What we know

3 PFLOPS of FP8 compute performance. 5 TB/s. These GPUs are designed specifically for AI inference acceleration in data center environments.

Why it matters

The GPU D-Series represents Intel’s push into the AI inference accelerator market, combining high compute throughput with large, fast memory to handle demanding AI workloads. This positions Intel to better compete with other AI hardware providers by addressing the growing need for efficient and powerful inference solutions in data centers.

What is still unknown

Details on the accuracy of the AI computations, the range of supported AI frameworks or languages, and the timing of a broader commercial release have not been disclosed. No independent tests of performance or energy efficiency are available yet.