Video Generation on Open-Sora-Plan 221 frames, 512×512
10.1LatencyTeaCache
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TeaCacheNumber of GPUs=8 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 10.1 | 28.1 | |
| TeaCacheNumber of GPUs=4 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 15.9 | 17.9 | |
| BWCacheNumber of GPUs=8 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 21.4 | 13.3 | |
| TeaCacheNumber of GPUs=2 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 26.9 | 10.6 | |
| BWCacheNumber of GPUs=4 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 32.8 | 8.7 | |
| PABNumber of GPUs=8 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 34.9 | 8.1 | |
| originalNumber of GPUs=8 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 47.1 | 6 | |
| TeaCacheNumber of GPUs=1 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 48.2 | 5.9 | |
| PABNumber of GPUs=4 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 60 | 4.7 | |
| BWCacheNumber of GPUs=2 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 60 | 4.7 | |
| originalNumber of GPUs=4 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 81.2 | 3.5 | |
| BWCacheNumber of GPUs=1 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 109.1 | 2.6 | |
| PABNumber of GPUs=2 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 112.5 | 2.5 | |
| originalNumber of GPUs=2 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 154.3 | 1.8 | |
| PABNumber of GPUs=1 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 211.4 | 1.3 | |
| originalNumber of GPUs=1 × A800, Dynamic Sequence Parallelism (DSP)=Enabled2025.09 | 284.3 | 1 |