Intel Gaudi 2 96GB vs NVIDIA A100 SXM4 80GB

The Gaudi 2 is the newer part: Gaudi 2, launched in 2022, against the A100's Ampere from 2020. Newer architectures typically add lower-precision formats and better throughput per watt, so check the precision rows your workload actually uses.

Dense throughput for the Gaudi 2 against the A100: FP32 45 vs 20 TFLOPS, FP16 432 vs 312 TFLOPS. Sparse figures, where the vendor publishes them, appear under each dense number in the performance table.

Memory is 96 GB against 80 GB, fed at 2.5 TB/s versus 2.0 TB/s. More memory per GPU means larger models fit before you have to shard across cards, which often matters more than raw TFLOPS for inference.

Power budgets are 600 W for the Gaudi 2 and 400 W for the A100. At FP32 that works out to 0.08 against 0.05 TFLOPS per watt.

For LLM training and inference, weight the FP8 and FP16 rows and memory capacity most heavily. For scientific and HPC workloads, the FP64 row is the one to read.