The strange bedfellows who want a little POWER

Thanks to reader Karl S for the initial report that got me down the rabbit hole, who discovered a quiet entry for a PowerPi workgroup at the OpenPOWER Foundation, clearly modeled on the ARM-based Raspberry Pi.

The obvious basis for a "little Power" device would be something like Microwatt, which Raptor is already using for Kestrel, and to which IBM staff regularly adds new features. While IBM also opened up other cores, notably the A2 family, Microwatt is (or at least will be) more like what's under our desk in terms of supported instructions and features, and is probably the easiest of the open synthesizeable cores to work with right now.

Still, Microwatt is "merely" a core, albeit a rapidly developing one, and not actually a product for sale or even a chip (though it may be in the near future). This makes the PowerPi concept particularly interesting in that it's trying to use real hardware to appeal to the same market segment that would rather experiment with something smaller, but end up unrealistically repelled by even Raptor's "low end" (relatively) pricing and reach instead for ARM or RISC-V boards because of their sensitivity to cost.

The PowerPi workgroup has four members. The first three are IBM (naturally); YADRO, a Russian manufacturer of high performance servers and storage solutions; and Van Tosh, a Belgian cloud solutions provider with their own RHEL offshoot. With the exception of IBM, none of them seem to have the expertise to work on a mini-POWER system, nor do their current sites indicate any particular proclivity.

But the fourth is noteworthy, a company called ChipEleven, a new, privately held VC-backed fabless design firm based in New Orleans that started just last year. Its CEO is a Master's student at Georgia Institute of Technology. They advertise a system on a chip with lots of peripheral interconnects, a quadcore ISA 3.0B-compliant CPU, an ARM Mali G77 GPU, onboard VGA up to 1920x1080, Ethernet, SPI flash and DDR4 RAM controller, with an ETA of 2023.

Does this sound a little familiar? It might, because an earlier project exists called Libre-SOC. Libre-SOC started with a RISC-V design to serve as a libre GPU, interestingly based on the CDC 6600, but shifted to OpenPOWER for higher performance and now plans to release an entire SoC instead. Raptor has provided material assistance (though Kestrel is proceeding on its own separate path) and development by volunteers is funded by NLNet donations. Allegedly, ChipEleven started with the Libre-SOC project and cleaved off, though to what extent (if any) ChipEleven's design derives from Libre-SOC is disputed.

It's not generally my habit to report on new speculative projects because a lot of them, even those with well-meaning and motivated people, will eventually flame out or their organizers lose interest. And ChipEleven may do just that: 2023 is a long way off (POWER10 will have certainly arrived by that point) and that doesn't mean there will be an actual physical product you can buy, let alone what it would cost. For its part, Libre-SOC wisely makes no promises. Still, the fact such a workgroup exists and has the backing of the OpenPOWER Foundation is notable, even if the current members are a motley bunch. Whether they can actually ship something is another story.