

>>105824181
Sounds more like a good idea executed poorly, it might have worked well if they had spent more time developing and testing it.

>>105824195
Probably a bit of both. Execution was garbage but so was the timing

>>105818344
the site's moderators (that spend all day on discord talking shit about users and sharing unfunny memes with each other) refuse to do their jobs and clean up this board. they are at a point now where any traffic generated by bots and spammers is good. this is why they removed the user counter in threads, just so they could allow bots to generate artificial traffic to make it seem like this website has users. it's like they completely forget that 4stats dot io exists.

posting in retarded larper thread.
the real issue is the frequency. IBM failed to do their fucking jobs and do proper timing design to get it to a stable and usable clock rate. all the other issues stem from this.
the SPEs were GPU replacements, not CPUs, and were meant to provide more flexibility than the former. the poor clocks meant sony had to include another gpu, but instead of making ibm give them an all ppe chip instead to get something out of their investment, they just bought it as-is like cucks, then went to nvidia and cucked out even harder there somehow. it is a valuable lesson for sony and the mutt simps they call japs to learn, to never trust the americans who treat anyone who's not an anglo as less than dirt.
VLIW gets a bad rep because of this and itanium too, but the latter is mostly because PC is an x86 monopoly and intel made no effort whatsoever to break that. then you had compiler devs with excessive egos for their IQ and the cpu-side retards who thought you could adapt a compiler written for stupid basic turing machine tier architectures to vliw in like a week. and instead of seeing the error of their ways both sides threw their own tantrums.

>>105824542
>instead of making ibm give them an all ppe chip instead to get something out of their investment
The PPE is just a PowerPC core but extra dogshit, all of the investment was in the SPEs. Making IBM give them an all PPE chip would still return essentially zero

>>105824607
yeah, but at the very least they're not getting negative returns by cramming that garbage into usable die space.

>>105824542
i dunno , ibm released power 6 around the same time, and it was a "boil the ocean design" that was in-order as well and ran up to 5ghz, its possible that the PPE /cell drew simply too much power, and was downclocked as a result, won't really know unless someone manages to figure out clock control.

>>105824697
It was targeted to be 4GHz and released at 3.2

>>105824697
if it isn't the ppe then it's probably the spe
I also assume you can improve thermals (reduce voltage) by having better timings and vice versa, so they are equivalent up to a certain point.
if not design then it's a fab issue, which would again be entirely the fault of ibm.

>>105824715
Yeah If thats true , I'm gonna go with power consumption, those types of designs clock to the moon easily, but they might need 150-200w to do it, well beyond what sony would tolerate.

>>105824747
the spes are pretty simple, so they shouldn't have too hard of a time clocking up. usually if you are frequency bound its because you have a few complicated stages, nothing in the cell iis like that lmao

>>105824753
did the spes even have thermal issues? if not why didn't they clock them higher than the ppe?

>>105824780
>>105824766

>>105824780
the 90nm that they were originally fabbed on was not particularly efficient, launch day ps3 consumes about 200w and they run hot, probably the highest ever up to that point. might not seem like much but you have to consider that the ps3 was already insanely expensive to make, its not really surprising that they didn't opt for a bigger heatsink and power supply, frequency/power curves are exponential, so its very possible that doing 4ghz woud have pushed it over 300w.

>>105824780
the spes are plenty fast already, the only thing that would have maybe saved the cell is if the PPE was significantly faster, would have helped low quality ports alot.

>>105824542
> the SPEs were GPU replacements, not CPUs, and were meant to provide more flexibility than the former. the poor clocks meant sony had to include another gpu
There was always going to be a GPU. We have a terminology issue. GPU used to refer to the device which rendered the polygons. But later GPUs combined compute with rendering. In another thread someone posted that sony could not make their custom GPU work, forcing them to go to Nvidia. Since the Nvidia GPU has its own compute, this left the SPEs without their original purpose.

>>105818344
Funnily enough (I want to kill myself), lack of moderation.

>>105820171
the xbox 360 shipped with a unified shader arch gpu, but not because it was standard. it was one of if not the first consumer device to have unified shaders
pc and xbox were going one way, and the ps3 was following on instead from where the ps2 left off, which was a very different way

>>105824715
https://arstechnica.com/features/2005/06/xbox360-2/

>The PPE's pipeline is 21 stages deep, the same number of pipeline stages as the Pentium 4 Northwood core. This 21-stage pipeline looks like a relic of an age prior to the recent multicore revolution, when clockspeed was still king.

well there you have it , definitely underclocked.
the cell's PPE and the xenon are very similar. so this pretty well confirms it. was likely intended for 4ghz +

>>105824542
>>105824835
My assumption is that the original architecture was envisioned as something similar to the PS2, where you have two VPUs that can either receive instructions on-demand or you can load small programs into and they'll run in parallel with the main CPU, with one connected directly to the GPU which was leveraged for computing effects and shit.
There would have been a GPU, but most of the heavy lifting would have been done by the SPEs.
Though PS2 is MIPS and PS3 PPC, the architecture does seem like it was supposed to be like the PS2 but more.

