Xenon core = Cell core?

[Danji]

There is a rumor, or perhaps more accurately, a hypothesis held by some of the workers at Arstechnica that the 64-bit PPC core in the Xenon (there are three) are the same as the 64-bit PPC core in the Cell.
(speaking about the Cell PPC core)All we know about it is that it's a static execution, two-issue core with an AltiVec/VMX unit, FPU and integer units, and a split L1 cache (32K code/32K data). I and others have speculated that this lack of disclosure might indicate that IBM is using that same PPC core in the XBox2's CPU, codenamed Xenon.
link (http://arstechnica.com/news.ars/post/20050309-4686.html)

This could possibly make the porting of Xbox Next games to the PS3 easier than previously suspected on the mere fact that if the PS3 is the Broadband Engine it would have 4 of that core and 32 APU's on top of it. Communication would be slower between the four cores on the PS3 (only slightly though) than that on the Xbox Next but there would be that fourth core, 32 APU's, and the extra capacity and celerity (that's speed) of the memory would make porting to the PS3 a smooth experience that could result in improvements. However, things that must be taken into consideration are games for the XBN that aren't hard-coded, games that utilize XNA too much, games that use DirectX instead of OpenGL (does the XBN support OpenGL??) and that transition from the XBN to the PS3 obviously isn't as easy as I make it sound. However, I believe that it will stand true that a transition of a game from the XBN to the PS3 would be smoother than any port to the PS2 has ever been.

[cpiasminc]

I'd have to wager that they would be highly similar, but not really "the same" per se. I'd expect Xenon's to have a little more aggressive on the IPC-saving features like branch prediction and speculative prefetches and so on. Namely because it wouldn't have 8 APUs attached to it after all. It would need to have a little bit better practical single-thread performance than Cell's PPE which would basically be a dedicated scheduler.

I don't doubt that they're pretty much based on the same idea since going multicore inside a console is bound to raise power consumption concerns, and so as long as there's existing IP for in-order two-way SMT cores that can clock really high and not consume a whole lot of power... well, it's certainly bound to save some time. However, if they were almost entirely identical, I would be lost as to why MS would grade down to 3 or 3.5 GHz, when the core has already been shown to easily clock to over 4 GHz without consuming that much power even at the 90nm node. Unless they're planning to keep saying 3 GHz, 3 GHz, and then when the beta hardware comes out, they say "Surprise! It's actually 4.5 GHz!! Fooled ya' didn't we?"

[Lumine]

Aren't PPC cores really hot and power hungry? Atleast that was my understanding. That could be why they bumped the clockspeed down to 3ghz. Or maybe that would explain why alot of people were thinking the Xenon was gonna have 6 cores and now it only looks like 4.

[Danji]

Aren't PPC cores really hot and power hungry? Atleast that was my understanding. That could be why they bumped the clockspeed down to 3ghz. Or maybe that would explain why alot of people were thinking the Xenon was gonna have 6 cores and now it only looks like 4.
Relative to the Pentium 4 it's not very hot or power consumptive. However since there are 3 cores they might need to clock down to keep it from getting hot.

[cpiasminc]

Aren't PPC cores really hot and power hungry? Atleast that was my understanding.
Far from it. G5 cores draw around 1/3 the power of P4 cores. The difference is that G5 core is also a lot smaller, so the actual thermal density is not that much lower, which means actual core temperatures will still be rather high.

However, you have to realize that "PPC" is just an instruction set. There's no rule that says PPC ISA needs to be implemented using the same hardware that IBM puts into Apple boxes. There's really no relation whatsoever between the console CPUs in question and a G5. They just share the ISA.

I'm just saying that if the hardware behind Xenon's cores was the same as Cell's PPE cores, there shouldn't be anything holding the clock speed back. So there's got to be something else unique about Xenon's CPU cores unless MS is hiding something. I know that every time we at my office ask MS when we'll have beta hardware or when we could at least get something closer to it, they just say "oh, we don't know yet..." or "we have no clue..." In the back of my mind, I just think "Yeah, like hell you don't know. You're just not telling us."