Microsoft’s Innovative 4-Processor PC


Microsoft’s Innovative 4-Processor PC

Microsoft Build is Microsoft’s most interesting event because it focuses on the people that build stuff, mostly code, but often, as is the case this year, hardware.

Last week, Microsoft held its latest Build event and I’m pretty sure it caused most of the PC OEMs to freak out. This is because Microsoft announced a new focused workstation for developers called Project Volterra. It has four processors and is based on ARM, not x86, and it’s coupled with a major effort to provide ARM native code that allows that platform, with Qualcomm’s help, to reach its full potential once the code is available toward the end of 2022.

But ARM is only one of four processors. We still have the GPU, but Microsoft added a NPU and an ACU (Azure compute unit), and that last one isn’t even in the PC. Let’s talk about how Microsoft is radically rethinking the PC in a cloud world, and how disruptive this needed change is likely to be.

Then we’ll close with my product of the week, which has to be Project Volterra because it reminds me of the old PCjr from IBM but done right. (IBM crippled the IBM PCjr because they were rightly afraid it would cannibalize their IBM PC sales, creating what is now a textbook product mistake.)

Inside the 4-Processor PC

Today, PCs have two processors, a CPU that handles numerically related information, and a GPU that is more focused on unstructured data and visual information. Together they define how a PC performs, with the current trend shifting loads from the CPU to the GPU because they are increasingly less structured and more visually focused, particularly when it comes to how PCs present their information.

But with the rise of artificial intelligence — and the fact that AI operates very differently than apps designed for CPUs or GPUs, by forming decision chains based on neural network capabilities premised on how we believe our brains work — these loads work inefficiently on CPUs, and though more efficiently on GPUs, beg for a very different hardware architecture designed specifically for those workloads.

Enter the NPU or neural processing unit. On paper it can outperform both the CPU and GPU with AI loads doing more with far less power and opening the door to developers who want to create applications that can use a focused and more efficient AI processing platform. It implies a much harder focus on AI capabilities going forward, and Microsoft has said that, in the future, all PCs will have NPUs.

But what about the APU? Well, this is an acronym I came up with. APU stands for Azure processing unit. This is that second shoe we’ve been waiting to drop ever since Satya took over Microsoft. It refers to a persistent connection to Azure in the cloud for additional processing power. It is really the first hardware implementation at the endpoint that addresses the hybrid world we live in today.

By hybrid I don’t mean working from home and in the office, though that does apply to the world we are in today. Nor does this apply to hybrid cloud as we currently talk about it which has to do with server loads. This is a new hybrid concept, one where the loads are shifted between the cloud and the desktop as needed.

Like PCjr – but in a Good Way

Project Volterra is a new class of workstation with all four processors based on ARM and focused on developers who develop for ARM-based PCs. As I mentioned earlier, this reminds me of the PCjr (pronounced “PC junior”) from IBM back in the 1980s but done right.

The PCjr was a revolutionary modular design that was incredibly well priced for the time and provided an easy upgrade path that would have anticipated the PC-as-a-service concept that came decades later.

But someone in IBM planning raised the concern that the PCjr, which was targeted at consumers, was too good because it made the vastly more expensive IBM PC look old and overpriced. So, they crippled the PCjr and effectively killed it, leading them to learn the lesson that you never cripple a product because it is too good. If customers prefer it, you pivot to that preference to assure that customer needs are prioritized over revenue.

 Microsoft’s Innovative 4-Processor PC

Microsoft Build is Microsoft’s most interesting event because it focuses on the people that build stuff, mostly code, but often, as is the case this year, hardware.

Last week, Microsoft held its latest Build event and I’m pretty sure it caused most of the PC OEMs to freak out. This is because Microsoft announced a new focused workstation for developers called Project Volterra. It has four processors and is based on ARM, not x86, and it’s coupled with a major effort to provide ARM native code that allows that platform, with Qualcomm’s help, to reach its full potential once the code is available toward the end of 2022.

But ARM is only one of four processors. We still have the GPU, but Microsoft added a NPU and an ACU (Azure compute unit), and that last one isn’t even in the PC. Let’s talk about how Microsoft is radically rethinking the PC in a cloud world, and how disruptive this needed change is likely to be.

Then we’ll close with my product of the week, which has to be Project Volterra because it reminds me of the old PCjr from IBM but done right. (IBM crippled the IBM PCjr because they were rightly afraid it would cannibalize their IBM PC sales, creating what is now a textbook product mistake.)

Inside the 4-Processor PC

Today, PCs have two processors, a CPU that handles numerically related information, and a GPU that is more focused on unstructured data and visual information. Together they define how a PC performs, with the current trend shifting loads from the CPU to the GPU because they are increasingly less structured and more visually focused, particularly when it comes to how PCs present their information.

But with the rise of artificial intelligence — and the fact that AI operates very differently than apps designed for CPUs or GPUs, by forming decision chains based on neural network capabilities premised on how we believe our brains work — these loads work inefficiently on CPUs, and though more efficiently on GPUs, beg for a very different hardware architecture designed specifically for those workloads.

Enter the NPU or neural processing unit. On paper it can outperform both the CPU and GPU with AI loads doing more with far less power and opening the door to developers who want to create applications that can use a focused and more efficient AI processing platform. It implies a much harder focus on AI capabilities going forward, and Microsoft has said that, in the future, all PCs will have NPUs.

But what about the APU? Well, this is an acronym I came up with. APU stands for Azure processing unit. This is that second shoe we’ve been waiting to drop ever since Satya took over Microsoft. It refers to a persistent connection to Azure in the cloud for additional processing power. It is really the first hardware implementation at the endpoint that addresses the hybrid world we live in today.

By hybrid I don’t mean working from home and in the office, though that does apply to the world we are in today. Nor does this apply to hybrid cloud as we currently talk about it which has to do with server loads. This is a new hybrid concept, one where the loads are shifted between the cloud and the desktop as needed.

Like PCjr – but in a Good Way

Project Volterra is a new class of workstation with all four processors based on ARM and focused on developers who develop for ARM-based PCs. As I mentioned earlier, this reminds me of the PCjr (pronounced “PC junior”) from IBM back in the 1980s but done right.

The PCjr was a revolutionary modular design that was incredibly well priced for the time and provided an easy upgrade path that would have anticipated the PC-as-a-service concept that came decades later.

But someone in IBM planning raised the concern that the PCjr, which was targeted at consumers, was too good because it made the vastly more expensive IBM PC look old and overpriced. So, they crippled the PCjr and effectively killed it, leading them to learn the lesson that you never cripple a product because it is too good. If customers prefer it, you pivot to that preference to assure that customer needs are prioritized over revenue.