News

Hard vs. Soft: Software may be eating the world, but hardware monetizes better

Source: TechSpot, Septemeber 11, 2022 [1]
Are software businesses less capital intensive than hardware?
Enterprise SaaS companies do not need a lot of capital to get to revenue. By contrast, it can cost a few hundred million just to get a chip to first revenue. For many venture investors, seeing all their dollars spent before the product can even be tested makes hardware seem like a bad bet.

ROS 2 Hardware Acceleration Working Group #11

Source: Acceleration Robotics, Septemeber 13, 2022 [2]
Join the #11th meeting of the ROS 2 Hardware Acceleration Working Group to discuss progress on the Robotic Processing Unit, a hardware ROS 2 microcontroller unit (MCU) and more!

We have exciting updates coming up including a) Robotic Processing Unit updates and release, b) Hardware acceleration in ROS 2 and Gazebo community survey updates and release schedule, c) REP-2008 updates, d) ROSCon and IROS hardware acceleration talks and e) a new WG subproject (roscore-v) that will design an open source software and hardware ROS 2 microcontroller unit (MCU) powered by RISC-V.

Intel CEO warns of more tough times ahead

Source: TechRadar, Septemeber 9, 2022 [3]
Intel CEO Pat Gelsinger has warned that the company's performance is likely to get worse before it gets better. He expects Intel to continue to lose share in the server processor market next year, with recovery starting in 2025. The chipmaker is also likely to take further steps to streamline its bloated product portfolio.

Nvidia claims 'record performance' for Hopper MLPerf debut

Source: The Register, Septemeber 9, 2022 [4]
Nvidia's Hopper-based H100 Tensor Core GPUs set new records in inferencing workloads. Company claims it delivers up to 4.5x more performance than previous GPUs. MLCommons analyzes the performance of systems using a machine learning model against new data.

Fully integrated AI platform for industrial robotics

Source: eenewseurope, Septemeber 7, 2022 [5]
Mov.ai has teamed up with Lanner Electronics in Taiwan to provide a fully integrated platform for industrial robotics. The two have combined Lanner’s Edge AI computing appliances, Mov.ai’s Robotics Engine Platform and Nvidia’s Isaac robotic operating system (ROS) platform.

NIST and Google team up on open-source computer chips for researchers and startups

Source: SiliconANGLE, Septemeber 14, 2022 [6]
Google and the U.S. Department of Commerce are teaming up to make silicon wafers available to academic researchers. The goal is to reduce costs associated with developing semiconductor devices. President Joe Biden's administration has been working to secure the chip supply chain.

Amazon acquiring warehouse robotics maker Cloostermans

Source: The Robot Report, Septemeber 9, 2022 [7]
Amazon acquires Cloostermans, a Belgium-based company that specializes in mechatronics. Amazon said this acquisition will ramp up its R&D and deployment in those areas. Financial terms of the deal were not disclosed. About 200 employees will join Amazon Global Robotics' growing presence in Europe.

SOSA-aligned single-board computers that combine CPU, GPU, and FPGA embedded computing offered by Aitech

Source: Military Aerospace, Septemeber 14, 2022 [8]
The boards are aligned to The Open Group Sensor Open Systems Architecture (SOSA) technical standard. The U-C850x series single-board computers offer three main pillars of modern data processing acceleration: central processing unit (CPU), Integrated graphics processing unit (iGPU), and field-programmable gate array (FPGA).

Scaling data pipelines: AT&T optimizes speed, cost, and efficiency with GPUs

Source: NVIDIA Technical Blog, Septemeber 8, 2022 [9]
AT&T's data teams were working to manage cloud costs while balancing simplicity at scale. The use of GPUs for each stage in the data-to-AI pipeline proved to be faster, cheaper, and simpler. We also provide insights on design considerations and explain how we optimized performance and price of our GPU cluster.

Using Vulkan SC for safety-critical graphics and real-time GPU processing

Source: NVIDIA Technical Blog, Septemeber 12, 2022 [10]
Vulkan SC (Safety Critical) is a newly released open standard to streamline the use of GPUs in markets where functional safety and hitch-free performance are essential. NVIDIA helped lead the creation of the new API and is now shipping production drivers on its NVIDIA DRIVE and NVIDIA Jetson platforms.

