Hugging Face's logo

mrpopsalot
/
sd-speed-list

Text-to-Image
English
list
comparison
Model card Files
xet
Community
7

Added 5090 generation times for Neta Lumina 2 and Illus 2.0 SDXL

#2
by Leyon - opened
base: refs/heads/main
←
from: refs/pr/2
Discussion Files changed
+6
-12
Files changed (1) hide show
  1. README.md +6 -12
README.md CHANGED
@@ -1,13 +1,3 @@
1
- ---
2
- license: gfdl
3
- language:
4
- - en
5
- pipeline_tag: text-to-image
6
- tags:
7
- - list
8
- - comparison
9
- ---
10
-
11
  # Methodology
12
  All settings use:
13
  - Euler sampler
@@ -20,8 +10,6 @@ Raw gen times are not recorded due to variance due to steps being variable. Inst
20
 
21
  The given speed value (it/s or s/it) is used, and then extrapolated using the formula `1/speed` to get the other value. If its under 0.01 then it will be expanded to four digits compared to the usual 2
22
 
23
- If you can contribute to the list, do so as well. Lets make the most comprehensive, curated list of local Image Gen speeds!
24
-
25
  ## Models Used
26
  The following are the models used for testing. The models you use can be the same architecture as the tested models
27
 
@@ -37,6 +25,7 @@ The following are the models used for testing. The models you use can be the sam
37
  ### 1024px
38
  | Chip | It/s | s/it | Backend | App | Notes
39
  |-------------|----|----|------|------|------|
 
40
  |NVIDIA RTX 3090|1.14it/s|0.87s/it|CUDA|ComfyUI|Sage Attention
41
  |NVIDIA RTX 3090|1.00it/s|1.00s/it|CUDA|ComfyUI
42
  |NVIDIA GTX 980|0.0599it/s|16.69s/it|CUDA|ComfyUI| FP32 CPU TE
@@ -46,6 +35,7 @@ The following are the models used for testing. The models you use can be the sam
46
  ### 512px
47
  | Chip | It/s | s/it | Backend | App | Notes
48
  |-------------|----|----|------|------|------|
 
49
  |NVIDIA RTX 3090|4.35it/s|0.23s/it|CUDA|ComfyUI|Sage Attention
50
  |NVIDIA GTX 980|0.28it/s|3.57s/it|CUDA|ComfyUI| FP8 CPU TE
51
  |NVIDIA GTX 980|0.25it/s|3.99s/it|CUDA|ComfyUI| FP32 CPU TE
@@ -54,6 +44,7 @@ The following are the models used for testing. The models you use can be the sam
54
  ### 256px
55
  | Chip | It/s | s/it | Backend | App | Notes
56
  |-------------|----|----|------|------|------|
 
57
  |NVIDIA RTX 3090|11.37it/s|0.0880s/it|CUDA|ComfyUI|Sage Attention
58
  |NVIDIA GTX 980|0.78it/s|1.27s/it|CUDA|ComfyUI| FP8 CPU TE
59
  |NVIDIA GTX 980|0.59it/s|1.68s/it|CUDA|ComfyUI| FP32 CPU TE
@@ -65,12 +56,14 @@ The following are the models used for testing. The models you use can be the sam
65
  ### 1536px
66
  | Chip | It/s | s/it | Backend | App | Notes
67
  |-------------|----|----|------|------|------|
 
68
  |NVIDIA RTX 3090|1.63it/s|0.61s/it|CUDA|ComfyUI
69
 
70
  ### 1024px
71
  Runs on 2GB of VRAM with tiled VAE.
72
  | Chip | It/s | s/it | Backend | App | Notes
73
  |-------------|----|----|------|------|------|
 
74
  |NVIDIA RTX 3090|4.00it/s|0.25s/it|CUDA|ComfyUI
75
  |NVIDIA GTX 980|0.18it/s|5.35s/it|CUDA|ComfyUI|
76
  |AMD Radeon Pro W5500|0.13it/s|7.35s/it|Vulkan|KoboldCPP
@@ -83,6 +76,7 @@ Runs on 2GB of VRAM with tiled VAE.
83
 
84
  | Chip | It/s | s/it | Backend | App | Notes
85
  |-------------|----|----|------|------|------|
 
