Skip to content

Commit f14388a

Browse files
authored
Add the benchmark of Jetson AGX Orin (#3733)
1 parent 3f6fbcc commit f14388a

File tree

1 file changed

+241
-0
lines changed

1 file changed

+241
-0
lines changed

benchmark/README.md

Lines changed: 241 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -71,6 +71,247 @@ echo "performance" > /sys/class/misc/mali0/device/devfreq/ff9a0000.gpu/governor
7171

7272
Typical output (executed in android adb shell)
7373

74+
### NVIDIA Jetson AGX Orin (Cortex-A78AE 2.2 GHz x 12 + [email protected] GHz Tensor Cores 64)
75+
```
76+
i@orin:~/projects/ncnn/benchmark$ ./benchncnn 64 1 0 -1 0
77+
loop_count = 64
78+
num_threads = 1
79+
powersave = 0
80+
gpu_device = -1
81+
cooling_down = 0
82+
squeezenet min = 11.66 max = 11.80 avg = 11.74
83+
squeezenet_int8 min = 12.24 max = 12.39 avg = 12.31
84+
mobilenet min = 19.56 max = 19.73 avg = 19.65
85+
mobilenet_int8 min = 16.06 max = 16.25 avg = 16.14
86+
mobilenet_v2 min = 13.20 max = 13.41 avg = 13.29
87+
mobilenet_v3 min = 11.39 max = 11.57 avg = 11.48
88+
shufflenet min = 8.07 max = 8.18 avg = 8.11
89+
shufflenet_v2 min = 8.41 max = 8.51 avg = 8.45
90+
mnasnet min = 12.74 max = 12.91 avg = 12.79
91+
proxylessnasnet min = 15.18 max = 15.32 avg = 15.25
92+
efficientnet_b0 min = 26.86 max = 26.96 avg = 26.90
93+
efficientnetv2_b0 min = 35.99 max = 36.15 avg = 36.07
94+
regnety_400m min = 16.81 max = 16.98 avg = 16.87
95+
blazeface min = 4.25 max = 4.37 avg = 4.29
96+
googlenet min = 48.73 max = 48.98 avg = 48.87
97+
googlenet_int8 min = 47.39 max = 47.60 avg = 47.49
98+
resnet18 min = 30.93 max = 31.24 avg = 31.08
99+
resnet18_int8 min = 55.44 max = 55.70 avg = 55.56
100+
alexnet min = 44.19 max = 44.43 avg = 44.33
101+
vgg16 min = 173.94 max = 174.97 avg = 174.46
102+
vgg16_int8 min = 475.10 max = 479.37 avg = 477.33
103+
resnet50 min = 89.50 max = 90.11 avg = 89.80
104+
resnet50_int8 min = 106.77 max = 107.14 avg = 106.96
105+
squeezenet_ssd min = 37.78 max = 38.35 avg = 37.93
106+
squeezenet_ssd_int8 min = 50.48 max = 50.88 avg = 50.74
107+
mobilenet_ssd min = 45.62 max = 46.12 avg = 45.74
108+
mobilenet_ssd_int8 min = 37.77 max = 38.00 avg = 37.88
109+
mobilenet_yolo min = 90.23 max = 90.49 avg = 90.35
110+
mobilenetv2_yolov3 min = 47.27 max = 47.48 avg = 47.33
111+
yolov4-tiny min = 60.41 max = 60.75 avg = 60.57
112+
nanodet_m min = 19.26 max = 19.43 avg = 19.35
113+
yolo-fastest-1.1 min = 8.16 max = 8.31 avg = 8.20
114+
yolo-fastestv2 min = 8.26 max = 8.39 avg = 8.32
115+
i@orin:~/projects/ncnn/benchmark$ ./benchncnn 64 2 0 -1 0
116+
loop_count = 64
117+
num_threads = 2
118+
powersave = 0
119+
gpu_device = -1
120+
cooling_down = 0
121+
squeezenet min = 6.83 max = 6.98 avg = 6.90
122+
squeezenet_int8 min = 7.39 max = 7.50 avg = 7.45
123+
mobilenet min = 10.40 max = 10.50 avg = 10.45
124+
mobilenet_int8 min = 8.92 max = 9.09 avg = 8.99
125+
mobilenet_v2 min = 7.67 max = 7.80 avg = 7.74
126+
mobilenet_v3 min = 6.86 max = 7.01 avg = 6.93
127+
shufflenet min = 6.34 max = 6.44 avg = 6.39
128+
shufflenet_v2 min = 5.71 max = 5.83 avg = 5.76
129+
mnasnet min = 7.47 max = 7.58 avg = 7.53