Papers

A survey on efficient Convolutional neural networks and hardware acceleration

Source: MDPI, March 18, 2022 [11]
Deep-learning-based representations have demonstrated remarkable performance in academia and industry. They often require substantial computation and memory resources while replacing traditional hand-engineered features. In this review, we focus on three aspects: quantized/binarized models, optimized architectures, and resource-constrained systems.

FSHMEM: Supporting partitioned global address space on FPGAs for large-scale hardware acceleration infrastructure

Source: arXiv, July 11, 2022 [12]
Partitioned Global Address Space (PGAS) has become one of the most promising parallel computing models. FPGA is getting attention as an alternative compute platform for HPC systems. FSHMEM is a software/hardware framework that enables the PGAS programming model on FPGAs.

Implementing hardware extensions for multicore RISC-V GPUs

Source: carrv, July 17, 2022 [13]
In this work, They present a generalized methodology for implementing hardware extensions for multi-core RISC-V-based GPUs. Part of the challenge involves extending the instruction set and register file. They also provide a generalized solution for supporting hardware monitoring counters for platforms with multiple custom accelerators onboard.

Previous Hardware Acceleration in Robotics Newsletters

Past ROS 2 Hardware Acceleration Working Group meetings

  1. Goldberg, J. (2022, September 11). Hard vs. soft: Software may be eating the world, but hardware monetizes better. TechSpot. https://www.techspot.com/news/95937-hard-vs-soft-software-may-eating-world-but.html ↩︎

  2. https://www.linkedin.com/events/ros2hardwareaccelerationworking6975460776610291713/comments/ ↩︎

  3. Khalili, J. (2022, September 9). Intel CEO warns of more tough times ahead. TechRadar. https://www.techradar.com/news/intel-ceo-warns-of-more-tough-times-ahead ↩︎

  4. Robinson, D. (2022, September 9). Nvidia claims 'record performance' for Hopper MLPerf debut. The Register: Enterprise Technology News and Analysis. https://www.theregister.com/2022/09/09/nvidia_hopper_mlperf/?utm_source=daily&utm_medium=newsletter&utm_content=article ↩︎

  5. Flaherty, N. (2022, September 7). Fully integrated AI platform for industrial robotics. eenewseurope. https://www.eenewseurope.com/en/fully-integrated-ai-platform-for-industrial-robotics/ ↩︎

  6. Wheatley, M. (2022, September 14). NIST and Google team up on open-source computer chips for researchers and startups. SiliconANGLE. https://siliconangle.com/2022/09/13/nist-google-team-open-source-computer-chips-researchers-startups/ ↩︎

  7. Crowe, S. (2022, September 9). Amazon acquiring warehouse robotics maker Cloostermans. The Robot Report. https://www.therobotreport.com/amazon-acquiring-warehouse-robotics-maker-cloostermans/ ↩︎

  8. SOSA-aligned single-board computers that combine CPU, GPGPU, and FPGA embedded computing offered by Aitech. (2022, September 14). Military Aerospace. https://www.militaryaerospace.com/computers/article/14282680/singleboard-computers-embedded-computing-sosa ↩︎

  9. Austin, M., Kanakadandila, S., Dabholkar, A., & Vo, C. (2022, September 8). Scaling data pipelines: AT&T optimizes speed, cost, and efficiency with GPUs. NVIDIA Technical Blog. https://developer.nvidia.com/blog/scaling-data-pipelines-att-optimizes-speed-cost-and-efficiency-with-gpus/ ↩︎

  10. Trevett, N., & Koch, D. (2022, September 12). Using Vulkan SC for safety-critical graphics and real-time GPU processing. NVIDIA Technical Blog. https://developer.nvidia.com/blog/using-vulkan-sc-for-safety-critical-graphics-and-real-time-gpu-processing/ ↩︎

  11. Ghimire, D., Kil, D., & Kim, S. H. (2022, March 18). A survey on efficient Convolutional neural networks and hardware acceleration. MDPI. https://www.mdpi.com/2079-9292/11/6/945 ↩︎

  12. Arthanto, Y. F., Ojika, D., & Kim, J. Y. (2022, July 11). FSHMEM: Supporting partitioned global address space on FPGAs for large-scale hardware acceleration infrastructure. arXiv.org. https://arxiv.org/abs/2207.04625 ↩︎

  13. Blaise, T., & Kim, H. (2022, July 17). Implementing hardware extensions for multicore RISC-V GPUs. carrv. https://carrv.github.io/2022/papers/CARRV2022_paper_11_Blaise.pdf ↩︎