VCK190 Evaluation Board - 3.0 English

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

The 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 C32B6 DPU core is implemented using AI Engines and delivers 61.4 TOPS INT8 peak performance for deep learning inference acceleration.

Refer to the following table for the throughput performance (in frames/sec or fps) for various neural network samples on VCK190 with AI Engines clocked at 1250 MHz and PL clocked at 333 MHz.

Table 1. VCK190 Performance with Batch 6
No Neural Network Input Size GOPS Performance (fps) (Single thread) Performance (fps) (Multiple thread)
1 3D-Unet_pt 128x128x128x1 1065.44 0.04 0.06
2 bcc_pt 800x1000 268.9 36.9 72.1
3 bevdet 256x704 407.6 6.5 12.9
4 c2d2_lite 512x512 6.86 20.1 26.1
5 centerpoint 2560x40x4 54 128.3 235.5
6 cflownet_pt 128x128 5.21 90.8 97.1
7 chen_color_resnet18_pt 224x224 3.627 2173.5 5461.6
8 clocs 12000x100x4 41 8.1 14.5
9 drunet_pt 528x608 2.59 259.5 461.9
10 efficientdet_d2_tf 768x768 11.06 18 39.8
11 efficientnet_lite_tf2 224x224 0.77 1671.2 4456.5
12 efficientnet-b0_tf2 224x224 0.36 1090 1819.3
13 efficientNet-edgetpu-L_tf 300x300 19.36 389.8 500.8
14 efficientNet-edgetpu-M_tf 240x240 7.34 879.7 1379.4
15 efficientNet-edgetpu-S_tf 224x224 4.72 1194.2 2166.5
16 ENet_cityscapes_pt 512x1024 8.6 25.7 54.3
17 face_mask_detection_pt 512x512 0.593 453.6 935.5
18 face-quality_pt 80x60 0.06 13317.8 28402.7
19 facerec-resnet20_mixed_pt 112x96 3.5 3693.9 5418.4
20 facereid-large_pt 96x96 0.5 8143.1 19327.2
21 facereid-small_pt 80x80 0.09 11935.6 25837.5
22 fadnet 576x960 441 8 13.5
23 fadnet_pruned 576x960 154 9 16
24 fadnet_v2_pt 576x960 412 9.0 16.9
25 fadnet_v2_pruned_pt 576x960 201 9.6 18.7
26 FairMot_pt 640x480 36 194.3 366.1
27 FPN-resnet18_covid19-seg_pt 352x352 22.7 485.6 797.9
28 HardNet_MSeg_pt 352x352 22.78 218.6 273.5
29 hfnet_tf 960x960 20.09 9.4 22.3
30 HRNet_pt (C32B3) 1024x2048 1511.9 5.2 5.7
31 inception_resnet_v2_tf 299x299 26.4 382.2 491.3
32 inception_v1_tf 224x224 3 1290.1 2438.2
33 inception_v2_tf 224x224 3.88 822.4 1184.4
34 inception_v3_pt 299x299 5.7 583.2 876.8
35 inception_v3_tf 299x299 11.5 586.1 884.4
36 inception_v3_tf2 299x299 11.5 646.2 1033.3
37 inception_v4_2016_09_09_tf 299x299 24.6 337.8 418.1
38 medical_seg_cell_tf2 128x128 5.3 1512.6 2972.8
39 MLPerf_resnet50_v1.5_tf 224x224 8.19 1344.8 2671.5
40 mlperf_ssd_resnet34_tf 1200x1200 433 11.9 20.8
41 mobilenet_1_0_224_tf2 224x224 1.1 1991.6 4950.2
42 mobilenet_edge_0_75_tf 224x224 0.62 1895.5 4876.6
43 mobilenet_edge_1_0_tf 224x224 0.99 1806.7 4806
44 mobilenet_v1_0_25_128_tf 128x128 0.027 4992.5 10220.3
45 mobilenet_v1_0_5_160_tf 160x160 0.15 3611.4 7915.5
