SiTime product launch boosts efficiency of AI data centres

Author: Simon Rowley

SiTime Corporation, a precision timing company, has announced the differential-ended SiT5977 Super-TCXO, the newest member of the SiTime Elite RF family.

The company states that this is the only single-chip timing product that delivers the most resilient performance for AI compute-nodes with high bandwidth and network synchronisation. Applications include smart network interface cards (Smart NICs), acceleration cards, switches and compute nodes in the $200 billion data centre infrastructure market.

According to a recent IDC press release, “IDC expects the surging demand for AI workloads will lead to a significant increase in data centre capacity, energy consumption and carbon emissions, with AI data centre capacity projected to have a compound annual growth rate (CAGR) of 40.5% through to 2027.”

Dave Altavilla, President and Principal Analyst at HotTech Vision & Analysis, comments, “Improving AI workload efficiency to reduce energy consumption and carbon emissions is an industry-wide challenge. SiTime is the only semiconductor company fully dedicated to developing innovative timing solutions required for the complex scaling of today’s AI data centres.”

An AI cluster’s efficiency results in lower total cost of ownership (TCO) and energy consumption. Efficient clusters require high-bandwidth interconnects and tightly synchronised orchestration to minimise AI accelerator idle time. Tighter synchronisation also provides precise network telemetry for AI operations. This network telemetry improves performance and reliability by identifying and mitigating underperforming hardware before it fails.

The SiT5977 replaces multiple timing components. It optimises the efficiency of AI compute clusters with three times tighter synchronisation for data transmission and communications in real-world environments. In addition, its four times smaller size, compared to competing architectures, enables larger processors in compact systems. This allows system architects to fully leverage high-speed 800G bandwidth network connectivity and maximise utilisation of the AI cluster.

“AI training and inference are fundamentally distributed computing applications, which require accurate timing to synchronise activities,” adds Piyush Sevalia, Executive Vice President of Marketing at SiTime. “By enabling a new, more efficient architecture, the Elite RF timing solution uniquely supports more efficient AI workload processing, which may lead to higher revenue and lower TCO for data centres.”

The SiT5977 Super-TCXO streamlines AI system architecture. This new timing chip is an ultra-stable, low-jitter TCXO with the industry’s highest frequency differential output. This chip further integrates digital control to provide additional system-level programmability.

Key features:

• Environmentally robust with ±1 ppb/°C frequency slope (dF/dT) for optimum performance under airflow, thermal shock.
• Capable of driving 800G and higher links via 80 fs phase jitter and LVDS outputs.
• Enables embedded control loops with precise digital tuning of output frequency (DCTCXO), ±400 ppm pull range and 0.05 ppt (5e-14) resolution via I2C/SPI.
• Eliminates link flaps from quartz timing activity dips or micro jumps.
• Resistant to shock, vibration and board bending.
• Eliminates external LDOs via on-chip voltage regulators.
• 156.25 MHz output frequency enabling high-speed SerDes and 800G links.

The SiT5977 Super-TCXO is in production and samples are available now.

For more from SiTime, click here.