Intel Shines at MLCommons New AI Benchmarks to Assess Performance

Intel said MLCommons’s latest results of the MLPerf v4.0 benchmark for inference has highlighted Intel’s advancements in AI performance.
Intel Gaudi 2 accelerators for AI
On Wednesday, MLCommons, an artificial intelligence benchmarking group, introduced a fresh array of tests and outcomes aimed at assessing the speed and responsiveness of cutting-edge hardware in running AI applications.

These new benchmarks, unveiled by MLCommons, focus on measuring the efficiency with which AI chips and systems process data-packed AI models to generate responses swiftly. The results provide insights into the rapidity with which AI applications, such as ChatGPT, can furnish responses to user queries.

One of the newly introduced benchmarks, dubbed Llama 2, evaluates the speed of a question-and-answer scenario for large language models, boasting 70 billion parameters. Developed by Meta Platforms, this benchmark showcases the prowess of AI chips in swiftly processing complex language interactions.

Additionally, MLCommons included a second text-to-image generator in its suite of benchmarking tools, known as MLPerf, which utilizes Stability AI’s Stable Diffusion XL model.

Servers powered by Nvidia’s H100 chips, utilized by industry giants like Alphabet’s Google, Supermicro, and Nvidia itself, emerged as clear winners in both new benchmarks due to their raw performance capabilities.

While Nvidia’s L40S chip also garnered attention, server builder Krai presented a design utilizing Qualcomm’s AI chip, highlighting its lower power consumption compared to Nvidia’s advanced processors.

Intel also entered the fray with its Gaudi2 accelerator chips, with the company expressing satisfaction with the performance outcomes.

However, beyond raw performance metrics, energy efficiency remains a crucial factor in deploying AI applications. Advanced AI chips consume substantial amounts of energy, posing a significant challenge for AI companies to strike a balance between performance and energy consumption, necessitating the development of chips that deliver optimal performance with minimal energy expenditure, Reuters news report said.

Intel Showcases Strong AI Performance

Intel’s results for the Intel Gaudi 2 accelerators and 5th Gen Intel Xeon Scalable processors with Intel Advanced Matrix Extensions (Intel AMX) underscore the company’s commitment to providing competitive AI solutions under its “AI Everywhere” initiative.

Intel said its Gaudi 2 AI accelerator stands out as the only benchmarked alternative to Nvidia H100 for generative AI (GenAI) performance, offering robust performance-per-dollar. Additionally, Intel remains the sole server CPU vendor to submit MLPerf results, showcasing its dedication to innovation in the AI space. The 5th Gen Xeon processors demonstrated an average improvement of 1.42x compared to the previous generation’s results, further highlighting Intel’s strides in AI inference performance.

Zane Ball, Intel corporate vice president and general manager of DCAI Product Management, emphasized the significance of these results, stating, “We continue to improve AI performance on industry-standard benchmarks across our portfolio of accelerators and CPUs. Today’s results demonstrate that we are delivering AI solutions that meet our customers’ diverse and evolving AI needs.”

The Intel Gaudi 2 software suite has expanded its model coverage for popular large language models (LLMs) and multimodal models, evident in its submissions for MLPerf Inference v4.0. Notably, Gaudi 2 delivered compelling performance in state-of-the-art models like Stable Diffusion XL and Llama v2-70B, with strong results in terms of samples-per-second and queries-per-second, reinforcing its competitive price/performance ratio.

Intel’s 5th Gen Xeon processors showcased significant improvements, with a geomean performance boost of 1.42x compared to the previous generation. Optimizations for models like GPT-J and DLRMv2 led to substantial performance gains and enhanced accuracy, demonstrating the efficacy of Intel’s hardware and software enhancements.

Intel’s collaboration with OEM partners, including Cisco, Dell, Quanta, Supermicro, and WiWynn, has facilitated the submission of MLPerf results, further validating the performance and versatility of Intel’s AI solutions.

Moreover, Intel’s commitment to providing accessible AI solutions is evident through the availability of 5th Gen Xeon processors and Intel Gaudi 2 accelerators for evaluation in the Intel Developer Cloud, offering users the opportunity to explore and deploy AI workloads at scale.

Intel’s strong showing in the MLPerf v4.0 benchmark reinforces its position as a leading provider of AI solutions, offering customers a comprehensive range of high-performance options to address diverse AI requirements.

Baburajan Kizhakedath

Related News

Latest News

Latest News