130+
proxylessnasnet min = 8.73 max = 8.83 avg = 8.78
131+
efficientnet_b0 min = 14.93 max = 15.13 avg = 15.03
132+
efficientnetv2_b0 min = 20.17 max = 20.70 avg = 20.29
133+
regnety_400m min = 12.50 max = 12.62 avg = 12.57
134+
blazeface min = 2.95 max = 3.06 avg = 3.00
135+
googlenet min = 26.25 max = 26.53 avg = 26.37
136+
googlenet_int8 min = 26.54 max = 26.79 avg = 26.66
137+
resnet18 min = 16.69 max = 16.90 avg = 16.80
138+
resnet18_int8 min = 29.70 max = 29.93 avg = 29.81
139+
alexnet min = 22.96 max = 23.12 avg = 23.03
140+
vgg16 min = 88.39 max = 89.16 avg = 88.79
141+
vgg16_int8 min = 245.86 max = 247.55 avg = 246.62
142+
resnet50 min = 46.55 max = 46.86 avg = 46.70
143+
resnet50_int8 min = 56.28 max = 56.63 avg = 56.43
144+
squeezenet_ssd min = 23.65 max = 24.29 avg = 23.81
145+
squeezenet_ssd_int8 min = 30.86 max = 31.27 avg = 30.99
146+
mobilenet_ssd min = 25.17 max = 25.31 avg = 25.24
147+
mobilenet_ssd_int8 min = 21.77 max = 21.97 avg = 21.84
148+
mobilenet_yolo min = 48.03 max = 48.33 avg = 48.14
149+
mobilenetv2_yolov3 min = 26.58 max = 26.81 avg = 26.66
150+
yolov4-tiny min = 35.31 max = 35.53 avg = 35.41
151+
nanodet_m min = 12.93 max = 13.08 avg = 13.01
152+
yolo-fastest-1.1 min = 6.00 max = 6.10 avg = 6.04
153+
yolo-fastestv2 min = 6.46 max = 6.61 avg = 6.52
154+
i@orin:~/projects/ncnn/benchmark$ ./benchncnn 64 4 0 -1 0
155+
loop_count = 64
156+
num_threads = 4
157+
powersave = 0
158+
gpu_device = -1
159+
cooling_down = 0
160+
squeezenet min = 4.54 max = 4.84 avg = 4.61
161+
squeezenet_int8 min = 4.96 max = 5.41 avg = 5.05
162+
mobilenet min = 5.96 max = 6.23 avg = 6.04
163+
mobilenet_int8 min = 5.21 max = 5.50 avg = 5.30
164+
mobilenet_v2 min = 5.05 max = 5.26 avg = 5.15
165+
mobilenet_v3 min = 4.83 max = 5.14 avg = 4.90
166+
shufflenet min = 5.11 max = 5.34 avg = 5.18
167+
shufflenet_v2 min = 4.13 max = 4.44 avg = 4.18
168+
mnasnet min = 4.93 max = 5.27 avg = 5.01
169+
proxylessnasnet min = 5.64 max = 5.89 avg = 5.72
170+
efficientnet_b0 min = 9.47 max = 10.60 avg = 9.60
171+
efficientnetv2_b0 min = 12.67 max = 13.06 avg = 12.82
172+
regnety_400m min = 10.27 max = 10.58 avg = 10.38
173+
blazeface min = 2.05 max = 2.27 avg = 2.10
174+
googlenet min = 15.57 max = 15.96 avg = 15.68
175+
googlenet_int8 min = 16.19 max = 16.65 avg = 16.32
176+
resnet18 min = 10.20 max = 11.76 avg = 10.35
177+
resnet18_int8 min = 16.89 max = 17.31 avg = 17.03
178+
alexnet min = 13.13 max = 13.70 avg = 13.32
179+
vgg16 min = 51.03 max = 52.46 avg = 51.35
180+
vgg16_int8 min = 131.08 max = 139.44 avg = 133.78
181+
resnet50 min = 26.74 max = 28.32 avg = 26.91
182+
resnet50_int8 min = 32.15 max = 32.74 avg = 32.38
183+
squeezenet_ssd min = 16.58 max = 16.99 avg = 16.70
184+
squeezenet_ssd_int8 min = 20.22 max = 21.67 avg = 20.51
185+
mobilenet_ssd min = 14.68 max = 16.07 avg = 14.83
186+
mobilenet_ssd_int8 min = 12.89 max = 13.27 avg = 13.01
187+
mobilenet_yolo min = 28.44 max = 28.85 avg = 28.58
188+
mobilenetv2_yolov3 min = 17.21 max = 21.31 avg = 17.44
189+
yolov4-tiny min = 23.68 max = 24.38 avg = 23.88
190+
nanodet_m min = 8.76 max = 9.17 avg = 8.86
