Paris-based digital signal processing (DSP) IP startup Vsora has introduced a suite of PetaFLOPS computational chips to accelerate Level 3 (L3) through Level 5 (L5) autonomous vehicle (AV) designs. Vsora just raised $4.2 million in a pre-Series B funding round to accelerate its development.
The Tyr chipset family uses Vsora’s AD1028 architecture, which combines DSP modules with Deep Learning Acceleration Modules (DLA). Tyr is claimed to perform between 258 trillion and 1,032 trillion operations per second (TOPS) and consumes 10 watts.
Khaled Maalaj, CEO of Vsora, told EE Times Europe: “In January 2022, we introduced a set of scalable algorithm accelerator designs based on our Vsora architecture that combines artificial intelligence in the form of DLA with DSP to exchange data across existing memory. on the slide.” “The combination of AI with DSP is essential to tackle Level 4 and Level 5 autonomous driving [AD]. The accelerators, fully programmable in real time in the field, are deployed as companion chips to the main processor, making them relevant for a range of applications, not limited to AD.”
The Tyr family now consists of three segments: Tyr1, Tyr2, and Tyr3. Maalaj said the difference lies in the sheer computing power offered. “This is due to the number of AI and DSP elements that are built into the chip in question. The higher the number, the higher the processing power with a moderate increase in power consumption, but roughly the same utilization factor and the same response time.”
Silicon is in the works, and Vsora is working with partners to roll out commercial products. It expects to complete an additional round of financing at the end of the process.
When asked if Vsora is already running on Tyr4 and Tyr5, Maalej said the architecture is scalable. However, “Algorithm development is happening very quickly, so we’re always looking to ensure our solutions are up-to-date. Fewer and fewer vendors today are able to handle new and advanced algorithms. We’re one of the few, and we’re very focused on staying that way.”
in the latestMicroprocessor reportLinley Group compared the Tyr1 against Nvidia Orin and Qualcomm AI 100 and concluded:
The advantage of Vsora is the impressive efficiency of its architecture. At 30 TOPS/W, the DLA core outperforms all competing IPs, including Ceva’s NeuPro-M. Thus, we expect it to outperform other DLA processors, and ResNet-50’s raw Tyr numbers are impressive. However, Nvidia and Qualcomm are experienced chip companies and know how to improve SoC design. Vsora may lose some of its advantage if the components around the DLA are not efficient. Until you receive and test the initial silicon, we won’t know if the company can achieve its aggressive power goal.
The company said it plans to make some “major” announcements within the next few months.
ADAS and AD
At this year’s Consumer Electronics Showthe automotive design ecosystem – mainly auto OEMs, Tier 1s, semiconductor suppliers and software developers – put Advanced Driver Assistance System (ADAS) technology in the spotlight.
For wizards, 2023 won’t just be about ADAS, which incorporates all abilities covered by Level 1 and Level 2+. He said, “What’s even more exciting is AD, from L3 to L5.”
Vsora expects to support this trend with its Tyr family, which Maalej describes as “the most advanced computational accelerator for mobile and high-end applications” that “excels in power delivery processing with low power consumption at an affordable price.”
In the “Microprocessor reportLinley Group noted:
Vsora’s architecture features AI and DSP engines connected to large SRAM that facilitate fast data sharing. The company believes this approach is critical for vehicles that must perform both functions — sometimes in the same algorithm. However, most AV designers seem willing to split AI and DSP between two separate chips. For those wanting a single-chip solution, Qualcomm’s AI Accelerator 100 also checks all the boxes, although it lags behind. [behind] Pictures in performance.
In his discussions with early clients, Maalej noted a growing interest in a new generation of algorithms that required “not just artificial intelligence but a combination of artificial intelligence and a digital signal processor.” He commented: “As the market improves in understanding what is necessary for autonomous driving, algorithms are being updated to reflect this. At the same time, we are seeing more sensors with higher resolutions exponentially increasing the amount of data being generated, which still has to be processed at the same time.” available limited.
Led by Otium Capital and existing investors, the funding round will be used to accelerate the development of the Tyr family and expand the company’s software and hardware teams to support customer needs.
“We started growing our research and development team at the end of last year, and at the end of 2022 we counted 12 engineers,” Maalaj said. “We are working to accelerate recruitment for 2023, and in parallel, we are also working with key partners to optimize our resources.”
He continued, “Our research and development [department] Spent significant time and brainpower to implement a software-oriented design flow. Tyr programming is done entirely at a high level via Matlab-like / Tensorflow-like / C++ selection. The Vsora compiler handles programming code transparently to the user, separating what runs from the main processor from what runs on the Tyr. This speeds up the job and simplifies the design and debugging flow. The program is constantly being improved, and part of the funding will be used to maintain this trend.”
The company was founded by Maalej and three other former DiBcom engineers in September 2015. Vsora product development is self-financed with loan support from Bpifrance, France’s public sector innovation financing agency.
In 2018, the startup raised $1.7 million in a Series A funding round from venture capital firms Omnes Capital, Partech Ventures, and angel investors. The funds were used to expand Vsora’s research and development and build sales channels in the United States and Asia.
Read also:
