Hi Terroni,
When the President/CEO of Frontgrade says
"Great job Jan!!! This product and partnership with Brainchip is a game changer!!!", I get a warm feeling*.
https://www.gaisler.com/products/gr801
Overview
This page describes a running development and no guarantees can be given concerning future product availability. All information on this page is subject to change without notice. Please click this link to sign up to receive notifications about product and documentation updates.
The GR801 architecture consists of a neuromorphic processing engine (Brainchip AkidaTM 1.0), a single-core RISC-V RV64GC processor (NOEL-V) for system management, and several digital interfaces.
GR801 is a reliable computing platform for AI applications such as image recognition, autonomous navigation, and data analysis. The neuromorphic engine provides fast and energy-efficient inference, enabling fast on-board decisions.
View attachment 90390
NOEL-V RISC-V Processor
AkidaTM neuromorphic processing accelerator
- General-purpose RISC-V processor (RV64GC)
- High-performance processing for traditional computing tasks
- System management and configuration of the neuromorphic engine
Interfaces
- Akida provides acceleration of AI models using event-based computing to minimize power consumption.
- Eight neural processing nodes connected in a mesh network
- Each node consisting of four convolutional or fully connected engines
- Each node supports 128 4x4 MACs, for total of 1024 MACs/clock
- Hardware support for 1, 2, or 4-bit hybrid quantized weights
- Multi-Pass Processing enables execution of large neural networks
- PCIe Gen 3 x4 port with capability to act both as a root port and as an endpoint
- 10/100/1000Mbit Ethernet interface
- 200Mbit/s SpaceWire port
- SPI,I2C and UART low speed interfaces.
Related project
NEUROSPACE: Pioneering Neuromorphic AI Acceleration in Space-Grade Microprocessors
Critical space applications calling for more onboard autonomy—such as in-orbit servicing, autonomous landing, docking, rovers, drones, space situational awareness, and debris removal—demand technologies that prioritize low latency, power efficiency, and fault tolerance.
GR801 is the integration of emerging neuromorphic computing technology with Gaisler proven space solutions, and the first product in our GRAIN roadmap.
Reference ESA - https://activities.esa.int/4000145267
This really is the first use of Akida IP in a customer SoC with a RISC-V processor, and it's radhard.
https://brainchip.com/frontgrade-gaisler-licenses-brainchips-akida-ip-to-deploy-ai-chips-into-space/
* ... but at my age ...
Now it seems my memory is playing tricks. I thought the FG licence announcement specified Akida 1, but the announcement only refers to "Akida IP":
https://brainchip.com/frontgrade-gaisler-licenses-brainchips-akida-ip-to-deploy-ai-chips-into-space/
Laguna Hills, Calif. – December 15, 2024 – BrainChip Holdings Ltd (ASX: BRN, OTCQX: BRCHF, ADR: BCHPY), the world’s first commercial producer of ultra-low power, fully digital, event-based, neuromorphic AI, today announced that Frontgrade Gaisler, a leading provider of radiation-hardened microprocessors for space applications, has licensed its Akida™ IP for incorporation into space-grade, fault-tolerant system-on-chip solutions for hardware AI acceleration. This commercial license agreement paves the path forward for AI chips to be deployed in space. The Akida neuromorphic computing solution was selected based upon providing real time data stream processing with resiliency, autonomy and accuracy at the lowest power, mass and volume.
The European Space Agency (ESA) has been leading the efforts to deploy neuromorphic computing to the furthest edge of AI. “While many fields no doubt can benefit from neuromorphic computing, one essential area is computer vision applications where current FPGA or GPU technology does not bring satisfactory results for edge deployments when considering mass, volume and power constraints,” said Laurent Hili, microelectronics and data handling engineer at the ESA. “This program has demonstrated the superiority of neuromorphic technology through several on-going activities with satellite primes and IP/components suppliers such as Frontgrade Gaisler and BrainChip.”
BrainChip’s neuromorphic AI will boost space-borne, on-board computers with superior power efficiency and inference performance while maintaining compatibility with existing Convolutional Neural Networks (CNNs).
“For the last 20 years, Gaisler’s space-grade microprocessors have been successfully deployed to every planet of our solar system in missions from ESA as well as most other space agencies,” said Sandi Habinc, General Manager at Frontgrade Gaisler, a Frontgrade company. “Together with BrainChip, we evaluated the Akida IP and decided that the next step of licensing it would beneficially augment our future space processors with neuromorphic AI.”
“This collaboration with Frontgrade Gaisler to license Akida IP for implementation into space SoCs represents an important step in satisfying the market demand for space-based AI deployments, turning into reality what once was considered unattainable,” said Sean Hehir, CEO of BrainChip. “We are pleased to expand on our trusted relationship with Frontgrade as they push the boundaries of space computing.”
About Frontgrade Gaisler
Frontgrade Gaisler, a Frontgrade company, is a leading provider of radiation-hardened microprocessors and IP cores for critical applications, particularly in the space industry. The company’s processors are known for their reliability, fault tolerance, and radiation tolerance, making them ideal for any space mission or other high-reliability application.
About Frontgrade
Frontgrade Technologies is the leading provider of high reliability and radiation-assured solutions for defense, space, intelligence, commercial, and civil applications. The company’s product portfolio is designed to perform in the harshest of environments and includes a complementary and integrated suite of mission-matched electronics ranging from radio frequency (RF) systems and microelectronics to motion control and power solutions. For more information, visit frontgrade.com.
About BrainChip Holdings Ltd (ASX: BRN, OTCQX: BRCHF, ADR: BCHPY) BrainChip is the worldwide leader in Edge AI on-chip processing and learning. The company’s first-to-market, fully digital, event-based AI processor, Akida™, uses neuromorphic principles to mimic the human brain, analyzing only essential sensor inputs at the point of acquisition, processing data with unparalleled efficiency, precision, and economy of energy. Akida uniquely enables Edge learning local to the chip, independent of the cloud, dramatically reducing latency while improving privacy and data security. Akida Neural processor IP, which can be integrated into SoCs on any process technology, has shown substantial benefits on today’s workloads and networks, and offers a platform for developers to create, tune and run their models using standard AI workflows like Tensorflow/Keras. In enabling effective Edge compute to be universally deployable across real world applications such as connected cars, consumer electronics, and industrial IoT, BrainChip is proving that on-chip AI, close to the sensor, is the future, for its customers’ products, as well as the planet. Explore the benefits of Essential AI at www.brainchipmay2022.flywheelstaging.com.
Of course, it would be far more sensible for FG to use Akida 2 IP with TENNs, especially as they are going to the expense of having the SoC incorporating a RISC-V processor and Akida IP manufactured. After all, would it have taken this long for FG to tapeout their RISC-V IP combined with Akida 1 IP, bearing in mind that GF had already done a radhard Akida 1500, albeit on their FD-SoI process?
In any case, it's worthwhile to refresh the memory re FG's pedigree:
"For the last 20 years, Gaisler’s space-grade microprocessors have been successfully deployed to every planet of our solar system in missions from ESA as well as most other space agencies"
)
