Intel Reportedly Cancels Thunder Bay Hybrid SoC
Intel has quietly canceled its hybrid Thunder Bay system-on-chip (SoC) that integrates general-purpose CPU cores with computer vision-oriented Movidius hardware. The chip maker hasn’t given a reason for the decision, but it looks like Intel’s CPU and vision processing unit (VPU) will remain separate for now.
“Due to product cancellation and no end customers or users, we are removing Thunder Bay specific code.” Linux patch discovered by phonics read.
Intel kept details about its Thunder Bay SoC a secret.Based on a Linux patch discovered by phonicsThe Thunder Bay SoC was intended to be a low-power design with Arm Cortex-A53 CPU cores and Movidius VPU hardware (acquired by Intel through its acquisition of Movidius in 2016). Still, the exact composition of the product remained unknown.
Intel’s Thunder Bay SoC was targeted at commercial and Internet of Things applications requiring computer vision acceleration and general-purpose processing capabilities. Such edge computing applications are expected to become more and more common in smart cities.
On the other hand, users with applications that require CPUs and VPUs will likely be happy with edge servers running Xeon and Movidius silicon, such as the Keem Bay accelerator cards introduced in 2019.
Moreover, as machine learning acceleration becomes ubiquitous, many applications are likely to adopt different hardware, such as Intel’s own Habana Gaudi, Nvidia’s GPUs, and Jetson SoCs (which use integrated GPU cores). I have. As a result, it remains to be seen if Intel will decide to offer a Thunder Bay-like SoC in the future and how this potential offering will be configured.
Movidius VPUs aren’t mentioned regularly, but they do have their advantages. The Movidius Vision Processing Unit features a general-purpose MIPS core with programmable 128-bit vector processing (called SHAVE core), various hardware accelerators, and image signal processing functions. VPUs are therefore more tailored for edge computing applications in terms of power consumption and footprint than high-performance AI/ML accelerators.