Is ADAS The Edge?


Debate is brewing over whether ADAS applications fall on the edge, or if they are better viewed squarely within the context of the automotive camp. There is more to this discussion than just semantics. The edge represents a huge greenfield opportunity for electronics of all sorts, and companies from the mobile market and from the cloud are both rushing to stake their claim. At this point the... » read more

Neural Network Performance Modeling Software


nnMAX Inference IP is nearing design completion. The nnMAX 1K tile will be available this summer for design integration in SoCs, and it can be arrayed to provide whatever inference throughput is desired. The InferX X1 chip will tape out late Q3 this year using 2x2 nnMAX tiles, for 4K MACs, with 8MB SRAM. The nnMAX Compiler is in development in parallel, and the first release is available now... » read more

Improving Edge Inferencing


Cheng Wang, senior vice president of engineering at Flex Logix, talks with Semiconductor Engineering about how to improve the efficiency and speed of edge inferencing chips, what causes bottlenecks, and why AI chips are different from other types of semiconductors. » read more

The Case For Embedded FPGAs Strengthens And Widens


The embedded FPGA, an IP core integrated into an ASIC or SoC, is winning converts. System architects are starting to see the benefits of eFPGAs, which offer the flexibility of programmable logic without the cost of FPGAs. Programmable logic is especially appealing for accelerating machine learning applications that need frequent updates. An eFPGA can provide some architects the cover they ne... » read more

Designing For The Edge


Chip and system architectures are beginning to change as the tech industry comes to grips with the need to process more data locally for latency, safety, and privacy/security reasons. The emergence of the intelligent edge is an effort to take raw data from endpoints, extract the data that requires immediate action, and forward other data to various local, regional or commercial clouds. The b... » read more

Multi-Layer Processing Boosts Inference Throughput/Watt


The focus in discussion of inference throughput is often on the computations required. For example, YOLOv3, a power real time object detection and recognition model, requires 227 BILLION MACs (multiply-accumulates) to process a single 2 Mega Pixel image! This is with the Winograd Transformation; it’s more than 300 Billion without it. And there is a lot of discussion of the large size ... » read more

Week in Review: IoT, Security, Auto


Internet of Things Smart-building technology is a factor in marketing new facilities to prospective tenants. The new Cambridge Crossing development in Cambridge, Mass., aspires to attract tech-oriented tenants much like nearby Kendall Square, this analysis notes. Philips has agreed to lease seven floors in Cambridge Crossing’s first office building, making that location its North American he... » read more

Week In Review: Design, Low Power


IP Flex Logix debuted its new InferX X1 edge inference co-processor, which incorporates the interconnect technology from its eFPGAs and its inference-optimized nnMAX clusters. The chip focuses on high throughput in edge applications with a single DRAM and is optimized for small batch sizes in edge applications where there is typically only one camera/sensor. InferX X1 will be available as chip... » read more

Spreading Intelligence From The Cloud To The Edge


The challenge of partitioning processing between the edge and the cloud is beginning to come into focus as chipmakers and systems companies wrestle with a massive and rapidly growing volume of data. There are widely different assessments of how much data this ultimately will include, but everyone agrees it is a very large number. Petabytes are simply rounding errors in this equation, and tha... » read more

Inference Acceleration: Follow The Memory


Much has been written about the computational complexity of inference acceleration: very large matrix multiplies for fully-connected layers and huge numbers of 3x3 convolutions across megapixel images, both of which require many thousands of MACs (multiplier-accumulators) to achieve high throughput for models like ResNet-50 and YOLOv3. The other side of the coin is managing the movement of d... » read more

← Older posts