Go back and read the posts. My "rebuttal" came before you made that post. The post I rebutted didn't say that.

I also find it odd that you claim that Nintendo not using 32 nm process (a process that IBM themselves are not using for the same chip) is an example of Nintendo not using cutting edge technology despite the fact that Power7 is the most cutting CPU on the planet.

And I am not saying there are no reasons for IBM not to do so, and for Nintendo not to do so. Quite the opposite, I am saying that there are reasons not to do so, and that those reasons are why Nintendo chose 45nm.

You seem to think that IBM not having released a POWER7 processor at 32nm today means that Nintendo could not have chosen to contract them to design an POWER7 derived processorat 32nm for next year, but why? Designing processors takes a lot of time so your choice is not just the processes available when you start your design but also the processes that are scheduled to be available when production starts. 32nm is available for design now, but if Nintendo believes it will not be mature enough when they need to start production and do not want to risk delaying launch until it is then they would choose to ask IBM to design it at 45nm.

I really don't understand why you can't get it: 32nm was an option for Nintendo's CPU, they didn't take it. This shows they tend to be conservatives with their process, This means we can expect them to be conservative with their GPU process choice too. This has implications on the power their GPU can have.

If you are not happy with this example, take the example of the Wii CPU and GPU. Even though both Sony and MS have transitioned from 90nm to 65nm and then to 45nm, the Wii still uses 90nm (at least I cannot find evidence otherwise, and I looked), again showing Nintendo's conservative nature when it comes to process technology.

I am sure Nintendo would love to have both their CPU and GPU at 32nm as it would allow to make them more powerful than at 45/40nm, but what they want and what is realistic is not necessarily the same thing.

I already told you that the Wii and GC CPU's use the same process as their corresponding IBM counterparts which also never received a processer drop.

Further, if Nintendo could have IBM set up the CPU for 32 nm by next year, why doesn't IBM have that on their Power7 roadmap themselves?   This is my point.  IBM currently has no publicly announced plans to bring the Power7 CPUs down to 45 nm.  Normally, they'd be nuts as hell not to do that so there must be one hell of a very good reason they don't.  And I'm willing to speculate that whatever that reason is is the same reason the Wii U's CPU is also listed for 45 nm processing.  Might have soemthing to do with high-k metal gates...go check it out.

And just as you stated the Cell and and Xenon have shrunk, why can't the Wii U's CPU shrink later in life?   Cell was not on the smallest possible process upon launch either and why isn't the Cell on 32 nm now?

Yes it did, from the post you rebutted to: "Anyway, this was just an example of Nintendo not using cutting edge technology. A good reason for not doing so is to avoid the yield ramp up problems typical of new processes. "

Of course, if you do not read my posts I cannot expect you to understand them.

As for Power7 being cutting edge, it is, but I didn't say cutting edge, I said mature, and it it that too, having been successfully deployed at 45nm for more than a year. IBM's 32nm process is also cutting edge but is not as mature as POWER7 and won't be for a while. From the look of things Nintendo doesn't seem to think that it is likely to be mature when they need it to be.

@Sri Lumpa

I guess the use of 45nm is more a decision from IBM than Nintendo... like I said the 32nm SOI is new and caused some problems to AMD/GF... so it's not mature yet.

But for 2012 the 32nm SOI could be a better option... so the project could be changed.

Sorry, I missed that post.

1. You only told half of what i did. You told about the Wii part but did not compare it to the PS3/360 part. The closest IBM counterpart to the PS3 360 part would be 2007's POWER6 as it is also an in-order CPU with high frequency and it would be foolish to think that IBM did not apply whatever lessons they learn from the PS3's PPE design to the POWER6 design. The closest IBM counterpart did not receive a 45nm shrink whereas the Cell and Xenon did.

2. IBM is keeping quite mum about their roadmap in general. All we know is that there is a POWER7+ coming sometime in 2011 or 2012 and that they are working on POWER8. The fact that IBM does not talk about their roadmap does not mean that they are not working on future power processors at 32nm, whether they be power7, power7+ or power8.

3. None of which disagrees with my contention that targeting 32nm for next year was a choice Nintendo had that they didn't take. Giving reasons while it is better to be conservative and choose 45nm does not undermine my point, it supports it.

