Mastering AI Chip Complexity
Explore challenges and solutions in AI chip development
Download eBook
Overview
NeuReality, based in Caesarea, Israel, is an innovative company in the AI industry, specifically focusing on developing purpose-built AI inference platforms. Founded in 2019 by a team of system engineers, NeuReality aims to make AI adoption more accessible and efficient for businesses of all sizes. With a small but ambitious team, the company has quickly become a significant player in the AI market, which is projected to grow at a compounded annual growth rate (CAGR) of 36.8% from 2023 to 2030.
"When we wanted to shift to a stronger emulation machine, the flexibility of ZeBu Cloud helped us to easily do it and benefit from the speed and the number of gates supported."
Yossi Kasus
|Co-Founder and VP of VLSI at NeuReality
"The Synopsys support team can log in to easily see what is happening and work in parallel with us when we need the assistance."
Yossi Kasus
|Co-Founder and VP of VLSI at NeuReality
"We look at Synopsys as a strategic partner. As a startup, needs change as the company grows, so it’s important to have a partner that is flexible with your needs and helps you grow."
Yossi Kasus
|Co-Founder and VP of VLSI at NeuReality
Challenges
NeuReality faced several significant challenges that demanded prompt and effective solutions:
- Limited Budget: As a startup, NeuReality had a constrained budget for developing its first silicon AI chip.
- Time-to-Market Pressure: The need to expedite the AI chip design and verification process to stay ahead in a competitive market.
- Complex Emulation Requirements: Simulating hardware and software together for a multi-billion-gate 7nm chip was highly challenging.
- Scalability: The necessity to scale emulation resources quickly without incurring the high costs and delays associated with building an on-premises infrastructure.
Solution
NeuReality adopted Synopsys ZeBu? Cloud for its emulation needs, which provided a flexible, scalable, and cost-effective solution. Key features of the solution included:
- Cloud-Based Emulation: Eliminated the need for costly on-premises infrastructure, offering scalability and flexibility as project needs evolved.
- High-Capacity Emulation: Transition from ZeBu Server 4 to ZeBu Server 5, supporting up to 30 billion gates, allowed for robust and accelerated emulation.
- Comprehensive Virtual Transactors: Enabled the use of all required high-speed interfaces without additional hardware, facilitating a complete system-level test environment.
- Quick Ramp-Up: Achieved rapid deployment within a couple of weeks, significantly faster than the months required for on-premises solutions.
Results
Through the implementation of ZeBu Cloud, NeuReality achieved remarkable results:
- Successful Tape-Out: The 7nm chip was developed and brought to production on schedule.
- Performance Gains: The NR1 NAPU demonstrated 10x the AI inference performance at the same cost compared to traditional CPU-centric architectures.
- Cost Savings: Achieved 90% cost savings for the same performance level.
- Market Readiness: The chip is now supporting various applications like natural language processing, computer vision, speech recognition, and recommendation systems.
- Future Growth: NeuReality is exploring new process nodes and architectures for their next device, focusing on generative AI and large language model (LLM) acceleration.
With Synopsys ZeBu Cloud, NeuReality not only accelerated their development process but also positioned themselves strongly for future innovations in the AI industry.
