Nvidia Stays In The Lead While AMD Shows Off Its Latest

The artificial intelligence (AI) market is rapidly evolving, with Nvidia standing out as a leader in innovation and performance. The company’s commitment to advancing silicon, software, and systems design has enabled it to maintain its top position in the AI market. Recently, the MLCommons organization released the latest benchmarks for inference processing, which has become a crucial aspect of AI hardware utilization. Inference processing now accounts for 80-90% of total AI hardware utilization once models are deployed at scale, and the market is projected to grow from $106 billion in 2025 to over $254 billion by 2030.

This growth is driven by the increasing demand for generative AI, agentic AI, and real-time application workloads. Nvidia has been at the forefront of this growth, publishing GB300 benchmarks and submitting serious benchmarks for the new Blackwell Ultra class GPUs. The company’s performance has been impressive, with the Blackwell Ultra outperforming Hopper GPUs by over five times. This significant performance boost is attributed to Nvidia’s innovative approach to disaggregating inference serving into context and generation, deploying each phase of AI inference on separate GPUs.

AMD’s Efforts to Compete with Nvidia

AMD has also been working to improve its position in the AI market, recently publishing a slew of benchmarks across the MI product line, including the new MI355X. While the performance of the AMD MI325 is about even with the Nvidia H200, Nvidia has already begun shipping the B300, two generations past the H200 Hopper architecture. AMD’s best opportunity to gain ground will come next year, with the MI400 and the “Helios” rack-scale system. The MI355X has shown a 2.7X performance improvement, largely due to the use of FP4, which has improved efficiency by up to 2X for all GPU vendors that support the smaller format while preserving accuracy.

AMD’s decision to run and publish more MLPerf benchmarks is a strategic move to help its partners better compete and sell AMD-powered AI servers. The company has submitted twice the number of benchmarks compared to the 4.1 round, and ran benchmarks with its latest ROCm software on three generations of GPUs. This increased transparency and willingness to participate in benchmarks will likely help AMD gain more traction in the AI market.

The Value of Running MLPerf Benchmarks

The MLPerf community has faced questions about the value of running these benchmarks, with some arguing that they are not representative of real-world workloads. However, the benchmarks provide a useful proxy for the various models in the real world, and can help companies demonstrate their performance and capabilities. For AMD, running MLPerf benchmarks is essential to helping its partners and customers justify the time and cost of running their real workloads on AMD hardware. The benchmarks also provide a way for companies to showcase their performance and gain bragging rights, which can be an important factor in attracting customers and investors.

As the AI market continues to evolve, the importance of running MLPerf benchmarks will only grow. Companies like Nvidia and AMD will need to continue innovating and improving their performance to stay ahead of the competition. With the market projected to grow significantly in the coming years, the stakes are high, and the companies that can demonstrate their performance and capabilities will be well-positioned to succeed.