The evolution of IBM’s Power architecture has been the subject of intense debate in the technology community. Over the past few years, this architecture has undergone significant strategic changes that have generated criticism and expectations alike. As with KVM, we almost guessed everything IBM was going to announce; let’s take a second shot at Power11. In this case, we don’t have the kernel.org lists to clue us in, but we do have 10 years of trajectory since Power8 and a market with very clear demands for alternative architectures to x86, even more so when Intel is going through one of its worst moments in its history..

Background and a little history

With Power8, came Power OEM/LC systems, NVIDIA GPUs, the NVLink connector and the possibility of having a first version of KVM on Power(not to be confused with the 2024 announcement). However, in practice, the challenges outweighed the opportunities… and we’ll leave it at that 🙂. Some felt that IBM was ahead of the market, while others felt that there was a lack of supported and proven solutions on these servers to achieve the anticipated impact; there was even talk of mass adoption by Google or Rackspace. Power9 represented a milestone in IBM’s strategy by offering a more open and accessible architecture for the community. Through the OpenPOWER Foundation, IBM released a significant portion of the specifications and technologies associated with Power9, allowing third parties to design and manufacture their own systems based on this architecture, similar to what is done with ARM or x86. Companies such as Raptor Computing Systems developed Power9-based systems using open source firmware and software, offering highly auditable and user-controllable platforms. However, in the next generation, development delays-perhaps exacerbated by the COVID-19 pandemic-led IBM, upon launching Power10, tolicense blocks of intellectual property from Synopsys for components such as the DDR4/5 PHY and PCIe 5.0. This decision introduced proprietary firmware into the system, breaking with the tradition of openness established with Power9 and limiting community participation in the development of these technologies. In addition, changes in NVIDIA’s strategy since Power9, by using alternative architectures such as ARM-based GPUs, made it difficult for GPUs to return to Power. In fact, in Power10, the commitment to inference in the processor core, without relying on GPUs, was a strategic response that allowed artificial intelligence processing to be performed directly on the processor.

With the potential arrival of Power11, there is an expectation that IBM will address previous challenges and align its strategy with current market demands. This involves reintegrating GPUs and other accelerators, improving support for open source workloads and Linux applications, and continuing the development of AIX and IBM i as key components of the Power ecosystem.

Our expectations

The decisions around Power10 have had tangible effects on the community and the market. The move away from an open architecture generated skepticism among developers and companies that value transparency and collaborative development. Competitors with open architectures, such as RISC-V, have gained ground by offering the freedom and flexibility that Power10 did not provide. This highlights the importance of openness as a competitive advantage in an ever-evolving technology environment, where open source solutions dominate the market for new workloads. Looking ahead to Power11, there is a clear expectation that IBM will address these concerns. At SIXE, we advocate a return to open development practices, allowing access to firmware source code and specifications that enable greater collaboration and innovation. We expect Power11 to rectify the shortcomings of Power10, regaining control over the development of key components such as DDR PHY and PCIe interfaces. Avoiding dependence on intellectual property from outside vendors is essential to achieving a fully open architecture. In doing so, IBM can realign with community expectations and leverage the collective experience of developers and companies committed to open source principles. On the other hand, the re-integration of GPUs and other accelerators is crucial to address the growing demand for heterogeneous computing. By supporting a variety of accelerators-including GPUs, FPGAs, and specialized AI processors-IBM can offer flexible and powerful solutions tailored to specific workloads. This approach aligns with industry trends favoring modular and scalable architectures capable of handling complex and changing computational requirements. Strengthening support for open source workloads and expanding support for Linux applications are key to Power11 adoption. Ensuring seamless integration with open source frameworks and tools will attract a broader community of developers and facilitate the migration of existing applications to the Power platform. This strategy promotes innovation and responds to market demand for flexible and cost-effective solutions. In addition, we want to see how all of these hardware innovations can be exploited from AIX and IBM i, demonstrating IBM’s commitment to its large customer base. It is critical that businesses that rely on these operating systems can benefit from Power11 advances for their critical systems without sacrificing stability, performance, compatibility or availability.

Conclusion

If there is one thing we know for sure, it is that there is no one operating system or architecture that fits all workloads. What is most valuable for Power customers is the possibility of integrating on the same machines the databases on which their business depends on AIX or IBM i, private clouds with KVM, front-ends with Kubernetes on Linux and, hopefully soon, also AI, ML, HPC, etc. workloads. At SIXE we think that, just as there is no perfect music for every moment, there is no universal operating system, database or programming language. In Power we can have them all, and that’s why we love it. For us, Power11 represents an opportunity for IBM to realign its strategy: integrating GPUs and accelerators to meet high-performance computing needs, enhancing support for open source workloads and Linux applications, and continuing to develop its leading-edge operating systems for mission-critical environments, such as AIX and IBM i. In doing so, IBM can deliver a versatile and powerful platform that appeals to a broad spectrum of users. The success of Power11 will depend on IBM’s ability to balance proprietary innovation with openness and collaboration with third parties.