November 13, 2023|Press Releases

Neuchips Unleashes Breakthrough: LLM Accelerator Achieves 800 Tokens/Second with Just 8 Inferencing Chips Across 8 Cards, All in a Single Server at Max. 800 Watts

SHARE

San Jose, November 12th, 2023

 

460x260g.png (86 KB)

Neuchips announces the availability of the Gen AI N3000 Accelerator which delivers industry leading performance and lower TCO than existing solutions. The product provides substantial savings for companies utilizing LLM services, enabling them to operate more efficiently and cost-effectively whether in the cloud or on-premises for Enterprise Gen AI applications. The Gen AI N3000 Accelerator chip, originally designed for recommendation workloads, sets a new standard in delivering exceptional power, scalability, and energy efficiency which benchmarks at 1.7x the competition in MLPerf 3.0 DLRM benchmarking.

 

The Gen AI N3000 Accelerator leads the industry in TCO affordability based on its power consumption at maximum 55W TDP per card, exceptional performance throughput of 800 tokens per second, and 100% linear scalability with a multi-card system all in a single server.

 

Featuring Neuchips accelerator chip and utilizing 32GB LPDDR5 memory with patented FFP8 (Flexible Float Point 8) precision, the Gen AI N3000 Accelerator is engineered to address the memory-bound challenges inherent in LLM applications, catering to both cloud and on-premise solutions, especially for enterprise applications. Different from traditional solutions, the Gen AI N3000 Accelerator increases resource utilization and concentration, maximizing efficiency and profitability for users.

 

The Gen AI N3000 Accelerator enables a seamless transition from existing development kits with its user-friendly software stack. Its patented FFP8 technology improves inferring accuracy, quantizes data for memory capacity, and optimizes bandwidth in LLM applications. With 6.4GHz LPDDR5 with ECC and support for up to 128 GB on-card LPDDR5, the Gen AI N3000 Accelerator ensures robust and reliable memory performance, providing users with a stable and efficient computing experience.

 

For more information on the Gen AI N3000 Accelerator and its transformative capabilities, please contact https://www.neuchips.ai/contact

Dive into the trial programs for an exclusive firsthand experience trial2023@neuchips.ai

Book a demo at SC23: contact@neuchips.ai

 

Get the latest NEUCHIPS news by email.