Use Inference Benchmarks Similar To Your Application


If an Inference IP supplier or Inference Accelerator Chip supplier offers a benchmark, it is probably ResNet-50. As a result, it might seem logical to use ResNet-50 to compare inference offerings. If you plan to use ResNet-50 it would be; but if your target application model is significantly different from Resnet-50 it could lead you to pick an inference offering that is not best for you. ... » read more

In-Memory Computing Challenges Come Into Focus


For the last several decades, gains in computing performance have come by processing larger volumes of data more quickly and with superior precision. Memory and storage space are measured in gigabytes and terabytes now, not kilobytes and megabytes. Processors operate on 64-bit rather than 8-bit chunks of data. And yet the semiconductor industry’s ability to create and collect high quality ... » read more

Mostly Upbeat Outlook For Chips


2019 has started with cautious optimism for the semiconductor industry, despite dark clouds that dot the horizon. Market segments such as cryptocurrencies and virtual reality are not living up to expectations, the market for smart phones appears to be saturated, and DRAM prices are dropping, leading to cut-backs in capital expenditures. EDA companies are talking about sales to China being pu... » read more

AI Market Ramps Everywhere


Artificial Intelligence (AI) has inspired the general populace, but its rapid rise over the past few years has given many people pause. From realistic concerns about robots taking over jobs to sci-fi scares about robots more intelligent than humans building ever smarter robots themselves, AI inspires plenty of angst. Within the technology industry, we have a better understanding about the pote... » read more

High Neural Inferencing Throughput At Batch=1


Microsoft presented the following slide as part of their Brainwave presentation at Hot Chips this summer: In existing inferencing solutions, high throughput (and high % utilization of the hardware) is possible for large batch sizes: this means that instead of processing say one image at a time, the inferencing engine processes say 10 or 50 images in parallel. This minimizes the number of... » read more

Inferencing In Hardware


Cheng Wang, senior vice president of engineering at Flex Logix, examines shifting neural network models, how many multiply-accumulates are needed for different applications, and why programmable neural inferencing will be required for years to come. https://youtu.be/jb7qYU2nhoo         See other tech talk videos here. » read more

Power/Performance Bits: Nov. 20


In-memory compute accelerator Engineers at Princeton University built a programmable chip that features an in-memory computing accelerator. Targeted at deep learning inferencing, the chip aims to reduce the bottleneck between memory and compute in traditional architectures. The team's key to performing compute in memory was using capacitors rather than transistors. The capacitors were paire... » read more

The Week In Review: Design


M&A GlobalFoundries formed Avera Semiconductor, a wholly-owned subsidiary focused on custom ASIC designs. While Avera will use its relationship with GF for 14/12nm and more mature technologies, it has a foundry partnership lined up for 7nm. The new company's IP portfolio includes high-speed SerDes, high-performance embedded TCAMs, ARM cores and performance and density-optimized embedded SR... » read more

Real-Time Object Recognition At Low Cost/Power/Latency


Most neural network chips and IP talk about ResNet-50 benchmarks (image classification at 224x224 pixels). But we find that the number one neural network of interest for most customers is real-time object recognition, such as YOLOv3. It's not possible to do comparisons here because nobody shows a YOLOv3 benchmark for their inferencing. But it's very possible to improve on the inferencing per... » read more

Making AI Run Faster


The semiconductor industry has woken up to the fact that heterogeneous computing is the way forward and that inferencing will require more than a GPU or a CPU. The numbers being bandied about by the 30 or so companies working on this problem are 100X improvements in performance. But how to get there isn't so simple. It requires four major changes, as well as some other architectural shifts. ... » read more

← Older posts