Hugging Face's logo

mrpopsalot
/
sd-speed-list

Text-to-Image
English
list
comparison
Model card Files
xet
 Community
7
mrpopsalot commited on
Commit
7f31fb3
·
1 Parent(s): 32d8feb

Initial Z-Image entries

Browse files
Files changed (3) hide show
  1. .gitignore +1 -0
  2. README.md +207 -3
  3. sd-speeds-v002.png → sd-speeds-v003.png +2 -2
.gitignore ADDED
@@ -0,0 +1 @@
 
 
1
+ .history/*
README.md CHANGED
@@ -76,9 +76,9 @@ tags:
76
  <div class="main-div">
77
  <h1>Pops' Stable Diffusion Speed List</h1>
78
  <div class="section-div">
79
- <img src="./sd-speeds-v002.png">
80
  <p>A hand curated list of generation speeds for various hardware and models.</p>
81
- <p><b>LAST UPDATED:</b> 2025.10.16</p>
82
  <p><b>Use the ComfyUI workflow above to start testing.</b></p>
83
  </div>
84
  <div class="section-div">
@@ -106,7 +106,7 @@ tags:
106
  <li>App versions are listed. If no version is listed then the commit version will be "Unknown".</li>
107
  </li>
108
  <p>Raw gen times are not recorded due to variance due to steps being variable. Instead iterations per second (and the inverse of it) are given since they are independent of steps.</p>
109
- <p>The given speed value (it/s or s/it) is used, and then extrapolated using the formula <code>1/speed</code> to get the other value. If its under 0.01 then it will be expanded to four digits compared to the usual 2</p>
110
  <p>If you can contribute to the list, do so as well. Lets make the most comprehensive, curated list of local Image Gen speeds!</p>
111
  <div class="subsection-div">
112
  <h3>Models Used</h3>
@@ -115,11 +115,215 @@ tags:
115
  <li>SD1.5: <a href="https://huggingface.co/jzli/Hassaku-1.3">jzli/Hassaku-1.3</a></li>
116
  <li>SDXL: <a href="https://huggingface.co/jzli/Hassaku-1.3">OnomaAIResearch/Illustrious-XL-v2.0</a></li>
117
  <li>Lumina 2: <a href="https://huggingface.co/neta-art/NetaLumina_Alpha">neta-art/NetaLumina_Alpha</a> Round NNNN EP6 S127716</li>
 
118
  </li>
119
  </div>
120
  </div>
121
  <div class="section-div">
122
  <h1>Benchmarks</h1>
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
123
  <div class="subsection-div">
124
  <h2>Lumina 2</h2>
125
  <h3>1536px</h3>
 
76
  <div class="main-div">
77
  <h1>Pops' Stable Diffusion Speed List</h1>
78
  <div class="section-div">
79
+ <img src="./sd-speeds-v003.png">
80
  <p>A hand curated list of generation speeds for various hardware and models.</p>
81
+ <p><b>LAST UPDATED:</b> 2025.11.29</p>
82
  <p><b>Use the ComfyUI workflow above to start testing.</b></p>
83
  </div>
84
  <div class="section-div">
 
106
  <li>App versions are listed. If no version is listed then the commit version will be "Unknown".</li>
107
  </li>
108
  <p>Raw gen times are not recorded due to variance due to steps being variable. Instead iterations per second (and the inverse of it) are given since they are independent of steps.</p>
109
+ <p>The given speed value (it/s or s/it) is used, and then extrapolated using the formula <code>1/speed</code> to get the other value. If its under 0.1 then it will be expanded to four digits compared to the usual 2</p>
110
  <p>If you can contribute to the list, do so as well. Lets make the most comprehensive, curated list of local Image Gen speeds!</p>
111
  <div class="subsection-div">
112
  <h3>Models Used</h3>
 
