Nvidia responds to Intel's claims about CPU-GPU AI performance, positioning its GPUs against Intel's processors in the data center AI chip race.

The machine-learning-performance conflict between Intel and Nvidia has intensified, with Nvidia accusing Intel of spreading misleading benchmark results. Intel, seeking to overtake Nvidia in deep-learning capabilities, has claimed that its 64-bit x86 chips outperform Nvidia's at neural-network computation tasks. This escalating dispute came to a head last week when Intel announced that its top-end two-socket Xeon Platinum 9282 data-center processor exceeded Nvidia's Tesla V100 GPU in running ResNet-50, a popular convolutional neural network used for object-recognition and computer vision workloads.

However, as observers noted at the time, the comparison proved problematic — comparing dissimilar products on different terms. In response, Nvidia issued a sarcastic rebuttal through a blog post of its own. Paresh Kharya, Nvidia's director of product marketing, opened his response with pointed irony: "It's not every day that one of the world's leading tech companies highlights the benefits of your products. Intel did just that last week, comparing the inference performance of two of their most expensive CPUs to NVIDIA GPUs."

The dispute underscores deeper tensions in how companies benchmark performance across fundamentally different processor architectures. Neither Intel's claims of CPU superiority nor Nvidia's dismissal of the comparison as fundamentally flawed emerged unscathed from the exchange, leaving questions about the legitimacy of such cross-platform performance metrics in the machine-learning space.