Edge AI platforms beat GPU/CPU on performance and efficiency - Embedded.com

Edge AI platforms beat GPU/CPU on performance and efficiency

Offering what it calls a breakthrough in artificial intelligence (AI) computing hardware and software for edge AI applications and overcoming legacy GPU/CPU limitations, Blaize has announced new embedded and accelerator platforms, Pathfinder and Xplorer which enable ‘code-free’ edge AI development.

The platforms are built on the Blaize graph streaming processor (GSP) architecture. According to the company, with 16 GSP cores and 16TOPS of AI inference performance within a 7W power envelope, GSP can deliver up to 60x better system-level efficiency compared to GPU/CPUs for edge AI applications. It also enables 50x less memory bandwidth, and 10x lower latency. The Blaize GSP is a 100% programmable processor and features advanced capabilities such as multi-threading and streaming.

Blaize said its early access customers for Pathfinder and Xplorer are looking to production deployments of previously infeasible real- time AI edge smart vision use cases across industrial, automobility, smart retail, metro and factory applications. The company’s co-founder and CEO, Dinakar Munagala, said, “Today’s edge solutions are either too small to compute the load or too costly and too hard to productize. Blaize AI edge computing products overcome these limitations of power, complexity and cost to unleash the adoption of AI at the edge, facilitating the migration of AI computing out of the data center to the edge.” Edge AI markets are expected to grow rapidly as AI functionality becomes a requirement for many new products. Omdia Research projects edge AI chipset revenue to reach $51 billion in 2025.

The two new products from Blaize are the Pathfinder P1600, an embedded system on module (SOM), and the Xplorer accelerator platform. Blaize said the P1600 brings the programmability and efficiency benefits of the GSP to embedded edge AI applications deployed at the sensor edge, or on the network edge. No host processor is needed for the P1600, so it is just a case of ‘plug it in and go’.

The Xplorer platforms accelerate AI applications at the edge of the enterprise via plug-in to the PCIe slot in a host server or appliance. The X1600E is a small form factor accelerator platform for small and power-constrained environments such as convenience stores or industrial sites. It can be easily added to accelerate AI apps in industrial PCs or as a rack of cards in a small 1U server. Or, there is the X1600P, which is a standard PCIe-based accelerator in a half-height, half-width form factor. The X1600P can replace a power-hungry desktop GPU in edge servers and provide anywhere from 16-64TOPS of AI inference performance within a very low power envelope.

Blaize X1600P PCIe Accelerator Platform
The Blaize X1600P, is a standard PCIe-based accelerator in a half-height, half-width form factor, to replace ‘power-hungry’ desktop GPUs in edge servers (Image: Blaize)

The Blaize AI software suite is built on open standards with ease of accessibility, which means product developers are not locked to proprietary development environments. Comprising the Picasso software development kit (SDK) and AI Studio, a completely code-free visual interface, the software suite offers tools for both traditional developers and non-coder domain experts. In addition, both tools sets utilize Blaize Netdeploy, a Blaize system with edge-aware algorithms to get the best accuracy and performance for edge deployments.

Proof of concept customers

Blaize said it has a number of proof of concept (POC) customers using its system to enable AI at the edge. The applications are in:

  • Industrial: running five independent neural networks at 50FPS with less than 100ms aggregate latency monitoring human and robot pose detection and position, product ID and count, and product quality.
  • Smart City: three independent neural networks running at 50FPS monitoring human detection, pose, position, auto detection and traffic intersection safety and security.
  • Sensor fusion: high resolution FHD video and lidar/radar sensor fusion running as a complete graph- native application on Blaize GSP.
  • Last mile delivery: sensor fusion of HD cameras at 30FPS & lidar/radar with less than 100ms latency and less than 15W power.
  • Retail: four independent neural networks running at 50FPS monitoring human detection, position, mask, and theft for store safety and security, and product detection.

Customer samples of both product lines are available now with full production expected starting this year. The Blaize Xplorer X1600E is available for $299 in volume quantity, the Pathfinder P1600 SOM is available in industrial grade for $399 in volume quantity, and the Xplorer X1600P is available for $999 in volume quantities.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.