115
  <li>SD1.5: <a href="https://huggingface.co/jzli/Hassaku-1.3">jzli/Hassaku-1.3</a></li>
116
  <li>SDXL: <a href="https://huggingface.co/jzli/Hassaku-1.3">OnomaAIResearch/Illustrious-XL-v2.0</a></li>
117
  <li>Lumina 2: <a href="https://huggingface.co/neta-art/NetaLumina_Alpha">neta-art/NetaLumina_Alpha</a> Round NNNN EP6 S127716</li>
118
+ <li>Z-Image: <a href="https://huggingface.co/Tongyi-MAI/Z-Image-Turbo">Tongyi-MAI/Z-Image-Turbo</a></li>
119
  </li>
120
  </div>
121
  </div>
122
  <div class="section-div">
123
  <h1>Benchmarks</h1>
124
+ <div class="subsection-div">
125
+ <h2>Z-Image</h2>
126
+ <h3>1536px</h3>
127
+ <table>
128
+ <thead>
129
+ <th>Chip</th>
130
+ <th>it/s</th>
131
+ <th>s/it</th>
132
+ <th>Backend</th>
133
+ <th>App (Commit)</th>
134
+ <th>OS</th>
135
+ <th>Notes</th>
136
+ </thead>
137
+ <tbody>
138
+ <tr>
139
+ <td>NVIDIA RTX 3090</td>
140
+ <td>0.20it/s</td>
141
+ <td>4.90s/it</td>
142
+ <td>CUDA 12.6</td>
143
+ <td>ComfyUI (5151cff)</td>
144
+ <td>Arch Linux</td>
145
+ <td></td>
146
+ </tr>
147
+ </tbody>
148
+ </table>
149
+ <h3>1024px</h3>
150
+ <table>
151
+ <thead>
152
+ <th>Chip</th>
153
+ <th>it/s</th>
154
+ <th>s/it</th>
155
+ <th>Backend</th>
156
+ <th>App (Commit)</th>
157
+ <th>OS</th>
158
+ <th>Notes</th>
159
+ </thead>
160
+ <tbody>
161
+ <tr>
162
+ <td>NVIDIA RTX 3090</td>
163
+ <td>0.50it/s</td>
164
+ <td>2.01s/it</td>
165
+ <td>CUDA 12.6</td>
166
+ <td>ComfyUI (5151cff)</td>
167
+ <td>Arch Linux</td>
168
+ <td></td>
169
+ </tr>
170
+ <tr>
171
+ <td>NVIDIA GTX 980</td>
172
+ <td>0.0226it/s</td>
173
+ <td>44.26s/it</td>
174
+ <td>CUDA 12.6</td>
175
+ <td>ComfyUI (5151cff)</td>
176
+ <td>Arch Linux</td>
177
+ <td>FP8 DiT, CPU TE, PCIe 2.0x1</td>
178
+ </tr>
179
+ <tr>
180
+ <td>NVIDIA GTX 980</td>
181
+ <td>0.0215it/s</td>
182
+ <td>46.51s/it</td>
183
+ <td>CUDA 12.6</td>
184
+ <td>ComfyUI (5151cff)</td>
185
+ <td>Arch Linux</td>
186
+ <td>Q3_K_S C96 GGUF DiT, CPU TE, PCIe 2.0x1</td>
187
+ </tr>
188
+ <tr>
189
+ <td>NVIDIA GTX 1650 Mobile</td>
190
+ <td>0.0147it/s</td>
191
+ <td>67.85s/it</td>
192
+ <td>CUDA 12.8</td>
193
+ <td>ComfyUI (5151cff)</td>
194
+ <td>Arch Linux</td>
195
+ <td>FP8 DiT and TE</td>
196
+ </tr>
197
+ <tr>
198
+ <td>NVIDIA GTX 1050 Ti</td>
199
+ <td>0.0115it/s</td>
200
+ <td>86.22s/it</td>
201
+ <td>CUDA 12.6</td>
202
+ <td>ComfyUI (5151cff)</td>
203
+ <td>Arch Linux</td>
204
+ <td>Q3_K_S C96 GGUF DiT, CPU TE, PCIe 2.0x1</td>
205
+ </tr>
206
+ </tbody>
207
+ </table>
208
+ <h3>512px</h3>
209
+ <table>
210
+ <thead>
211
+ <th>Chip</th>
212
+ <th>it/s</th>
213
+ <th>s/it</th>
214
+ <th>Backend</th>
215
+ <th>App (Commit)</th>
216
+ <th>OS</th>
217
+ <th>Notes</th>
218
+ </thead>
219
+ <tbody>
220
+ <tr>
221
