VCK5000 Performance with a 6PE DPUCVDX8H-aieMISC @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-aieMISC running at 6PE@350 MHz.

Table 1. VCK5000 Performance with a 6PE DPUCVDX8H-aieMISC @350MHz
No Neural Network Input Size GOPS DPU Frequency (MHz) Performance (fps) (Multiple thread)
1 chen_color_resnet18_pt 224x224 3.627 350 5210.5
2 clocs 12000x100x4 41 350 23.0
3 drunet_pt 528x608 2.59 350 133.2
4 ENet_cityscapes_pt 512x1024 8.6 350 89.9
5 fadnet 576x960 441 350 11.0
6 inception_resnet_v2_tf 299x299 26.4 350 426.3
7 inception_v1_pruned_0_087_tf 224x224 2.73 350 2285.6
8 inception_v1_pruned_0_157_tf 224x224 2.52 350 2317.6
9 inception_v1_tf 224x224 3 350 2157.0
10 inception_v3_pt 299x299 5.7 350 717.5
11 inception_v3_pruned_0_3_pt 299x299 8 350 814.8
12 inception_v3_pruned_0_4_pt 299x299 6.8 350 871.8
13 inception_v3_pruned_0_5_pt 299x299 5.7 350 936.9
14 inception_v3_pruned_0_6_pt 299x299 4.5 350 1060.5
15 inception_v3_tf 299x299 11.5 350 713.5
16 inception_v3_pruned_0_2_tf 299x299 9.1 350 769.9
17 inception_v3_pruned_0_4_tf 299x299 6.9 350 748.0
18 inception_v3_tf2 299x299 11.5 350 711.8
19 inception_v4_2016_09_09_tf 299x299 24.6 350 398.4
20 inception_v4_pruned_0_2_tf 299x299 19.56 350 397.4
21 inception_v4_pruned_0_4_tf 299x299 14.79 350 464.2
22 medical_seg_cell_tf2 128x128 5.3 350 1167.5
23 MLPerf_resnet50_v1.5_tf 224x224 8.19 350 2216.3
24 mlperf_ssd_resnet34_tf 1200x1200 433 350 70.5
25 ofa_rcan_latency_pt 360x640 45.7 350 3182.6
26 ofa_resnet50_0_9B_pt 160x160 0.9 350 3187.0
27 ofa_resnet50_baseline_pt 224x224 15 350 979.5
28 ofa_resnet50_pruned_0_45_pt 224x224 8.2 350 1288.2
29 ofa_resnet50_pruned_0_60_pt 224x224 6 350 1292.7
30 ofa_resnet50_pruned_0_74_pt 192x192 3.6 350 2089.1
31 ofa_yolo_pruned_0_30_pt 640x640 34.71 350 365.8
32 ofa_yolo_pruned_0_50_pt 640x640 24.62 350 409.2
33 ofa_yolo_pt 640x640 48.88 350 251.6
34 pmg_pt 224x224 2.28 350 2948.9
35 pointpainting 40000x64x16 112 350 17.7
36 pointpillars_kitti_12000_pt 12000x100x4 10.8 350 3.1
37 rcan_pruned_tf 360x640 86.95 350 52.8
38 refinedet_VOC_tf 320x320 81.9 350 287.0
39 RefineDet-Medical_EDD_baseline_tf 320x320 81.28 350 287.3
40 RefineDet-Medical_EDD_pruned_0_5_tf 320x320 41.42 350 474.8
41 RefineDet-Medical_EDD_pruned_0_75_tf 320x320 20.54 350 565.8
42 RefineDet-Medical_EDD_pruned_0_85_tf 320x320 12.32 350 772.3
43 RefineDet-Medical_EDD_tf 320x320 9.8 350 795.1
44 resnet_v1_101_pruned_0_346_tf 224x224 9.4 350 1650.9
45 resnet_v1_101_pruned_0_568_tf 224x224 6.21 350 1834.6
46 resnet_v1_101_tf 224x224 14.4 350 1541.2
47 resnet_v1_152_pruned_0_51_tf 224x224 10.68 350 1269.2
48 resnet_v1_152_pruned_0_60_tf 224x224 8.82 350 3056.9
49 resnet_v1_152_tf 224x224 21.8 350 1102.3
50 resnet_v1_50_pruned_0_38_tf 224x224 4.3 350 2573.7
51 resnet_v1_50_pruned_0_65_tf 224x224 2.45 350 3056.9
52 resnet_v1_50_tf 224x224 7 350 2467.1
53 resnet50_pruned_0_3_pt 224x224 5.8 350 2293.8
54 resnet50_pruned_0_4_pt 224x224 4.9 350 2374.1
55 resnet50_pruned_0_5_pt 224x224 4.1 350 2459.5
56 resnet50_pruned_0_6_pt 224x224 3.3 350 2597.4
57 resnet50_pruned_0_7_pt 224x224 2.5 350 2732.6
58 resnet50_pt 224x224 4.1 350 2267.8
59 resnet50_tf2 224x224 7.7 350 2467.4
60 salsanext_pt 64x2048 20.4 350 149.8
61 salsanext_v2_pt 64x2048 32 350 77.5
62 semantic_seg_citys_tf2 512x1024 54 350 76.7
63 SemanticFPN_cityscapes_pt 256x512 10 350 926.6
64 SESR_S_pt 360x640 7.48 350 160.9
65 solo_pt 640x640 107 350 59.6
66 squeezenet_pt 224x224 0.82 350 4323.2
67 ssd_resnet_50_fpn_coco_tf 640x640 178.4 350 99.8
68 vehicle_make_resnet18_pt 224x224 3.627 350 5212.8
69 vehicle_type_resnet18_pt 224x224 3.627 350 5208.7
70 vgg_16_pruned_0_43_tf 224x224 17.67 350 888.1
71 vgg_16_pruned_0_5_tf 224x224 15.64 350 950.6
72 vgg_16_tf 224x224 31 350 475.7
73 vgg_19_pruned_0_24_tf 224x224 29.79 350 527.5
74 vgg_19_pruned_0_39_tf 224x224 23.78 350 669.6
75 vgg_19_tf 224x224 39.3 350 424.2
76 xilinxSR_pt 360x640 182.44 350 11.4
77 yolov3_coco_416_tf2 416x416 65.9 350 381.6
78 yolov3_voc_tf 416x416 65.6 350 385.4
79 yolov4_csp_pt 640x640 121 350 171.2
80 yolov4_leaky_416_tf 416x416 60.3 350 311.6
81 yolov4_leaky_512_tf 512x512 91.2 350 156.5
82 yolov5_large_pt 640x640 109.6 350 192.1
83 yolov5-nano_pt 640x640 4.6 350 843.2
84 yolov5s6_pt 640x640 17 350 157.5
85 yolov6m_pt 640x640 82.2 350 278.6