4. I never said that it couldn't shrink later in life. I only said that Nintendo was being conservative by choosing to design at 45nm today for a chip to be produced next year. When the 32nm process is more mature I am sure that Nintendo will consider transitioning to it and if they think it makes sense to do so they will, if they think it doesn't make sense, they won't.

5. Maybe because the yield are too poor still, maybe because shrinking a die takes time and they haven't finished the work yet, maybe because the cost savings of manufacturing at 32nm do not justify the cost of shrinking it. Take your pick, there are plenty of reasons, and none of them mean that Sony does not have the choice to do so.

To put it another way, just because you can do something (target your design at 32nm) does not mean that it is the best course of action. Your argument is that it can't be done, mine is that it can be done but isn't the best course of action (or rather that Nintendo doesn't think it is). You prop your argument with possible reasons why it is not the best course of action and that such a chip not existing today means it cannot be in design stage today.

The former reinforce my argument and the later is asinine as the whole point of Nintendo paying IBM to design a custom chip is to have a chip that doesn't exist today. It is the equivalent of saying, in 2004, that the next Microsoft console (Xbox 360) cannot have a 3-core in-order PowerPC running at 3.2GHz because IBM does not produce any 3-core in-order PowerPC nor do they produce PowerPC with a frequency as high as 3.2GHz (at that time of course).

Actually that would have made more sense than your argument, because we didn't know that IBM was working on such a chip, whereas we know that both POWER7 architecture and 32nm process work, so the only work remaining is to port the former to the latter.

The Cell is its own CPU line of products and is independent of what happens to Power6. Gekko and Broadway are not independent processors as they are modified PowerPC 750's. You also forget that Sony and Toshiba also manufacture the Cell processor and are not wholly dependent on IBM's East Fishkill fab plant.


Look man, right now, 45 nm is the process the Power7 comes on. So Nintendo and IBM are not going to announce the CPU on a process not yet ready. If they move to 32 nm later on, they'll announce it. But for now, it's a 45 nm process. Given that they haven't announced Power7 or Power7+ on 32 nm, they sure enough aren't going to make that announcement just for Nintendo's CPU.

I think the focus on the manufacturing process of the CPU is (kind of) pointless ...

Most games are not (really) CPU limited at this point in time, and if IBM/Nintendo believe they can get the performance out of the CPU they want with energy consumption/heat they think is adequate and a yield/cost to manufacture that is reasonable they might not see the need to move from 45nm to 32nm.

This (of course) doesn't imply anything about what Nintendo and AMD/ATI will do with the GPU though ...

So does Viper1, but I guess the opposite.

Like you said 32nm SOI should be mature in 2012 (at least a lot more than now), but when in 2012? If it is likely to be mature at the beginning of the year it would make more sense for Nintendo to ask IBM to target their design for the Wii U's CPU at that process, but if it is likely to be mature later in the year, or if you want to play it safe (i.e. be conservative in your choice of process technology, my point) it makes more sense to have the design targeted at 45nm. In either case the 32nm process was available to Nintendo as a potential target for their CPU.

1.Both the Cell and Broadway are custom designs. That one was bases on a preexisting design and another wasn't is irrelevant. 

2. And Ms used Chartered to manufactured its PPC. That Nintendo chose to use IBM for both the design and the manufacture does not mean they were wholly dependent to IBM, just that they had no need to go elsewhere. If Nintendo had thought there was value in funding a sub-90nm shrink for Broadway, not only IBM would likely have obliged, but Nintendo could also have gone to another foundry like MS and Sony did.

3. If Nintendo had chosen to go 32nm they could either have said so (why not, Intel announces processors on not yet mature processes all the time) or, if they preferred to keep it quiet they could have simply said nothing and let us guess like we are guessing for the GPU (actually that might possibly be pointing to the GPU coming in at 32nm, or at least pointing out that that process is not ruled out yet for the GPU). That they chose to design for a 45nm process does not mean that they did not have a choice to design for a 32nm process.

4. And if Nintendo had chosen to design at 32nm they wouldn't have had to do so. Given that they chose 45nm there was no problem ammouncing it so they did. No contradiction there.