Future Power CPU's and ideas

[Casper]

NXP's QorIQ Power line has a bit of an uncertain future, so I was pondering what else could be a future candidate for a future project.

I was thinking, since the Power8 and Power9 cores are open for modification, could a potential dual core SMT4 (2 cores, 8 threads) or a quad core SMT4 (4 cores, 16 threads) be a viable option for a laptop? (I think a quad core design might guzzle up too much power). If a dual core full Power9 CPU could be utilized (maybe at a lower clockspeed as well), what would it enable us to do? Who would design it and who would build it?

Thoughts?

[Carlos]

Hmm, difficult. I think, after Apple jumped ship in 2006 no commercial company sees a market in new desktop/ mobile PowerPC chips. To build a POWER derivate chip, you need a lot of experience. Even IBM sees no one able to do such a development.

It's also a question of ROI (return on investment). You need a lot of money and a long breathing. As long as we don't have a lender of capital (such as e. g. BlackRock ) we can put that idea right out of our mind

BUT, I love your idea anyway, no question about that
Would be cool to see IBM doing this, pulling a hat (chip) overnight with twice the performance and half power consumption in comparison of Intels best offerings, like in 1997

[Casper]

I don't really see it as a POWER derivative chip, but rather a severely cut down version of their server chips (because that's what I think is hurting the TALOS Secure Workstation, it's prohibitively expensive). but I can imagine even making a cut down version would cost quite a bit of money.

However if NXP --> Qualcomm stops making Power based QorIQ chips, companies who are dependant on the Power chips like A-EON, ACube and such might pool together for such a project. i can even see Raptor Engineering opt for a more low-end Power chip to reduce costs.

[Carlos]

That's what I mean with "POWER derivative". The G5 (PPC970--) was a derivative of POWER4.
Yes, if Qualcomm leaves the PowerPC, only a few companies are left. IBM and AMCC, AFAIK. It seems that even AMCC is switching from PowerPC to ARM.

Even if IBM sees no one in doing a derivate of POWER chips, it's not looking good:
The number of "vendors who have the skills to be able to take a sophisticated processor and do derivatives of it for the server market is actually perhaps a smaller universe than we thought," Balog said

And IBM is only talking about the server market.

The only road that makes sense in a commercial way is to build a ultra low-power PowerPC SoC that fits in embedded and mobile devices, from phone to notebook. Then there is a bigger market maybe. But it would be really though to fight against ARM and Intel at once.

[Casper]

AMCC has seemingly ceased all development of Power based chips before Freescale. But since Freescale needed to design their cores from their own R&D, I can imagine Qualcomm passing these in favor of pre-made ARM cores.

I wonder if we grouped together with some companies and asked IBM to design a cut down POWER9 core how much that would cost (since the cores themselves already have been designed). Maybe if a large enough group of smaller companies buy a smaller design chip. If Intel can make more than 10 variants of every Intel Core design, then surely we should be able to get one more design out of POWER?

Maybe I'm a bit too optimistic.

[Carlos]

Why not. Visions are good. We will see what the future brings. But I'm pretty sure, you have to pay IBM a lot of money in doing this. And you can't compare with Intel, because they have a bigger market.

But I'm also pretty sure, if someone gives IBM enough money, the can build a chip that beats any Intel offerings.