Geekbench ML Score

327
TensorFlow Lite NNAPI Inference Score
Geekbench ML 0.6.0 Corporate for Android AArch64

Result Information

Upload Date December 08 2023 02:34 PM
Views 1

System Information

System Information
Operating System Android 13
Model Google Pixel 5
Model IDGoogle Pixel 5
Motherboard redfin
Governorschedutil
CPU Information
Name Qualcomm Qualcomm
Topology 1 Processor, 8 Cores
Identifier ARM implementer 81 architecture 8 variant 15 part 2052 revision 14
Base Frequency 1.80 GHz
Cluster 16 Cores @ 1.80 GHz
Cluster 21 Core @ 2.21 GHz
Cluster 31 Core @ 2.40 GHz
Memory Information
Size 7.29 GB
Inference Information
Framework TensorFlow Lite
Backend NNAPI
Device Default

Inference Performance

Workload Accuracy Score
Image Classification (F32)
100% 175
32.8 IPS
 
Image Classification (F16)
99% 247
46.2 IPS
 
Image Classification (I8)
97% 809
151.4 IPS
 
Image Segmentation (F32)
100% 232
3.87 IPS
 
Image Segmentation (F16)
100% 306
5.12 IPS
 
Image Segmentation (I8)
98% 851
14.2 IPS
 
Pose Estimation (F32)
100% 500
0.61 IPS
 
Pose Estimation (F16)
100% 1246
1.51 IPS
 
Pose Estimation (I8)
100% 6504
7.88 IPS
 
Object Detection (F32)
100% 64
4.78 IPS
 
Object Detection (F16)
100% 71
5.28 IPS
 
Object Detection (I8)
59% 87
6.51 IPS
 
Face Detection (F32)
100% 439
5.21 IPS
 
Face Detection (F16)
97% 532
6.33 IPS
 
Face Detection (I8)
88% 2375
28.2 IPS
 
Depth Estimation (F32)
100% 327
2.53 IPS
 
Depth Estimation (F16)
100% 392
3.04 IPS
 
Depth Estimation (I8)
95% 4649
36.1 IPS
 
Style Transfer (F32)
100% 521
0.69 IPS
 
Style Transfer (F16)
100% 570
0.75 IPS
 
Style Transfer (I8)
98% 2630
3.46 IPS
 
Image Super-Resolution (F32)
100% 306
10.9 IPS
 
Image Super-Resolution (F16)
100% 584
20.9 IPS
 
Image Super-Resolution (I8)
98% 2978
106.3 IPS
 
Text Classification (F32)
64% 3
4.12 IPS
 
Text Classification (F16)
64% 3
4.04 IPS
 
Text Classification (I8)
94% 94
135.6 IPS
 
Machine Translation (F32)
100% 112
2.06 IPS
 
Machine Translation (F16)
100% 285
5.24 IPS
 
Machine Translation (I8)
58% 218
4.02 IPS
 

Compare

Set Baseline Geekbench 6 now available