VCK190 Evaluation Board - 2.0 English

Vitis AI Library User Guide (UG1354)
Document ID
UG1354
Release Date
2022-01-20
Version
2.0 English

VCK190 is the first Versal AI Core series evaluation kit, enabling designers to develop solutions using AI and DSP engines capable of delivering over 100X greater compute performance compared to current server class CPUs. For this release, a C32B3 DPU core is implemented using AI Engine.

Refer to the following table for the throughput performance (in frames/sec or fps) for various neural network samples on VCK190 with DPU running at 1250 MHz.

Table 1. VCK190 Performance with Batch 3
No Neural Network Input Size GOPS Performance (fps) (Single thread) Performance (fps) (Multiple thread)
1 bcc_pt 800x1000 268.9 31.2 53.5
2 centerpoint_0_ptcenterpoint_1_pt 2560x40x4 54 113.5 218.8
3 densebox_320_320 320x320 0.49 1054.5 2311.2
4 densebox_640_360 360x640 1.1 515.7 1094
5 efficientnet-b0_tf2 224x224 0.36 441.3 540.3
6 efficientNet-edgetpu-L_tf 300x300 19.36 257.6 311.2
7 efficientNet-edgetpu-M_tf 240x240 7.34 587 801.6
8 efficientNet-edgetpu-S_tf 224x224 4.72 805.7 1205.1
9 ENet_cityscapes_pt 512x1024 8.6 24.1 55.3
10 face_landmark 96x72 0.14 6436.9 12013.9
11 face-quality 80x60 0.06 10145.5 25104.5
12 face-quality_pt 80x60 0.06 10135.7 25144.8
13 facerec_resnet20 112x96 3.5 2067.3 2598.3
14 facerec-resnet20_mixed_pt 112x96 3.5 2073.6 2600.1
15 facerec_resnet64 112x96 11 1080.7 1208.5
16 facereid-large_pt 96x96 0.5 5529.3 11537.5
17 facereid-small_pt 80x80 0.09 8575.4 21872
18 fpn 256x512 8.9 94.8 219.9
19 FPN_Res18_Medical_segmentation 320x320 45.3 68 188.2
20 FPN-resnet18_covid19-seg_pt 352x352 22.7 365.6 548.3
21 FPN-resnet18_Endov 240x320 13.75 105.7 241.3
22 hourglass-pe_mpii 256x256 10.2 73.8 139.1
23 inception_resnet_v2_tf 299x299 26.4 243.2 290.9
24 inception_v1 224x224 3.2 1015 1721.5
25 inception_v1_tf 224x224 3 1022.7 1754
26 inception_v2 224x224 4 833.3 1273.5
27 inception_v2_tf 224x224 3.88 566.9 739.8
28 inception_v3 299x299 11.4 422.6 589.1
29 inception_v3_pt 299x299 5.7 421.4 587.1
30 inception_v3_tf 299x299 11.5 422.8 590.4
31 inception_v3_tf2 299x299 11.5 439 623.6
32 inception_v4 299x299 24.5 240 285.8
33 inception_v4_2016_09_09_tf 299x299 24.6 241 287.4
34 medical_seg_cell_tf2 128x128 5.3 1164.9 1958.4
35 MLPerf_resnet50_v1.5_tf 224x224 8.19 827.8 1251
36 mlperf_ssd_resnet34_tf 1200x1200 433 10.8 23.4
37 mobilenet_1_0_224_tf2 224x224 1.1 1463.1 3166.2
38 mobilenet_edge_0_75_tf 224x224 0.62 1361.2 2911.7
39 mobilenet_edge_1_0_tf 224x224 0.99 1276 2597.2
40 mobilenet_v1_0_25_128_tf 128x128 0.027 4124.1 9165.5
41 mobilenet_v1_0_5_160_tf 160x160 0.15 2827.4 7082.6
42 mobilenet_v1_1_0_224_tf 224x224 1.1 1463.6 3245.7
43 mobilenet_v2 224x224 0.6 1348.4 2911.4
44 mobilenet_v2_1_0_224_tf 224x224 0.6 1336.2 2914.7
45 mobilenet_v2_1_4_224_tf 224x224 1.2 1104.2 2021.3
46 mobilenet_v2_cityscapes_tf 1024x2048 132.74 5.1 13.2
47 MT-resnet18_mixed_pt 512x320 13.65 126.4 276.9
48 multi_task 288x512 14.8 153.3 317.5
49 multi_task_v3_pt 320x512 25.44 73.6 190.1
50 openpose_pruned_0_3 368x368 49.9 20.9 39.9
51 personreid-res18_pt 176x80 1.1 2826.4 4673.2
52 personreid-res50_pt 256x128 5.4 1119.1 1712
53 plate_detection 320x320 0.49 1243.6 2591.6
54 plate_num 96x288 1.75 916.1 1557.1
55 pmg_pt 224x224 2.28 1176 1990.6