86
  |NVIDIA RTX 3090|12.39it/s|0.0807s/it|CUDA|ComfyUI
87
  |NVIDIA GTX 980|0.69it/s|1.45s/it|CUDA|ComfyUI|
88
  |AMD Radeon Pro W5500|0.54it/s|1.85s/it|Vulkan|KoboldCPP|Windows 11
 
 
 
 
 
 
 
 
 
 
 
1
  # Methodology
2
  All settings use:
3
  - Euler sampler
 
10
 
11
  The given speed value (it/s or s/it) is used, and then extrapolated using the formula `1/speed` to get the other value. If its under 0.01 then it will be expanded to four digits compared to the usual 2
12
 
 
 
13
  ## Models Used
14
  The following are the models used for testing. The models you use can be the same architecture as the tested models
15
 
 
25
  ### 1024px
26
  | Chip | It/s | s/it | Backend | App | Notes
27
  |-------------|----|----|------|------|------|
28
+ |NVIDIA RTX 5090|2.58it/s|0.39s/it|CUDA 12.8|ComfyUI
29
  |NVIDIA RTX 3090|1.14it/s|0.87s/it|CUDA|ComfyUI|Sage Attention
30
  |NVIDIA RTX 3090|1.00it/s|1.00s/it|CUDA|ComfyUI
31
  |NVIDIA GTX 980|0.0599it/s|16.69s/it|CUDA|ComfyUI| FP32 CPU TE
 
35
  ### 512px
36
  | Chip | It/s | s/it | Backend | App | Notes
37
  |-------------|----|----|------|------|------|
38
+ |NVIDIA RTX 5090|8.85it/s|0.11s/it|CUDA 12.8|ComfyUI
39
  |NVIDIA RTX 3090|4.35it/s|0.23s/it|CUDA|ComfyUI|Sage Attention
40
  |NVIDIA GTX 980|0.28it/s|3.57s/it|CUDA|ComfyUI| FP8 CPU TE
41
  |NVIDIA GTX 980|0.25it/s|3.99s/it|CUDA|ComfyUI| FP32 CPU TE
 
44
  ### 256px
45
  | Chip | It/s | s/it | Backend | App | Notes
46
  |-------------|----|----|------|------|------|
47
+ |NVIDIA RTX 5090|13.42it/s|0.075s/it|CUDA 12.8|ComfyUI
48
  |NVIDIA RTX 3090|11.37it/s|0.0880s/it|CUDA|ComfyUI|Sage Attention
49
  |NVIDIA GTX 980|0.78it/s|1.27s/it|CUDA|ComfyUI| FP8 CPU TE
50
  |NVIDIA GTX 980|0.59it/s|1.68s/it|CUDA|ComfyUI| FP32 CPU TE
 
56
  ### 1536px
57
  | Chip | It/s | s/it | Backend | App | Notes
58
  |-------------|----|----|------|------|------|
59
+ |NVIDIA RTX 5090|3.38it/s|0.296s/it|CUDA 12.8|ComfyUI
60
  |NVIDIA RTX 3090|1.63it/s|0.61s/it|CUDA|ComfyUI
61
 
62
  ### 1024px
63
  Runs on 2GB of VRAM with tiled VAE.
64
  | Chip | It/s | s/it | Backend | App | Notes
65
  |-------------|----|----|------|------|------|
66
+ |NVIDIA RTX 5090|8.95it/s|0.11s/it|CUDA 12.8|ComfyUI
67
  |NVIDIA RTX 3090|4.00it/s|0.25s/it|CUDA|ComfyUI
68
  |NVIDIA GTX 980|0.18it/s|5.35s/it|CUDA|ComfyUI|
69
  |AMD Radeon Pro W5500|0.13it/s|7.35s/it|Vulkan|KoboldCPP
 
76
 
77
  | Chip | It/s | s/it | Backend | App | Notes
78
  |-------------|----|----|------|------|------|
79
+ |NVIDIA RTX 5090|21.52it/s|0.046s/it|CUDA 12.8|ComfyUI
80
  |NVIDIA RTX 3090|12.39it/s|0.0807s/it|CUDA|ComfyUI
81
  |NVIDIA GTX 980|0.69it/s|1.45s/it|CUDA|ComfyUI|
82
  |AMD Radeon Pro W5500|0.54it/s|1.85s/it|Vulkan|KoboldCPP|Windows 11