New Intel Gaudi 3 AI chip is 50% faster than current industry leader, NVIDIA H100

Reading time icon 2 min. read

Readers help support MSpoweruser. We may get a commission if you buy through our links. Tooltip Icon

Read our disclosure page to find out how can you help MSPoweruser sustain the editorial team Read more

Intel made a show at its Vision 2024 conference, announcing a new AI accelerator, the Intel Gaudi 3, and a commitment to open and secure enterprise AI solutions. The new chip shows significant performance and efficiency improvements compared to Nvidia’s H100 GPU. Not just this, upcoming Intel Lunar Lake PC processors will deliver up to 100 TOPS of AI performance.

Intel claims Gaudi 3 is:

  • Twice as energy-efficient
  • Performs AI tasks 1.5 times faster than the NVIDIA H100
  • 50% faster at training large language models
  • 40% better power efficiency for inference tasks

These improvements can reduce the cost of training large AI models. Gaudi 3 comes in various modules, including motherboard configurations and existing server designs.

Testing with Meta’s Llama project shows Gaudi 3’s capability for complex AI applications. Benchmarks indicate Gaudi 3 outperforms Nvidia’s H100 in training time and inference tasks for various large language models.

This launch marks Intel’s strongest challenge yet to Nvidia’s dominance in the AI chip market, currently estimated at 80%. 

Intel also emphasized its commitment to open standards and collaboration in the AI space. They announced plans to create an open platform for enterprise AI with leading industry partners. This platform will allow businesses to leverage best-in-class solutions from various vendors.

The company also unveiled its next-generation Xeon processors, codenamed Sierra Forest and Granite Rapids. These processors offer improved performance and efficiency for AI workloads. Intel also announced updates to its edge and client AI offerings, including new Core Ultra processors and edge-focused silicon.

Intel also introduced the Intel Tiber portfolio, a suite of business solutions designed to simplify the deployment of enterprise software and services, including those for GenAI.

More here.