videorebirth commited on
Commit
c8f74e4
·
verified ·
1 Parent(s): 61f1311

Model card: add RTX PRO 6000 to the RTX benchmark table

Browse files
Files changed (1) hide show
  1. README.md +5 -4
README.md CHANGED
@@ -229,20 +229,21 @@ model.set_attention_backend("_flash_3_hub")
229
 
230
  All benchmarks measure end-to-end generation (text encoding, 6-step transformer, and video and audio VAE decode) at **1344×768 and 6 steps**. Clip lengths: `5.2 s` = 124 frames; `15 s` = 362 frames.
231
 
232
- ### 1. Consumer RTX GPUs (`5.2 s / 15 s` total time)
233
 
234
  <a id="comfyui-int8-benchmarks"></a>
235
 
236
  | GPU | INT8 (ComfyUI)<br>`*_comfyui/` file | GGUF `Q4_K_M` (ComfyUI)<br>≥ 64 GB RAM | GGUF `--fast-disk`<br>32 GB PC | GGUF `--fast-disk`<br>16 GB PC | INT8 (`hyperflow-h3-flash`)¹<br>`--offload stream` |
237
  |---|---|---|---|---|---|
 
238
  | **RTX 5090 (32 GB)** | 83 s / 518 s | 99 s / 575 s | 118 s / 618 s | 130 s / 615 s | 92 s / 551 s |
239
  | **RTX 4090 (24 GB)** | 152 s / 747 s | 157 s / 830 s | 151 s / 815 s | 153 s / 812 s | 130 s / 750 s |
240
  | **RTX 3090 (24 GB)** | — | 315 s / 1835 s | 318 s / 1869 s | 325 s / 1866 s | 266 s / 1684 s |
241
- | *Peak VRAM (5.2 s / 15 s)* | *5090: 31 GiB; 24 GB: 21 GiB (5.2 s)* | *5090: 28 / 27 GiB; 24 GB: 23 / 22 GiB* | *same as ≥ 64 GB* | *same as ≥ 64 GB* | *12.0–12.5 / 21.8–22.4 GiB* |
242
 
243
- - **ComfyUI INT8 column:** ComfyUI 0.44.0 with PyTorch 2.13 + CUDA 13.0 and the `nvfp4_awq` text encoder, on separate RunPod pods (60 GB / 31 GB of system RAM), without saving the MP4. On the same pods, `Q4_K_M` took 96 s / 562 s (5090) and 160 s / 834 s (4090), so INT8 is 5–15% faster. With 12 GiB of system RAM left to ComfyUI (a 16 GB PC) and no `--fast-disk`, INT8 took 88 s (5090) and 153 s / 821 s (4090) ([details](https://github.com/Video-Rebirth/hyperflow-h3-flash/blob/main/integrations/README.md#int8-performance)).
244
  - **GGUF columns:** Use the 15.7 GB `nvfp4_awq` text encoder. `--fast-disk` streams unresident weights from the OS file page cache instead of pinned RAM, keeping anonymous RSS at ~8–16 GiB.
245
- - **¹ `hyperflow-h3-flash` INT8 column:** Requires ≥ 91 GiB system RAM with the `bf16` text encoder, or ~69 GiB with `--text-encoder-precision fp8`.
246
 
247
  **Reference to video (`ref2va`) on one RTX GPU (`5.2 s / 15 s` total time):**
248
 
 
229
 
230
  All benchmarks measure end-to-end generation (text encoding, 6-step transformer, and video and audio VAE decode) at **1344×768 and 6 steps**. Clip lengths: `5.2 s` = 124 frames; `15 s` = 362 frames.
231
 
232
+ ### 1. RTX GPUs (`5.2 s / 15 s` total time)
233
 
234
  <a id="comfyui-int8-benchmarks"></a>
235
 
236
  | GPU | INT8 (ComfyUI)<br>`*_comfyui/` file | GGUF `Q4_K_M` (ComfyUI)<br>≥ 64 GB RAM | GGUF `--fast-disk`<br>32 GB PC | GGUF `--fast-disk`<br>16 GB PC | INT8 (`hyperflow-h3-flash`)¹<br>`--offload stream` |
237
  |---|---|---|---|---|---|
238
+ | **RTX PRO 6000 (96 GB)** | 53 s / 333 s | — | — | — | 61 s / 357 s |
239
  | **RTX 5090 (32 GB)** | 83 s / 518 s | 99 s / 575 s | 118 s / 618 s | 130 s / 615 s | 92 s / 551 s |
240
  | **RTX 4090 (24 GB)** | 152 s / 747 s | 157 s / 830 s | 151 s / 815 s | 153 s / 812 s | 130 s / 750 s |
241
  | **RTX 3090 (24 GB)** | — | 315 s / 1835 s | 318 s / 1869 s | 325 s / 1866 s | 266 s / 1684 s |
242
+ | *Peak VRAM (5.2 s / 15 s)* | *PRO 6000: 39 / 49 GiB; 5090: 31 GiB; 24 GB: 21 GiB (5.2 s)* | *5090: 28 / 27 GiB; 24 GB: 23 / 22 GiB* | *same as ≥ 64 GB* | *same as ≥ 64 GB* | *PRO 6000: 77 / 86 GiB; others: 12.0–12.5 / 21.8–22.4 GiB* |
243
 
244
+ - **ComfyUI INT8 column:** ComfyUI 0.44.0 with PyTorch 2.13 + CUDA 13.0 and the `nvfp4_awq` text encoder, on separate RunPod pods (188 GB / 60 GB / 31 GB of system RAM on the PRO 6000 / 5090 / 4090), without saving the MP4. On the same pods, `Q4_K_M` took 96 s / 562 s (5090) and 160 s / 834 s (4090), so INT8 is 5–15% faster. With 12 GiB of system RAM left to ComfyUI (a 16 GB PC) and no `--fast-disk`, INT8 took 88 s (5090) and 153 s / 821 s (4090) ([details](https://github.com/Video-Rebirth/hyperflow-h3-flash/blob/main/integrations/README.md#int8-performance)).
245
  - **GGUF columns:** Use the 15.7 GB `nvfp4_awq` text encoder. `--fast-disk` streams unresident weights from the OS file page cache instead of pinned RAM, keeping anonymous RSS at ~8–16 GiB.
246
+ - **¹ `hyperflow-h3-flash` INT8 column:** Requires ≥ 91 GiB system RAM with the `bf16` text encoder, or ~69 GiB with `--text-encoder-precision fp8`. On the RTX PRO 6000, `--offload auto` selects `none` instead: all components stay in VRAM, and the process peaks at 39 GiB of system RAM.
247
 
248
  **Reference to video (`ref2va`) on one RTX GPU (`5.2 s / 15 s` total time):**
249