191+
yolo-fastest-1.1 min = 4.83 max = 5.04 avg = 4.88
192+
yolo-fastestv2 min = 4.93 max = 5.17 avg = 5.00
193+
i@orin:~/projects/ncnn/benchmark$ ./benchncnn 64 8 0 -1 0
194+
loop_count = 64
195+
num_threads = 8
196+
powersave = 0
197+
gpu_device = -1
198+
cooling_down = 0
199+
squeezenet min = 3.52 max = 4.28 avg = 3.65
200+
squeezenet_int8 min = 3.85 max = 4.11 avg = 3.93
201+
mobilenet min = 3.78 max = 4.12 avg = 3.85
202+
mobilenet_int8 min = 3.57 max = 3.85 avg = 3.63
203+
mobilenet_v2 min = 4.14 max = 4.44 avg = 4.22
204+
mobilenet_v3 min = 3.89 max = 4.26 avg = 3.97
205+
shufflenet min = 4.78 max = 4.95 avg = 4.84
206+
shufflenet_v2 min = 3.49 max = 3.84 avg = 3.54
207+
mnasnet min = 3.94 max = 4.09 avg = 3.99
208+
proxylessnasnet min = 4.41 max = 4.68 avg = 4.47
209+
efficientnet_b0 min = 7.01 max = 7.85 avg = 7.13
210+
efficientnetv2_b0 min = 9.22 max = 9.46 avg = 9.32
211+
regnety_400m min = 9.34 max = 9.66 avg = 9.44
212+
blazeface min = 1.86 max = 1.98 avg = 1.89
213+
googlenet min = 10.37 max = 10.76 avg = 10.48
214+
googlenet_int8 min = 11.03 max = 11.34 avg = 11.16
215+
resnet18 min = 6.83 max = 7.12 avg = 6.93
216+
resnet18_int8 min = 10.25 max = 11.50 avg = 10.42
217+
alexnet min = 8.88 max = 9.71 avg = 9.01
218+
vgg16 min = 31.26 max = 31.97 avg = 31.44
219+
vgg16_int8 min = 71.31 max = 74.53 avg = 72.18
220+
resnet50 min = 16.43 max = 16.84 avg = 16.52
221+
resnet50_int8 min = 19.07 max = 20.28 avg = 19.42
222+
squeezenet_ssd min = 13.50 max = 13.69 avg = 13.56
223+
squeezenet_ssd_int8 min = 15.16 max = 16.06 avg = 15.30
224+
mobilenet_ssd min = 9.73 max = 10.85 avg = 9.90
225+
mobilenet_ssd_int8 min = 9.27 max = 9.46 avg = 9.36
226+
mobilenet_yolo min = 17.58 max = 17.79 avg = 17.67
227+
mobilenetv2_yolov3 min = 12.80 max = 13.50 avg = 12.90
228+
yolov4-tiny min = 17.98 max = 21.31 avg = 18.24
229+
nanodet_m min = 7.01 max = 7.18 avg = 7.09
230+
yolo-fastest-1.1 min = 4.76 max = 4.86 avg = 4.80
231+
yolo-fastestv2 min = 4.76 max = 4.88 avg = 4.82
232+
i@orin:~/projects/ncnn/benchmark$ ./benchncnn 64 12 0 -1 0
233+
loop_count = 64
234+
num_threads = 12
235+
powersave = 0
236+
gpu_device = -1
237+
cooling_down = 0
238+
squeezenet min = 3.50 max = 5.21 avg = 3.65
239+
squeezenet_int8 min = 3.97 max = 4.44 avg = 4.12
240+
mobilenet min = 3.49 max = 7.73 avg = 3.78
241+
mobilenet_int8 min = 3.40 max = 3.86 avg = 3.49
242+
mobilenet_v2 min = 4.07 max = 4.39 avg = 4.17
243+
mobilenet_v3 min = 3.92 max = 4.17 avg = 4.03
244+
shufflenet min = 5.08 max = 6.63 avg = 5.18
245+
shufflenet_v2 min = 3.64 max = 5.11 avg = 3.75
246+
mnasnet min = 3.86 max = 4.16 avg = 3.95
247+
proxylessnasnet min = 4.30 max = 5.39 avg = 4.38
248+
efficientnet_b0 min = 6.42 max = 9.19 avg = 6.61
249+
efficientnetv2_b0 min = 8.96 max = 9.43 avg = 9.12
250+
regnety_400m min = 10.11 max = 10.89 avg = 10.27
251+
blazeface min = 1.93 max = 2.16 avg = 1.99
252+
googlenet min = 9.72 max = 10.84 avg = 10.01
253+
googlenet_int8 min = 10.91 max = 13.03 avg = 11.17
254+
resnet18 min = 6.70 max = 7.27 avg = 6.92
255+
resnet18_int8 min = 9.62 max = 12.93 avg = 10.14
256+
