Tenstorrent and Samsung's Collaboration for the Quasar AI Chiplet

Tenstorrent, a rising name in the field of artificial intelligence, has recently announced its decision to utilize Samsung’s SF4X (4 nm-class) process technology for its upcoming Quasar chiplet. This low-cost, low-power chiplet is specifically designed for machine learning workloads.

Nov 30, 2023 - 17:32
 55
Tenstorrent and Samsung's Collaboration for the Quasar AI Chiplet

The Quasar Chiplet: Innovations and Design

Core Design and RISC-V Architecture

The Quasar chiplet represents a significant addition to Tenstorrent's product roadmap. It is expected to feature at least 80 Tensix cores, which are based on the RISC-V instruction set architecture. This design is tailored to efficiently handle various AI workloads, including formats such as BF4, BF8, INT8, FP16, and BF16.

Integration and Functionality

Tenstorrent's Quasar chiplets are designed to function cohesively in groups and will be paired with the company's CPU chiplets. They are equipped with advanced, non-blocking die-to-die interfaces to ensure seamless integration and communication between different chip components.

Samsung’s SF4X Process Technology

Tailored for High-Performance Computing

Samsung’s SF4X process technology, chosen for the Quasar chiplet, is specially designed for high-performance computing applications. It emphasizes high clocks and high voltages, key attributes that ensure optimal performance for demanding computing tasks.

Samsung’s Role and Expansion in the U.S.

Samsung Foundry's expansion in the U.S., particularly with its new fab near Taylor, Texas, set to become operational in 2024, plays a crucial role in this collaboration. Samsung’s advanced silicon manufacturing capabilities are expected to significantly bolster Tenstorrent's AI and RISC-V based innovations.

Also Check Corsair iCUE AF120 RGB Slim 120mm PWM Fan

Performance Expectations of Quasar Chiplet

Estimating Quasar's Capabilities

While Tenstorrent has not disclosed specific performance metrics for the Quasar chiplet, insights can be drawn from its Wormhole chiplet, which features a similar core count and achieves 328 FP8 TOPS performance. The Quasar chiplet, fabricated using a performance-enhanced process, is expected to deliver comparable, if not superior, performance metrics.

Strategic Choices in Power and Performance

Tenstorrent officially positions its Quasar chiplets as low-power, cost-effective solutions for machine learning. This raises questions about whether the company will prioritize maximizing performance or opt for a more balanced power strategy.

Financial and Strategic Implications

Investment and Strategic Interests

Tenstorrent's recent securing of $100 million in financing, co-led by Hyundai Motor Group and the Samsung Catalyst Fund, highlights the strategic interests of these major players. Both Hyundai and Samsung have vested interests in AI processor technology, making their investment in Tenstorrent a strategic move.

Samsung’s Potential Strategic Interest in Chipmaking

Samsung’s involvement goes beyond just being a foundry partner. The company’s investment in Tenstorrent, coupled with its advanced chipmaking capabilities, suggests a deeper strategic interest in the AI and machine learning market, potentially aiming to produce chips for broader applications.

The Broader Impact on the AI and Machine Learning Industry

The Role of RISC-V and AI in Data Centers and Automotive Solutions

The development of the Quasar chiplet, leveraging the RISC-V architecture, is set to have a significant impact on the AI and machine learning landscape, particularly in data center and automotive applications. This collaboration could pave the way for more advanced, efficient, and cost-effective AI solutions in these sectors.

Advancements in Machine Learning Workloads

The introduction of the Quasar chiplet is expected to bring new advancements in handling diverse machine learning workloads. Its design and capabilities could lead to more efficient processing of complex AI tasks, ultimately benefiting a range of applications from cloud computing to edge AI.

Conclusion: The Significance of the Tenstorrent-Samsung Collaboration

Pioneering Low-Cost, High-Performance AI Solutions

The partnership between Tenstorrent and Samsung, marked by the development of the Quasar chiplet, represents a significant step towards creating low-cost, high-performance solutions in the AI industry. This collaboration combines Tenstorrent's innovative design with Samsung's advanced manufacturing prowess.

Shaping the Future of AI and Machine Learning

As Tenstorrent prepares to launch its Quasar chiplets using Samsung's SF4X process technology, this collaboration is poised to shape the future landscape of AI and machine learning. The extended availability, advanced design, and strategic investments underscore the potential of this partnership to drive innovation and set new benchmarks in the AI sector.