A US startup has emerged from stealth mode with a programmable AI chip architecture and software for artificial intelligence and machine learning.
Blaize, formerly called Thinci, has backing from automotive companies such as Toyota and Denso, and currently has 325 staff around the world, including two design teams in the UK.
The AI chip technology, called a Graph Streaming Processor (GSP), uses a multicore architecture controlled by a state machine. The flow of the machine learning data is determined in the compiler beforehand. This means the compiler and the software development kit (SDK) is a vital part of the development, and the company claims that the architecture can process data ten to a hundred times faster than other AI chips with lower power consumption.
The company has developed prototype chips on a multiproject wafer that is being used with the software by fifteen early adopters around the world. However it is not revealing details of the architecture or power consumption figures until the first half of next year, other than to say it uses small amounts of SRAM with every execution unit to reduce the need to go to power hungry off chip high speed DRAM memory such as HBM.
“We are developing native computation of complex graphs,” said Dinakar Munagala, Co-founder and CEO of Blaize (above). “We had this vision of how to implement this in a small power envelope and silicon footprint. The way we run graphs natively is applicable to edge computing where we let the hardware manage the hardware scheduling and data dependencies without needing to go to DRAM as often as possible.
“We have a complex compiler and driver that understands the graph, and an SDK compiler tool chain that can understand any network – for example in automotive applications using a point cloud behind radar, lidar, etc.”
Next: Early architecture details
The graph compiler constructs a data flow graph, and the control flow is implicit in that, says the company. At run time that map of the metaflow is passed to the state machine hardware in the chip which can switch the context and flow based on the incoming data in real time. A scheduler calls in the resources it needs.
“This is a scalable architecture,” says Munagala. “It can be small for the edge camera, larger for an aggregator box to cloud server, so we are building products to multiple application points.”
The compiler works with the popular machine learning frameowrks such as Tensorflow or Caffe, but Blaize has also developed a kernel compiler. “Customers have custom code and we are probably one of three companies on the planet that could compiler the C++ code,” said Munagala. “There are some workloads that don’t exist in a library or are proprietary, and our ability to add a true compiler allows them to develop proprietary code.”
“It’s a modular compiler so its all integrated together. We started on automotive qualification from day one of working on the compiler as well as the hardware and the software components,” he added.
Eighteen months ago the company acquired the UK MIPS design team from Imagination Technologies when the US team was acquired by competitor Wave Computing. It also hired a team in Leeds that had been working on compiler technology for Sega, and together these form the core of the platform engineering tram. The development engineers in Hyderbad are working on the chip and compiler technology.
Next: Investor comments
“The coming out of Blaize and its leading Graph Streaming Processor is extremely exciting,” says David (Dadi) Perlmutter, an angel investor, entrepreneur and former EVP and Chief Product Officer of Intel corporation. “As an initial investor in Blaize, I recognized early on the great efficiency of one of the first to market a complete solution designed from scratch, fully optimized for AI and Neural Network applications. The unprecedented efficiency is great for a wide range of edge applications, particularly the automotive market. I am proud of the team in delivering on the promise.”
“Blaize is very innovative, our important business partner,” says Yukihide Niimi, CEO of NSITEXE and DENSO Advisory Board member. “DENSO is demonstrating leadership in many areas as the automotive industry undergoes extraordinary technology changes. NSITEXE was established to catch up such technology change and to accelerate development of flexible compute IP solutions like DFP. NSITEXE is willing to work together with Blaize to boost the flexible Graph (Data Flow) compute technology eco-system.”
The technical storage or access is strictly necessary for the legitimate purpose of enabling the use of a specific service explicitly requested by the subscriber or user, or for the sole purpose of carrying out the transmission of a communication over an electronic communications network.
The technical storage or access is necessary for the legitimate purpose of storing preferences that are not requested by the subscriber or user.
The technical storage or access that is used exclusively for statistical purposes.The technical storage or access that is used exclusively for anonymous statistical purposes. Without a subpoena, voluntary compliance on the part of your Internet Service Provider, or additional records from a third party, information stored or retrieved for this purpose alone cannot usually be used to identify you.
The technical storage or access is required to create user profiles to send advertising, or to track the user on a website or across several websites for similar marketing purposes.