Maximizing GPU Utilization to Near ~100%

Revolutionary Intelligence! Your Supercomputing '24 Recap 

Our NR1-S® AI Inference Appliance demonstrated software readiness, ease of use, and a 50-90% price/performance advantage versus AI-host CPU architecture in AI Inference servers at Supercomputing 2024 (SC24) in Atlanta in November. NeuReality demonstrated how it’s shaping the future of AI Inference - by thinking radically differently at the deep systems and architecture level.

But the real news was never-before-seen Large Language Model demonstrations of Llama 3 and Mistral running on NR1 server-on chips to super boost complimentary AI Accelerators to near ~100% maximum utilization! Unlike other solutions, NR1 replaces today's AI-host CPUs to unleash the full capability of GPUs; and in fact, any XPU.

Customer-Friendly and 3x Faster Time-to-Market

NeuReality makes rolling out your Generative AI projects a breeze with ready-to-go software and APIs, helping any business or government ramp AI solutions 3x faster and more cost-effectively than ever before.

To the right, we demonstrated an enterprise-ready Conversational AI LLM that handles the entire end-to-end function of a customer call center with Llama 3-ready software. It's so easy to install and to use, that it reduces the need for complex software porting and high labor hours for IT.

Our NR1-S Appliance housing Qualcomm® Cloud 100 Ultra accelerators is now deployed with Cirrascale cloud services, and separately with on-premise AI servers with a major Fortune 500 financial services customer.

It’s a once-in-a-generation architecture shift, where NR1 removes the traditional AI-host CPU bottlenecks that make AI Inferencing so costly and energy inefficient. Plus, the chip handles the AI pre and post-processing with ready-made compute graphs for each AI query type.

Value and Benefits to You

• Ready-Made, Plug-In AI Inference Solution: unique, simple experience with one-of-a-kind super boosted NR1-S AI Inference Appliance with ready-made software and APIs all in one out-of-the solution. You’ll be up and running in less than one day to hit the ground running. Go to market 3x faster with your LLM solutions (like Llama, Mistral), while reducing both your cost and time porting from Python to NR1 in far fewer steps.
• Unmatched Performance with Any AI Workload: With NR1-S, you can super boost any AI Model/Workload to near 100% utilization plus linear scalability versus limiting, conventional AI-host CPU architecture with <30-50% GPU utilization today. NeuReality demonstrated Llama 3, Mistral, Computer Vision, Natural Language Processing and Automatic Speech Recognition with anywhere from 50-90% performance gains and 13-15x greater energy efficiency at the lowest cost per AI query.
• Your Ticket to Acquiring New Customers: Your ability to win new markets and customers by delivering superior performance per dollar and per watt is unlimited. The NR1-S improves your time-to-market for great customer experiences with our easy-to-use and affordable-to-run AI Appliance. You will avoid *surprise* high operational with costs with NR1 versus AI Inference built on more costly and inefficient CPU architecture. And that means best-in-class Total Cost of Ownership.