VCK5000 Performance with an 8PE DPUCVDX8H @350MHz - 3.0 English

Vitis AI Library User Guide (UG1354)
Document ID
UG1354
Release Date
2023-01-12
Version
3.0 English

The following table lists the throughput performance (in frames/sec or fps) for various neural network samples on the Versal ACAP VCK5000 Gen4x8 with DPUCVDX8H running at 8PE@350 MHz.

Table 1. VCK5000 Performance with an 8PE DPUCVDX8H @350MHz
No Neural Network Input Size GOPS DPU Frequency (MHz) Performance (fps) (Multiple thread)
1 chen_color_resnet18_pt 224x224 3.627 350 8433.2
2 drunet_pt 528x608 2.59 350 200.4
3 ENet_cityscapes_pt 512x1024 8.6 350 175.1
4 fadnet 576x960 441 350 13.4
5 inception_v1_pruned_0_087_tf 224x224 2.73 350 4587.5
6 inception_v1_pruned_0_157_tf 224x224 2.52 350 4693.2
7 inception_v1_tf 224x224 3 350 4202.4
8 medical_seg_cell_tf2 128x128 5.3 350 1452.2
9 MLPerf_resnet50_v1.5_tf 224x224 8.19 350 4521.6
10 mlperf_ssd_resnet34_tf 1200x1200 433 350 91.0
11 ofa_rcan_latency_pt 360x640 45.7 350 66.2
12 ofa_resnet50_baseline_pt 224x224 15 350 1511.6
13 ofa_resnet50_pruned_0_45_pt 224x224 8.2 350 2306.8
14 ofa_resnet50_pruned_0_60_pt 224x224 6 350 2243.6
15 rcan_pruned_tf 360x640 86.95 350 71.1
16 refinedet_VOC_tf 320x320 81.9 350 397.5
17 RefineDet-Medical_EDD_baseline_tf 320x320 81.28 350 398.2
18 RefineDet-Medical_EDD_pruned_0_5_tf 320x320 41.42 350 687.1
19 RefineDet-Medical_EDD_pruned_0_75_tf 320x320 20.54 350 794.1
20 RefineDet-Medical_EDD_pruned_0_85_tf 320x320 12.32 350 1200.1
21 RefineDet-Medical_EDD_tf 320x320 9.8 350 1247.6
22 resnet_v1_101_pruned_0_346_tf 224x224 9.4 350 3340.1
23 resnet_v1_101_pruned_0_568_tf 224x224 6.21 350 4145.6
24 resnet_v1_101_tf 224x224 14.4 350 2971.0
25 resnet_v1_152_pruned_0_51_tf 224x224 10.68 350 2769.6
26 resnet_v1_152_pruned_0_60_tf 224x224 8.82 350 3064.3
27 resnet_v1_152_tf 224x224 21.8 350 2109.9
28 resnet_v1_50_pruned_0_38_tf 224x224 4.3 350 5358.6
29 resnet_v1_50_pruned_0_65_tf 224x224 2.45 350 7214.6
30 resnet_v1_50_tf 224x224 7 350 4972.4
31 resnet50_pruned_0_3_pt 224x224 5.8 350 4648.8
32 resnet50_pruned_0_4_pt 224x224 4.9 350 4946.0
33 resnet50_pruned_0_5_pt 224x224 4.1 350 5295.8
34 resnet50_pruned_0_6_pt 224x224 3.3 350 5826.4
35 resnet50_pruned_0_7_pt 224x224 2.5 350 6399.7
36 resnet50_pt 224x224 4.1 350 4548.2
37 resnet50_tf2 224x224 7.7 350 4970.0
38 salsanext_pt 64x2048 20.4 350 258.1
39 salsanext_v2_pt 64x2048 32 350 97.9
40 semantic_seg_citys_tf2 512x1024 54 350 118.9
41 SemanticFPN_cityscapes_pt 256x512 10 350 1072.7
42 SESR_S_pt 360x640 7.48 350 234.6
43 squeezenet_pt 224x224 0.82 350 7664.4
44 ssd_resnet_50_fpn_coco_tf 640x640 178.4 350 113.9
45 vehicle_make_resnet18_pt 224x224 3.627 350 8419.5
46 vehicle_type_resnet18_pt 224x224 3.627 350 8442.7
47 xilinxSR_pt 360x640 182.4 350 14.4
48 yolov3_coco_416_tf2 416x416 65.9 350 472.6
49 yolov3_voc_tf 416x416 65.6 350 478.7