+ <td>NVIDIA RTX 3090</td>
222
+ <td>1.76it/s</td>
223
+ <td>2.01s/it</td>
224
+ <td>CUDA 12.6</td>
225
+ <td>ComfyUI (5151cff)</td>
226
+ <td>Arch Linux</td>
227
+ <td></td>
228
+ </tr>
229
+ <tr>
230
+ <td>NVIDIA GTX 980</td>
231
+ <td>0.082it/s</td>
232
+ <td>12.24s/it</td>
233
+ <td>CUDA 12.6</td>
234
+ <td>ComfyUI (5151cff)</td>
235
+ <td>Arch Linux</td>
236
+ <td>Q3_K_S C96 GGUF DiT, CPU TE, PCIe 2.0x1</td>
237
+ </tr>
238
+ <tr>
239
+ <td>NVIDIA GTX 1650 Mobile</td>
240
+ <td>0.0536t/s</td>
241
+ <td>18.69s/it</td>
242
+ <td>CUDA 12.8</td>
243
+ <td>ComfyUI (5151cff)</td>
244
+ <td>Arch Linux</td>
245
+ <td>FP8 DiT and TE</td>
246
+ </tr>
247
+ <tr>
248
+ <td>NVIDIA GTX 980</td>
249
+ <td>0.0474it/s</td>
250
+ <td>21.06s/it</td>
251
+ <td>CUDA 12.6</td>
252
+ <td>ComfyUI (5151cff)</td>
253
+ <td>Arch Linux</td>
254
+ <td>FP8 DiT, CPU TE, PCIe 2.0x1</td>
255
+ </tr>
256
+ <tr>
257
+ <td>NVIDIA GTX 1050 Ti</td>
258
+ <td>0.0457it/s</td>
259
+ <td>21.87s/it</td>
260
+ <td>CUDA 12.6</td>
261
+ <td>ComfyUI (5151cff)</td>
262
+ <td>Arch Linux</td>
263
+ <td>Q3_K_S C96 GGUF DiT, CPU TE, PCIe 2.0x1</td>
264
+ </tr>
265
+ </tbody>
266
+ </table>
267
+ <h3>256px</h3>
268
+ <table>
269
+ <thead>
270
+ <th>Chip</th>
271
+ <th>it/s</th>
272
+ <th>s/it</th>
273
+ <th>Backend</th>
274
+ <th>App (Commit)</th>
275
+ <th>OS</th>
276
+ <th>Notes</th>
277
+ </thead>
278
+ <tbody>
279
+ <tr>
280
+ <td>NVIDIA RTX 3090</td>
281
+ <td>4.06it/s</td>
282
+ <td>0.24s/it</td>
283
+ <td>CUDA 12.6</td>
284
+ <td>ComfyUI (5151cff)</td>
285
+ <td>Arch Linux</td>
286
+ <td></td>
287
+ </tr>
288
+ <tr>
289
+ <td>NVIDIA GTX 1650 Mobile</td>
290
+ <td>0.16it/s</td>
291
+ <td>6.02s/it</td>
292
+ <td>CUDA 12.8</td>
293
+ <td>ComfyUI (5151cff)</td>
294
+ <td>Arch Linux</td>
295
+ <td>FP8 DiT and TE</td>
296
+ </tr>
297
+ <tr>
298
+ <td>NVIDIA GTX 980</td>
299
+ <td>0.15it/s</td>
300
+ <td>6.71s/it</td>
301
+ <td>CUDA 12.6</td>
302
+ <td>ComfyUI (5151cff)</td>
303
+ <td>Arch Linux</td>
304
+ <td>Q3_K_S C96 GGUF DiT, CPU TE, PCIe 2.0x1</td>
305
+ </tr>
306
+ <tr>
307
+ <td>NVIDIA GTX 1050 Ti</td>
308
+ <td>0.0935it/s</td>
309
+ <td>10.69s/it</td>
310
+ <td>CUDA 12.6</td>
311
+ <td>ComfyUI (5151cff)</td>
312
+ <td>Arch Linux</td>
313
+ <td>Q3_K_S C96 GGUF DiT, CPU TE, PCIe 2.0x1</td>
314
+ </tr>
315
+ <tr>
316
+ <td>NVIDIA GTX 980</td>
317
+ <td>0.0523it/s</td>
318
+ <td>19.11s/it</td>
319
+ <td>CUDA 12.6</td>
320
+ <td>ComfyUI (5151cff)</td>
321
+ <td>Arch Linux</td>
322
+ <td>FP8 DiT, CPU TE, PCIe 2.0x1</td>
323
+ </tr>
324
+ </tbody>
325
+ </table>
326
+ </div>
327
  <div class="subsection-div">
328
  <h2>Lumina 2</h2>
329
  <h3>1536px</h3>
sd-speeds-v002.png → sd-speeds-v003.png RENAMED
File without changes