56 pointpainting-pointpainting_nuscenes_40000_64_0_ptpointpainting_nuscenes_40000_64_1_pt 40000x64x16 112 3.7 6.8
57 pointpillars_kitti_12000_0_ptpointpillars_kitti_12000_1_pt 12000x100x4 10.8 36.4 57.5
58 pointpillars_nuscenes_40000_64_0_ptpointpillars_nuscenes_40000_64_1_pt 40000x64x5 108 7.5 17.6
59 rcan_pruned_tf 360x640 86.95 46.4 59.5
60 refinedet_baseline 480x360 123 115.5 142.2
61 RefineDet-Medical_EDD_tf 320x320 9.8 413.7 887
62 refinedet_pruned_0_8 360x480 25 263.4 443.2
63 refinedet_pruned_0_92 360x480 10.1 334.6 673.7
64 refinedet_pruned_0_96 360x480 5.1 403 775.6
65 refinedet_VOC_tf 320x320 81.9 77.2 175.7
66 reid 80x160 0.95 2750.4 4856.9
67 resnet18 224x224 3.7 1310.6 2691
68 resnet50 224x224 7.7 898.4 1419.4
69 resnet50_pt 224x224 4.1 827.5 1252.6
70 resnet50_tf2 224x224 7.7 899.8 1422.3
71 resnet_v1_101_tf 224x224 14.4 591.4 781.6
72 resnet_v1_152_tf 224x224 21.8 440.6 538.7
73 resnet_v1_50_tf 224x224 7 895.9 1421.1
74 resnet_v2_101_tf 299x299 26.78 267 324.1
75 resnet_v2_152_tf 299x299 40.47 201.9 232.8
76 resnet_v2_50_tf 299x299 13.1 393.3 530.8
77 retinaface 360x640 1.11 344.1 790.8
78 SA_gate_pt 360x360 59.71 14.4 24.2
79 salsanext_pt 64x2048 20.4 12.3 24.8
80 salsanext_v2_pt 64x2048 32 10.3 23.9
81 SemanticFPN_cityscapes_pt 256x512 10 102.1 221.5
82 SemanticFPN_Mobilenetv2_pt 512x1024 5.4 25.2 56
83 semantic_seg_citys_tf2 512x1024 54 20.1 52.2
84 sp_net 128x224 0.55 2161 4399.8
85 squeezenet 227x227 0.76 2584.8 4111.4
86 squeezenet_pt 224x224 0.82 2605.4 4256.9
87 ssd_adas_pruned_0_95 360x480 6.3 386.3 830.9
88 ssd_inception_v2_coco_tf 300x300 9.6 208.2 379
89 ssdlite_mobilenet_v2_coco_tf 300x300 1.5 347.6 513.8
90 ssd_mobilenet_v1_coco_tf 300x300 2.5 370.9 528.3
91 ssd_mobilenet_v2 360x480 6.6 77.8 172.6
92 ssd_mobilenet_v2_coco_tf 300x300 3.8 323.3 526
93 ssd_pedestrian_pruned_0_97 360x360 5.9 316.6 708.2
94 ssd_resnet_50_fpn_coco_tf 640x640 178.4 10.8 12.5
95 ssd_traffic_pruned_0_9 360x480 11.6 283.8 677.7
96 tiny_yolov3_vmss 416x416 5.46 536.1 1339.3
97 unet_chaos-CT_pt 512x512 23.3 86.5 217.3
98 vgg_16_tf 224x224 31 308.5 353.3
99 vgg_19_tf 224x224 39.3 276.1 311.6
100 vpgnet_pruned_0_99 480x640 2.5 272.1 678.8
101 yolov2_voc 448x448 34 273.7 432
102 yolov2_voc_pruned_0_66 448x448 11.6 384.7 793.9
103 yolov2_voc_pruned_0_71 448x448 9.9 406 893.1
104 yolov2_voc_pruned_0_77 448x448 7.8 425.5 992.4
105 yolov3_adas_pruned_0_9 256x512 5.5 425.4 944.6
106 yolov3_bdd 288x512 53.7 148.8 191.1
107 yolov3_voc 416x416 65.4 151.6 189.9
108 yolov3_voc_tf 416x416 65.6 151.9 189.7
109 yolov4_leaky_spp_m 416x416 60.1 115.1 158.9
110 yolov4_leaky_spp_m_pruned_0_36 416x416 38.2 127.8 182.8
111 ultrafast_pt 288x800 8.4 278.6 581
112 HardNet_MSeg_pt 352x352 22.78 172.7 211.4
113 drunet_pt 528x608 2.59 163.9 268.5
114 person-orientation_pruned_558m_pt 224x112 0.558 3771.3 6992.2
115 ofa_resnet50_0_9B_pt 160x160 0.9 1460.5 2329.7
116 SESR_S_pt 360x640 7.48 287.6 471.8
117 c2d2_lite 512x512 6.86 14.6 17.7
118 ofa_depthwise_res50_pt 176x176 1.25 279.9 456.3
119 FairMot_pt 640x480 36 158 298.2
120 mobilenet_v3_small_1_0_tf2 224x224 0.132 1271.7 2595.7
121 clocs 12000x100x4 41 8.4 15.7
122 tsd_yolox_pt 640x640 73 106.3 147.4
123 fadnet_pruned 576x960 154 8.1 13.9
124 ssr_pt 256x256 39.72 61.6 64.4
125 fadnet 576x960 441 7 11.5
126 psmnet 576x960 696 0.3 0.7
127 solo_pt 640x640 107 2.0 3.5