Flopper.io GPU Comparison Sheet
August 26, 2026
flopper.io/compare/intel-gaudi-3-128gb-vs-intel-gaudi-3-hl-338Intel Gaudi 3 128GB vs Intel Gaudi 3 HL-338
Side by side datacenter GPU specifications from the Flopper.io database.
Overview
| Intel Gaudi 3 128GB | Intel Gaudi 3 HL-338 | |
|---|---|---|
| FP32 TFLOPS | 229.0 | 229.0 |
| VRAM | 128 GB | 128 GB |
| Memory Bandwidth | 3.7 TB/s | 3.7 TB/s |
| TDP | 900 W | 600 W |
Performance by Precision
| Precision | Intel Gaudi 3 128GB peak TFLOPS (dense) | Intel Gaudi 3 HL-338 peak TFLOPS (dense) |
|---|---|---|
| FP8 | 1678.0 | 1678.0 |
| FP16 | 459.0 | 459.0 |
| BF16 | 1678.0 | 1678.0 |
| TF32 | 459.0 | 459.0 |
| FP32 | 229.0 | 229.0 |
Specifications
| Intel Gaudi 3 128GB | Intel Gaudi 3 HL-338 | |
|---|---|---|
| Architecture | Gaudi 3 | Gaudi 3 |
| Process Node | 5nm | 5nm |
| Launch Year | 2024 | — |
| Form Factor | OAM | PCIe |
| Memory Type | HBM2E | HBM2E |
| Interface Width | — | — |
| Interconnect | RoCE v2 RDMA (24x 200GbE), 1200 GB/s bidirectional | RoCE v2 RDMA over top bridge (HLTB-304A/304B), 900 GB/s per card in a 4-card quad |
| TDP | 900 W | 600 W |
| FP32 Efficiency | 0.25 TFLOPS/W | 0.38 TFLOPS/W |