IBM backstabbing sony and neglected Apple ? (long read)

[ADEX]

So basically, according to the article, the the Xbox360 CPU is a derivative of the Cell power core?

The Power core is derived from an earlier research processor originally done in 1997/98 and updated in 2000 called GuTS.

The control and Integer processing portions of the 2 cores appear to be identical.

The important part in the XBox is the VMX units, the Cache and the interconnect between them. These parts are either not present in the Cell's PPE or are different (e.g. the PPE has standard VMX, the XBox has an enhanced version).

They both have cache locking but this is also present in other PowerPCs (G3, G4).

The PPE may also appear in a separate IBM product (embedded processor).

 

[Fox5]

Well since Cell research began ~2000 and XeCPU research began in 2003 (or is it 2004?) I'm thinking that it probably wasn't the other way around.

Not saying it's *either* way really, but I think we know which way it's not for sure.

EDIT: You could be right on the OOE sacrifice simply being the logical choice due to die space Mefisutoferesu. I guess we'll need another interview or two from the MS side to be sure what exactly prompted that move on their part.

Maybe IBM sacrificed OOE because it wasn't their biggest strength. I don't think any apple cpus match the modern AMD or Intel cpus in OOE, so rather than try to play catch up, maybe IBM just said "Let's just go with what we're good at and not waste time or effort." (on the other hand, I believe the transistor count of the G5 is much lower than an athlon 64 or prescott, maybe even lower than dothan, but the transistor space on the console cpus that could have been used to add OOE of dubious performance benefit was instead used for extra cores)

 

[Brimstone]

Maybe IBM sacrificed OOE because it wasn't their biggest strength. I don't think any apple cpus match the modern AMD or Intel cpus in OOE, so rather than try to play catch up, maybe IBM just said "Let's just go with what we're good at and not waste time or effort." (on the other hand, I believe the transistor count of the G5 is much lower than an athlon 64 or prescott, maybe even lower than dothan, but the transistor space on the console cpus that could have been used to add OOE of dubious performance benefit was instead used for extra cores)

Cutting out OOE (reducing transistor count) and making the core simpler results in the cpu core consuming less power, therefore reducing heat.

 

[one]

He hasn't designed any part of a microprocessor in his entire career so of course he cannot be head engineer of a microprocessor project. That would be suicide.

He once designed a microprocessor, albeit not quite advanced one like EE or Cell.

This is a page in the alumini association at his university (machine translation) with Kutaragi's career and it contains "LCD TV, microprocessor, and LSI design" in it as you see in the page.

He's a digital engineer from the beginning of his career, which was not very honorable in Sony back then.
http://www.eetimes.com/futureofsemis/showArticle.jhtml?articleId=18301417

But Kutaragi's work did not always generate such high-profile praise. Long before he became chairman of Sony Computer Entertainment America, Kutaragi's first design efforts were as obscure as one could imagine — developing dot-matrix LCDs and code for stripped-down controllers.

The 80 x 100 matrix LCD Kutaragi designed as a young engineer at Sony never got off the ground. "It was too early," he said. But the assemblers, debuggers and compilers he helped write for proprietary 4- and 8-bit Sony controllers was more of a success, albeit only inside the walls of Sony's engineering labs.

Kutaragi describes those processors as something less ambitious than a full-blown microcontroller, since they were streamlined and optimized around specific Sony systems such as tape decks and videotape players. The code itself ran on the CP/M operating system. "There were no Microsoft tools," he said. "I realized it was important to have integrated tools to help optimize those cores."

 

[Fox5]

Cutting out OOE (reducing transistor count) and making the core simpler results in the cpu core consuming less power, therefore reducing heat.

Does the triple core x360 cpu really produce less heat than say a single or dual core opteron?(not sure which it's closer to in die size)

 

[Mefisutoferesu]

Well, I believe it's been said that the XeCPU eats up about 85Watts. Being simplistic that's 28Watts per 3.2GHz core. That's a good deal cooler than any processor at similar speeds. Whether that's solely due to the removal OOOE, I dunno, but I'm betting it has something to do with it.

 

[Fox5]

Well, I believe it's been said that the XeCPU eats up about 85Watts. Being simplistic that's 28Watts per 3.2GHz core. That's a good deal cooler than any processor at similar speeds. Whether that's solely due to the removal OOOE, I dunno, but I'm betting it has something to do with it.

It's a 90nm core right? Wonder how the performance per core compares to the 90nm P-M's...well it destroys them in flops.
There's also 2.2ghz Turion cpus to compare to. If that 3.2ghz XeCPU only performed on par with a P4, it wouldn't look so good after all. (but it's strengths are in completely different areas, so it will lose and beat a p4 depending on the test)