Wii U hardware discussion and investigation *rename

[liolio]

TSMC can fab eDRAM as well. It was actually cheaper for Microsoft as NEC was charging more IIRC.

I didn't know that but I guess they are in the same situation as NEC this process has to be lagging, the best TSMC or NEC can provide is still 65nm? (I asked because even if they are multiple providers the option GPU embedding some edram is highly unlikely for this reason).

 

[The_legend_of_drtre]

That was a very good read.:smile:

I actually did my own little analysis awhile back.

http://www.youtube.com/watch?v=l-Bng1LF-_Y&feature=youtube_gdata_player

Before that...

http://www.youtube.com/watch?v=W6FQCY_1OAY&feature=youtube_gdata_player

 

[Espresso]

We know that the CPU will include "lot of EDRAM" (they did not state if it's all cache by the way)

This was a bit of an exaggeration. It is only "a lot of EDRAM" relative to amount of EDRAM in the previous Nintendo IBM CPU, and it's all cache. Both the L1 and L2 in the previous generation were SRAM. Any amount is "a lot" compared to 0.

The target for the CPU should be somewhere around or over the X360’s Xenon, which means at least triple core and a comparable clock. The rumour also has it that the mysterious amount of eDRAM available on the CPU wouldn’t only be measured by words like “a lot” but also by number. That number is 16.

That number is more than 300% over the actual total.

Being a POWER 7 derivative, the CPU would also be capable of four-way simultaneous multithreading (SMT), whereas the PPC cores of the PlayStation 3 (Cell) and Xbox 360 (Xenon) are only capable of dual-way SMT. If the CPU has indeed 3 cores then, we’d be talking about 12 simultaneous hardware threads, putting an end to Cell’s record of 8 (2 SMT PPC + 6 usable SPUs, one is OS only) in the console space. Although Cell would keep its crown when it comes to concurrent simultaneous hardware threads, with its 7 way heterogeneous SMP setup (1 PPC + 6 SPUs).

It's not a Power 7 derivative. It's directly descended from the CPU core in the Wii, there are just more of them and they are clocked a little faster. It does come up about the same as Xenon for processing power, but the clock is much, much closer to Wii than X360.

 

[stiftl]

That number is more than 300% over the actual total.

So around 4MB then?

It's not a Power 7 derivative. It's directly descended from the CPU core in the Wii, there are just more of them and they are clocked a little faster. It does come up about the same as Xenon for processing power, but the clock is much, much closer to Wii than X360.

WTF? Do you really know something or is this just a wild guess? Much, much closer to Wii than X360 would mean ~1,5GHz max, are you sure about this? How could this be equal performance wise to Xenon? 8 cores?

 

[Entropy]

WTF? Do you really know something or is this just a wild guess? Much, much closer to Wii than X360 would mean ~1,5GHz max, are you sure about this? How could this be equal performance wise to Xenon? 8 cores?

The Power7 rumour never made much sense - why would Nintendo, even with a cut down version, spend all that die area on the CPU? The IBM A2, or, given Espressos claim about being a direct descendant of above, a version of the PPC 470, makes a lot more sense. Both have been fabbed using IBMs 45nm SOI.

It could easily be comparable to the Xenon by for instance having more cache, higher IPC through OoO execution, better branch prediction and lower mispredict penalties, faster interconnect to the GPU and a higher performance main memory subsystem.

Even without modification, the PPC 476 PLB6 supports 8-processor systems, and the A2 goes beyond that. Of course, a new design is more flexible. Not that it's needed, a four way design based on either of these cores should definitely be enough to provide performance comparable to the Xenon. Depending on where your bottlenecks are, a two way design could probably be said to do the job. The 476 in particular looks to be a nice foundation for a low cost, well performing CPU.

Edit: The 1Mbit eDRAM macros used in the POWER7 are 0.24mm2 (with a 1.05V supply and 1.7ns/1.35 cycle and access times). That is, 4MByte of eDRAM on the Nintendo CPU, on the same 45nm process, would consume just under 10mm2. Could make sense.

 

[Urian]

I agree that the 45nm CPU from IBM that has more numbers to go inside the Wii U is the PowerPC 476FP and the reason is the size of the box and power consumption, the second possibility is a derivative of the one used in the Xbox 360.

This is an old pdf about the processor but something curious is that Nintendo puts a Wii in Page 4 as a possible application for this architecture:

https://www-01.ibm.com/chips/techlib/techlib.nsf/techdocs/D393643EC6B662E78525763200547AED/$file/476fp_wp_04_07_2011.pdf

The question is... how many cores?

 

[stiftl]

The question is... how many cores?

Could a quad core PowerPC 476FP derivat at 1458 MHz (2xWii) with four way SMT and 4MB eDRAM do the deal?

 

[Entropy]

Could a quad core at 1458 MHz (2xWii) PowerPC 476FP derivat with four way SMT and 4MB eDRAM do the deal?

Definitely. Remember that IBM can modify it freely to suit its customers. And that the overall environment of the cores is likely to be an improvement over what is accessible to the Xenos. Of course, with new cores in a new environment, it is never going to be possible to say something like "50% faster". It will depend strongly on the particular code you are running. What I like about the 476 is that it seems to be a straightforward and, for lack of a better word, robust processor design with no obvious performance cliffs to fall off. Give it a sympathetic environment, and it should do nicely. I'll raise a question mark about SIMD implementation though, if that's needed.

 

[Teasy]

It's not a Power 7 derivative. It's directly descended from the CPU core in the Wii, there are just more of them and they are clocked a little faster. It does come up about the same as Xenon for processing power, but the clock is much, much closer to Wii than X360.

I seriously doubt that.

 

[Entropy]

I seriously doubt that.

