image/video generation benchmarks on edge AI hardware
70 published rows. Each table compares only results with the same model-size class and batch size. Search this workload · JSON.
Fastest in the best-covered like-for-like group (unknown, batch unknown): NVIDIA GB300 NVL72 (4x NVIDIA GB300) at 0.655 FPS A 2026-09-16 (published). 1 groups, 16 platforms.
Show data — 1 like-for-like tables
unknown · batch unknown (16 platforms)
✓ Like-for-like: model-size class unknown, batch unknown (size or batch not stated by source — unverified). Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| NVIDIA GB300 NVL72 (4x NVIDIA GB300) | 0.655 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b fp8 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Dell PowerEdge XE9712 (4x NVIDIA GB300) | 0.646 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b fp8 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA GB200 NVL72 (4x NVIDIA GB200) | 0.496 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b fp8 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA Jetson AGX Orin 64GB | 0.0800 FPS latency_single_stream · b=unknown | MLPerf stable-diffusion-xl per MLPerf v4.0 submission · TensorRT 9.0.1, JetPack 5.1.1, CUDA 11.4 · unknown | D | 65.00 W spec platform spec envelope (not measured with this result) | NVIDIA Jetson Benchmarks (MLPerf v4.0 results table) Older data (2024) |
| smci355-ccs-aus-m13-13 (8x AMD Instinct MI355X 288GB HBM3e) | 0.0731 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b bf16, fp8, fp4 · PyTorch 2.10.0+git8514f05, ROCm 7.2.2 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE ProLiant Compute XD685 (8x AMD Instinct MI355X 288GB HBM3e) | 0.0707 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b bf16, fp8, fp4 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| 8xMI355X_2xEPYC_9575F (8x AMD Instinct MI355X 288GB HBM3e) | 0.0706 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b bf16, fp8, fp4 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| G894-AD3-AAX7 (8x NVIDIA B300-SXM-270GB) | 0.0629 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b fp8 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| G894-ZD3-AAX7 (8x NVIDIA B300-SXM-270GB) | 0.0625 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b fp8 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA DGX B300 (8x NVIDIA B300-SXM-270GB) | 0.0622 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b fp8 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB) | 0.0613 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b fp8 · TensorRT 10.14, CUDA 13.1 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| G894-SD3-AAX7 (8x NVIDIA B300-SXM-270GB) | 0.0583 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b fp32 · TensorRT 10.14, CUDA 13.1 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Quanta Grand Teton 1.5 MI350 (8x AMD Instinct MI350X 288GB HBM3e) | 0.0565 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b bf16, fp8, fp4 · xDiT, xfuser 0.4.5, PyTorch 2.9.1+gitff65f5b, ROCm 7.13.0 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA DGX B200 (8x NVIDIA B200-SXM-180GB) | 0.0464 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b fp32 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| XE7745_8xMI350P_2xEPYC_9455 (8x AMD Instinct MI350P) | 0.0346 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b bf16, fp8, fp4 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE Cray XD670 (8x NVIDIA H200-SXM-141GB) | 0.0241 FPS throughput_offline · b=unknown | MLPerf wan-2.2-t2v-a14b bfloat16 · KISS-V: Krai Inference Serving Solution for Video (Fast Mode) · unknown | B | unknown power n/a no power published | MLPerf Inference v6.0, open division, data-centre suite 2026-04-01 (published) |
How to read these numbers
Every number is labelled. Confidence: A MLPerf closed, audited · B documented third-party method · C community / review · D vendor-published, not independently verified · E estimate. Power basis: measured with the result · spec vendor spec/envelope · power n/a not published.
Machine-readable: /find.json (see llms.txt); full tables are in “Show data”. Confidence grades: A audited MLPerf closed · B documented third-party · C community · D vendor-published · E estimate. Unknown values are shown as unknown, never estimated. Generated 2026-10-16 from public sources.