alexnet min = 7.21 max = 7.47 avg = 7.32
257+
vgg16 min = 29.61 max = 63.73 avg = 30.86
258+
vgg16_int8 min = 64.91 max = 75.06 avg = 68.72
259+
resnet50 min = 15.35 max = 16.28 avg = 15.73
260+
resnet50_int8 min = 17.47 max = 18.98 avg = 18.09
261+
squeezenet_ssd min = 13.40 max = 28.74 avg = 14.07
262+
squeezenet_ssd_int8 min = 15.35 max = 16.77 avg = 15.67
263+
mobilenet_ssd min = 9.51 max = 11.49 avg = 9.88
264+
mobilenet_ssd_int8 min = 9.43 max = 10.08 avg = 9.58
265+
mobilenet_yolo min = 16.88 max = 17.45 avg = 17.09
266+
mobilenetv2_yolov3 min = 11.91 max = 31.90 avg = 12.50
267+
yolov4-tiny min = 17.85 max = 18.87 avg = 18.36
268+
nanodet_m min = 6.88 max = 7.64 avg = 7.06
269+
yolo-fastest-1.1 min = 5.02 max = 5.53 avg = 5.12
270+
yolo-fastestv2 min = 4.95 max = 5.60 avg = 5.05
271+
i@orin:~/projects/ncnn/benchmark$ ./benchncnn 128 1 0 0 0
272+
[0 NVIDIA Tegra Orin (nvgpu)] queueC=2[8] queueG=0[16] queueT=1[2]
273+
[0 NVIDIA Tegra Orin (nvgpu)] bugsbn1=0 bugbilz=0 bugcopc=0 bugihfa=0
274+
[0 NVIDIA Tegra Orin (nvgpu)] fp16-p/s/a=1/1/1 int8-p/s/a=1/1/1
275+
[0 NVIDIA Tegra Orin (nvgpu)] subgroup=32 basic=1 vote=1 ballot=1 shuffle=1
276+
loop_count = 128
277+
num_threads = 1
278+
powersave = 0
279+
gpu_device = 0
280+
cooling_down = 0
281+
squeezenet min = 2.13 max = 3.37 avg = 2.31
282+
squeezenet_int8 min = 12.31 max = 12.51 avg = 12.42
283+
mobilenet min = 2.03 max = 2.73 avg = 2.23
284+
mobilenet_int8 min = 16.86 max = 17.91 avg = 16.99
285+
mobilenet_v2 min = 2.59 max = 3.59 avg = 2.91
286+
mobilenet_v3 min = 3.22 max = 4.23 avg = 3.71
287+
shufflenet min = 2.57 max = 3.27 avg = 2.80
288+
shufflenet_v2 min = 3.20 max = 4.03 avg = 3.47
289+
mnasnet min = 2.45 max = 3.06 avg = 2.69
290+
proxylessnasnet min = 2.50 max = 3.14 avg = 2.72
291+
efficientnet_b0 min = 4.23 max = 8.73 avg = 4.85
292+
efficientnetv2_b0 min = 8.15 max = 8.60 avg = 8.41
293+
regnety_400m min = 3.25 max = 4.17 avg = 3.54
294+
blazeface min = 1.29 max = 1.48 avg = 1.33
295+
googlenet min = 4.95 max = 12.34 avg = 6.36
296+
googlenet_int8 min = 47.49 max = 47.78 avg = 47.61
297+
resnet18 min = 3.18 max = 9.49 avg = 4.04
298+
resnet18_int8 min = 55.57 max = 55.88 avg = 55.73
299+
alexnet min = 3.22 max = 14.56 avg = 4.25
300+
vgg16 min = 6.82 max = 14.75 avg = 8.18
301+
vgg16_int8 min = 473.55 max = 479.07 avg = 476.22
302+
resnet50 min = 4.75 max = 15.06 avg = 6.08
303+
resnet50_int8 min = 106.99 max = 107.48 avg = 107.22
304+
squeezenet_ssd min = 6.87 max = 9.12 avg = 7.76
305+
squeezenet_ssd_int8 min = 50.87 max = 51.17 avg = 51.01
306+
mobilenet_ssd min = 4.44 max = 6.22 avg = 5.23
307+
mobilenet_ssd_int8 min = 37.80 max = 38.03 avg = 37.92
308+
mobilenet_yolo min = 5.41 max = 7.36 avg = 6.29
309+
mobilenetv2_yolov3 min = 7.20 max = 9.96 avg = 7.30
310+
yolov4-tiny min = 16.48 max = 28.81 avg = 18.40
311+
nanodet_m min = 5.75 max = 8.54 avg = 6.85
312+
yolo-fastest-1.1 min = 4.03 max = 4.75 avg = 4.35
313+
yolo-fastestv2 min = 4.27 max = 5.23 avg = 4.71
314+
```
74315

75316
### AMD Ryzen Threadripper 3970X (Zen2 3.7 GHz ~ 4.5 GHz x 32)
76317
```

0 commit comments

Comments
 (0)