Of course you do, given the anonymous source and his/her lack of posting history.
But it makes more sense than the other claim floating around, that it would be a Power7 derivative.
So - would you care to elaborate?

 

[Rangers]

lherre drops a lot of hints on neogaf and he knows what he's talking about i think. he is a dev. Ubisoft I imagine.

he says WII U CPU is triple core.

http://www.neogaf.com/forum/showpost.php?p=29784657&postcount=4165

This is pretty exciting, I definitely believe these new infos from espresso and lherre are likely to be true and it's the first hard infos.

It's probably a triple core higher clocked Wii CPU.

 

[Gubbi]

I agree that the 45nm CPU from IBM that has more numbers to go inside the Wii U is the PowerPC 476FP and the reason is the size of the box and power consumption, the second possibility is a derivative of the one used in the Xbox 360.

Something based on 476FP seems likely. I'd expect four cores. Four 4-way isssue OoOe cores running at 1.6GHz would match the cpu in the 360.

Cheers

 

[Entropy]

lherre drops a lot of hints on neogaf and he knows what he's talking about i think. he is a dev. Ubisoft I imagine.

he says WII U CPU is triple core

http://www.neogaf.com/forum/showpost.php?p=29784657&postcount=4165

Frankly, barring IP ownership conflicts, the easiest way out for Nintendo would be to simply use an updated Xenon. But even if the "triple core" statement is true, there is still ambiguity. For instance, does that mean that there are only three cores, or that there are only three cores available to the application programmer? (Note that an 8-core Power7 is 567 mm2 on a 45nm SOI process!)

All speculation I have seen about the WiiU have been had high estimations of CPU performance/complexity, and therefore cost, compared to the speculation regarding the graphics processor. There is a dissonance there that doesn't sit well with me.

 

[Rangers]

Frankly, barring IP ownership conflicts, the easiest way out for Nintendo would be to simply use an updated Xenon. But even if the "triple core" statement is true, there is still ambiguity. For instance, does that mean that there are only three cores, or that there are only three cores available to the application programmer? (Note that an 8-core Power7 is 567 mm2 on a 45nm SOI process!)

All speculation I have seen about the WiiU have been had high estimations of CPU performance/complexity, and therefore cost, compared to the speculation regarding the graphics processor. There is a dissonance there that doesn't sit well with me.

Why could it not be three higher clocked Wii CPU cores as Espresso stated?

Also, what speculation are you referring to exactly?

 

[Entropy]

Why could it not be three higher clocked Wii CPU cores as Espresso stated?
Also, what speculation are you referring to exactly?

He did not state that it would be higher clocked Wii cores, he explicitly said that they were "descended from...". Now, if you look at the lineage of CPUs the Gekko/Broadway belong to, the 476FP is the latest of that line, and has variants fabbed on the same process that the Wii U CPU has been said by IBM to be targeted for. Of course, this is IBM IP, and they are free to modify it anyway they please to suit a customer, but it really doesn't make sense to spend a lot of effort to try to recast the Gekko in MP form, when in a sense they have already done that job, and have perfectly fine compatible solutions on the shelf. Better and cheaper to take one of these evolved and more modular designs and ask how it needs to be tweaked to suit the customer.

Most speculation I've seen regarding the WiiU CPU has focussed on Power7, along with less frequent Cell/BBE references, and Xenon derivatives. There has been some targeting the IBM A2 as well. I think that the Power7, unless cut down to where it effectively is a new core, can be safely dismissed and that the referral to Power7 technology implied the eDRAM rather than the processor cores themselves. As I said, out of these, an updated Xenon would probably be the simplest road to parity+ with the XBOX360 for easy porting, but I'm not aware if Microsoft has any IP rights that would cause trouble, nor do I know if Nintendo has other requirements in terms of cost or power dissipation that makes this a less attractive option. The IBM press release contains statements such as "all-new, Power-based microprocessor" "energy saving" that doesn't really smell like Xenon Mk2. A higher IPC, lower clocked CPU utilizing a better cache/memory subsystem could probably be both smaller and draw a bit less power.

 

[Teasy]

Of course you do, given the anonymous source and his/her lack of posting history.
But it makes more sense than the other claim floating around, that it would be a Power7 derivative.
So - would you care to elaborate?

I've only heard some vague info on mhz and cache. But what I heard was 3.5ghz in the last development kit. Maybe that's just a stopgap until they come out with the real CPU but even then I can't see a multicore Broadway derivative running a little faster than 750mhz being that CPU.

 

[Entropy]

I've only heard some vague info on mhz and cache. But what I heard was 3.5ghz in the last development kit. Maybe that's just a stopgap until they come out with the real CPU but even then I can't see a multicore Broadway derivative running a little faster than 750mhz being that CPU.

No CPU from that family of processors would hit those kinds of clocks, that's for sure.
Rather, it sounds just like the PPE (derivatives) we already know well from consoles. Regardless of core, I should think it gets more cache, faster connection to the GPU, and higher main memory bandwidth than its predecessors.

 

[Espresso]

Why could it not be three higher clocked Wii CPU cores as Espresso stated?

A modified and speed-bumped version of the Wii core is the easiest path for the CPU side of the backwards compatibility.

 

[The_legend_of_drtre]

Broadway can't even be clocked over 900mhz so...no. Why would they even attempt to overclock and multicore a chip like that when it'd be easier and cheaper to use a more recent chip?

derivative of Wii cpu could mean anything. The 360 cpu is "based on" the Gc's Cpu which is based on a Ibm chip from 1998...it means nothing.

Oh and wassup Teasy... long time.

 

[The_legend_of_drtre]

A modified and speed-bumped version of the Wii core is the easiest path for the CPU side of the backwards compatibility.

If the CPU were really an overclocked Wii CPU then Wii U would also have 100% BC with the GAMECUBE.

Game over good Sir.