Geekbench AI Score

1117
Single Precision Score
1122
Half Precision Score
747
Quantized Score
Geekbench AI 1.4.0 for Linux AVX2

Result Information

User nikhilr
Upload Date August 15 2025 04:39 AM
Views 9

System Information

AI Information
Framework TensorFlow Lite
Backend CPU
Device AMD EPYC 7K62 48-Core Processor
System Information
Operating System Debian GNU/Linux 12 (bookworm)
Model HUANANZHI H12D-8D
Motherboard HUANANZHI H12D-8D
CPU Information
Name AMD EPYC 7K62 48-Core Processor
Topology 1 Processor, 48 Cores, 96 Threads
Identifier AuthenticAMD Family 23 Model 49 Stepping 0
Base Frequency 3.31 GHz
Cluster 10 Cores
Memory Information
Size 503.54 GB

AI Performance

Workload Accuracy Score
Image Classification (SP)
100% 886
164.7 IPS
 
Image Classification (HP)
100% 896
166.7 IPS
 
Image Classification (Q)
99% 568
105.9 IPS
 
Image Segmentation (SP)
100% 1555
25.2 IPS
 
Image Segmentation (HP)
100% 1515
24.6 IPS
 
Image Segmentation (Q)
98% 573
9.32 IPS
 
Pose Estimation (SP)
100% 1932
2.25 IPS
 
Pose Estimation (HP)
100% 1968
2.30 IPS
 
Pose Estimation (Q)
96% 2686
3.15 IPS
 
Object Detection (SP)
100% 860
68.2 IPS
 
Object Detection (HP)
100% 866
68.7 IPS
 
Object Detection (Q)
85% 707
56.9 IPS
 
Face Detection (SP)
100% 1441
17.1 IPS
 
Face Detection (HP)
100% 1470
17.5 IPS
 
Face Detection (Q)
97% 737
8.79 IPS
 
Depth Estimation (SP)
100% 1299
10.0 IPS
 
Depth Estimation (HP)
99% 1302
10.0 IPS
 
Depth Estimation (Q)
63% 893
8.37 IPS
 
Style Transfer (SP)
100% 1839
2.36 IPS
 
Style Transfer (HP)
100% 1827
2.35 IPS
 
Style Transfer (Q)
98% 3189
4.11 IPS
 
Image Super-Resolution (SP)
100% 775
28.6 IPS
 
Image Super-Resolution (HP)
100% 774
28.6 IPS
 
Image Super-Resolution (Q)
97% 1113
41.3 IPS
 
Text Classification (SP)
100% 332
443.0 IPS
 
Text Classification (HP)
100% 341
455.6 IPS
 
Text Classification (Q)
92% 183
245.9 IPS
 
Machine Translation (SP)
100% 1501
25.9 IPS
 
Machine Translation (HP)
100% 1482
25.5 IPS
 
Machine Translation (Q)
58% 205
4.82 IPS
 

Compare

Set Baseline Geekbench 6 now available