Programmable Solutions Group
While Intel’s (INTC) memory business is benefiting from a strong market, its FPGA (field-programmable gate array) is not performing well. Intel created the PSG (Programmable Solutions Group) in 2016 after the acquisition of Altera.
In fiscal 2Q17, PSG revenue fell 5% YoY (year-over-year) due to weakness in the data center and communication sectors. However, rival Xilinx (XLNX) reported 7% YoY revenue growth as weakness in data centers and communications was more than offset by strength in the industrial, aerospace and defense, broadcast, consumer, and automotive markets.
On a sequential basis, PSG revenue grew 3.5%, whereas Xilinx’s revenue grew only 1%. Intel expects PSG revenue to grow by mid-single digits in fiscal 2017, driven by increasing adoption of its 14-nanometer Stratix 10 FPGAs and several design wins.
Intel’s recent FPGA design wins
At its Build conference in May 2017, Microsoft (MSFT) disclosed the industry’s fastest public cloud network powered by Intel FPGAs. Intel has secured a design win for Audi’s Level 3 autonomous driving system, which will be powered by Cyclone V SoC (system on chip) FPGA technology. The autonomous driving system will feature object and map fusion, parking pilot, and sensor data pre-processing.
New Technology Group
Intel has created a separate group to develop and commercialize new technologies, which has higher expenses than revenue as it includes projects in testing. In fiscal 2Q17, NTG (New Technology Group) revenue fell 75% YoY to $144 million. However, its operating loss narrowed from $2.2 billion in fiscal 2Q16 to $985 million in fiscal 2Q17.
Movidius Neural Compute Stick
At the 2017 Computer Vision and Pattern Recognition conference in July 2017, the Movidius Neural Compute Stick was released, the first USB-based development kit that runs embedded DNN (deep neural networks) on ultra-low power embedded devices. In an interview with InfoQ, Intel NTG marketing director for Movidius Gary Brown talked about the features of the new Movidius Neural Compute Stick.
He explained that the stick features a Movidius Myriad 2 VPU (vision processing unit), which accelerates machine vision tasks such as image and computer vision processing. The standard USB 3 port allows users to deploy deep learning inference and AI (artificial intelligence) applications to a wide range of devices without cloud connectivity. The stick supports the Caffe neural network framework, which means developers can deploy custom networks trained on the Caffe framework.
The compute stick is ideal for DNN applications that involve segmenting scenes and tracking, classifying, and recognizing objects. The compute stick can be used for robotic vision, autonomous drones, or intelligent security camera development. Intel is leveraging its supplementary products to widen its product offerings in the IoT (Internet of Things), data center, and automotive markets.