>>105825221
yea, if you want to know what the original plan for the CELL was, you need only look at how the PS2 is architected
even the ps2 was not easy to make full use of, they surely believed they could do it again, and in the context of the ps2, the ps3 doesn't sound like such a bad idea
https://youtu.be/iEQLbbyToxw

>>105825221
Using the same design philosophy does not mean being literally the same thing. The fact that they use a different cpu architecture does not matter.

>>105824831
evidently not fast enough to avoid the need for an nvidia gpu
>>105824824
yeah if the thermals are an issue then it's an issue, no solving with pumping more watts. guess it's a fab thing then
>>105824835
fair, but they went from amd to nvidia because of the cell's failure, not in spite of it.

>>105814132 (OP)
Because it became obsolete you stupid shit head, do you use a fucking 386 with msdos or something? Everything fucking day 24/7 we have to put up with these stupid fucking trolls threads that if organic would have died 10 years ago when they stopped being funny.

>>105827323
this nigga hasn't developed a theory of mind yet

>>105827575
QRD?

>>105825208
These old chips with long pipelines and no branch prediction were actually the shit. They forced good programming, it's the whole reason why devs found it hard to program for the PS3. No branch prediction, long pipelines, and parallel processing by using the SPEs along the GPU. Sexy. Incredible. Alien. I want more like this. I wish processors followed this kind of logic.

It still is the strongest cpu ever. But developers don't know how to use it

If you create this thread tomorrow, i could provide some documentation to prove it

Cell was a succes for IMB: scammed Sony into paying billions in R&D costs, lied about issues during development, and then sold a design with 3 times the central core and no braindead SPE's. It was basically america cucking the Japs while stealing their money.

And then nvidia gave them a cum-covered roasty as a very overpriced GPU, while Microsoft got a tight virgin jailbait from AMD. That gen was really Sony getting fucked in the ass by everyone.

>>105824424
True

>>105818293
ive heard that the last of us managed to use both the spe and rsx to render graphics at the same time to cope with how underpowered the rsx gpu was

>>105828539
fucking grim

>>105828539
And they completely deserved it

>>105827892
Yes, there’s a lot of things on PC architectures that could be fixed using other methods.

My opinion is that the procesor should “notice” these things, like pipeline stalls, missed branch prediction, etc, and have a runtime that re-orders or re-writes the code based on what the processor sees.

Kind of like what java could theoretically do with hotspot.

>>105828032
> the strongest cpu ever
I heard it was as fast as the ps4 apu in compute.
It was kind of like if they started selling F1 cars at regular dealerships. Maybe not extremely popular for getting groceries.
My opinion is they liked the PPC because it was weird and helped prevent piracy and forced lock-in and probably prevents emulation.

>>105829025
>I heard it was as fast as the ps4 apu in compute.
Not that hard, as the Jaguar was a shitty design for very cheap netbook. It's performance was equal to the Pentium 4 clock per clock (13 years old at the time)

>>105829197
Had 8 cores tho

>>105828622
That's basically what you were supposed to do. Use the CELL SPEs to offload heavier/more specialised graphics tasks from the rather basic RSX G70 core. Basically every high profile PS3 exclusive does this.

>>105818830
correct, that’s a workstation. A personal computer is a consumer product, indigo never marketed their computers to the general public, even the cheapest one cost thousands of dollars.

>>105829316
It's fucking HARD bro

>>105829338
Yeah I never said it was easy lol. For what it's worth the games that do it well still look pretty good today (720p and stodgy frame rates aside). I played through Killzone 2 recently and just started 3 and I'm quite impressed with what they managed to shove on screen with such arcane bullshit going on.

>>105828539
Pretty sure IBM scammed Microsoft too, the PPE was utter trash.

>>105829487
I think that if the core had been clocked to 4-5ghz like power6 it probably wouldnt have been that bad. seems like ibm basically used sony and microsoft for r&d funding of their own products, since power 6 isnt nearly as bad and appears similar in design (long pipeline , inorder architecture)

>>105829307
They were literally cell phone cores designed at AMD's nadir. Even at the time there were netbooks with as much compute power as the PS4.

>>105830768
>150W dual core
How exactly were they supposed to fit that into a console?

If any of you actually saw dev diaries of PS3 era games you'd know that Cell was a massive failure. Best looking games on the PS3 just used the SPEs for shit like anti aliasing and not game logic e.g. The Last of Us. Xbox 360 just did that on the GPU which was faster and more straightforward.
Cell was the same kind of misguided design as Intel's Knights Landing. Why would you ever want to do on a CPU what the GPU can do faster? Instead of making a needlessly large multi purpose computing unit, make better standards so your compute doesn't waste silicon to handle 1 in 100 edge cases. Just avoid them altogether.

>>105829487
Yeah but MS paid pennies for it compared to what they'd have to pay Intel while Sony funded the entire thing for IBM and MS.

>>105830905
not really what I was getting at, the PPE in the ps3 and xbox360 appears to be a precursor to power6, and can probably hit those frequencies , at the cost of power consumption.