Intel has quietly cancelled its hybrid Thunder Bay system-on-chip (SoC), which combines general-purpose CPU cores with computer vision-focused Movidius hardware. The chipmaker does not explain its decisions, but it appears that Intel’s CPUs and vision processing units (VPUs) will remain separate for the time being.
Intel withheld information about its Thunder Bay SoC. The Thunder Bay SoC was supposed to be a low-power design with Arm Cortex-A53 CPU cores and Movidius VPU hardware, according to Linux patches discovered by Phoronix (which Intel acquired by taking over Movidius in 2016). The product’s exact configuration remained unknown.
The Thunder Bay SoC from Intel was designed for commercial and Internet-of-Things applications that required computer vision acceleration and general-purpose processing power.
Meanwhile, it appears that users of CPU and VPU-intensive applications may be satisfied with their edge servers running Xeon and Movidius silicon, such as the Keem Bay accelerator card introduced in 2019.
Furthermore, as machine learning acceleration becomes more common, many applications may switch to different hardware, such as Intel’s Habana Gaudi, Nvidia’s GPUs, or Jetson SoCs (with integrated GPU cores). As a result, whether the chip maker decides to offer a Thunder Bay-like SoC in the future and how this potential product will be configured remains to be seen.
While Movidius VPUs are not frequently mentioned, they do have advantages. The Movidius vision processing unit combines general-purpose MIPS cores with programmable 128-bit vector processing (referred to as SHAVE cores), a variety of hardware accelerators, and image signal processing capabilities. As a result, VPUs are better suited for edge computing applications in terms of power consumption and footprint than high-performance AI/ML accelerators.