46 mobilenet_v1_1_0_224_tf 224x224 1.1 2001 4931.7
47 mobilenet_v2_1_0_224_tf 224x224 0.6 1887.7 4886.1
48 mobilenet_v2_1_4_224_tf 224x224 1.2 1579 3646.8
49 mobilenet_v2_cityscapes_tf 1024x2048 132.74 5.3 11.9
50 mobilenet_v3_small_1_0_tf2 224x224 0.132 2021.3 4924.5
51 monodepth2_pt 192x640 257.21 427.2 725.5
52 movenet_ntd_pt 192x192 0.5 243.8 431.3
53 MT-resnet18_mixed_pt 512x320 13.65 143.4 258.8
54 multi_task_v3_pt 320x512 25.44 76.3 174.2
55 ocr_pt 960x960 875.7 8.6 18.4
56 ofa_depthwise_res50_pt 176x176 1.25 309.8 450.2
57 ofa_rcan_latency_pt 360x640 45.7 59.5 81.1
58 ofa_resnet50_0_9B_pt 160x160 1.8 2029 3849.3
59 ofa_yolo_pruned_0_30_pt 640x640 34.71 145.2 253.9
60 ofa_yolo_pruned_0_50_pt 640x640 24.62 166.7 294.1
61 ofa_yolo_pt 640x640 48.88 127.7 219.3
62 person-orientation_pruned_558m_pt 224x112 0.558 6080.5 16496.2
63 personreid-res18_pt 176x80 1.1 4581.2 10087.9
64 personreid-res50_pt 256x128 5.3 1062.9 2220.5
65 pmg_pt 224x224 2.28 1748.2 3553
66 pointpainting 40000x64x16 112 3.8 6.7
67 pointpillars_kitti_12000_pt 12000x100x4 10.8 25 34.5
68 pointpillars_nuscenes 40000x64x5 108 7.6 16
69 psmnet 576x960 696 0.4 0.7
70 rcan_pruned_tf 360x640 86.95 44.6 56.8
71 refinedet_VOC_tf 320x320 81.9 102.1 225.4
72 RefineDet-Medical_EDD_tf 320x320 9.8 550.1 1288.3
73 resnet_v1_101_tf 224x224 14.4 1063 1728.6
74 resnet_v1_152_tf 224x224 21.8 830.2 1200.8
75 resnet_v1_50_tf 224x224 7 1425.6 3011.7
76 resnet_v2_101_tf 299x299 26.78 408.9 531.8
77 resnet_v2_152_tf 299x299 40.47 328.2 402.1
78 resnet_v2_50_tf 299x299 13.1 542.3 783.5
79 resnet50_pt 224x224 4.1 1347.3 2689.3
80 resnet50_tf2 224x224 7.7 1437.9 3015.4
81 SA_gate_base_pt 360x360 178 7.6 9.6
82 salsanext_pt 64x2048 20.4 30.1 63.2
83 salsanext_v2_pt 64x2048 32 21.8 56.3
84 semantic_seg_citys_tf2 512x1024 54 20.5 46.6
85 SemanticFPN_cityscapes_pt 256x512 10 110.3 224.2
86 SemanticFPN_Mobilenetv2_pt 512x1024 5.4 27.2 55.6
87 SESR_S_pt 360x640 7.48 351.4 685.4
88 solo_pt 640x640 107 4.4 7.4
89 squeezenet_pt 224x224 0.82 3124.3 5862.1
90 ssd_inception_v2_coco_tf 300x300 9.6 273.3 416.2
91 ssd_mobilenet_v1_coco_tf 300x300 2.5 430.6 523.4
92 ssd_mobilenet_v2_coco_tf 300x300 3.8 352 518.5
93 ssd_resnet_50_fpn_coco_tf 640x640 178.4 11.1 12.2
94 ssdlite_mobilenet_v2_coco_tf 300x300 1.5 410.7 518.2
95 ssr_pt 256x256 39.72 74.1 77.6
96 superpoint_tf 480x640 52.4 49.9 105.5
97 textmountain_pt 960x960 575.2 20.8 29.3
98 tsd_yolox_pt 640x640 73 134 191.4
99 ultrafast_pt 288x800 8.4 389.3 959.4
100 unet_chaos-CT_pt 512x512 23.3 93.5 237.1
101 vehicle_make_resnet18_pt 224x224 3.627 2172.1 5561.7
102 vehicle_type_resnet18_pt 224x224 3.627 2181.1 5578.8
103 vgg_16_tf 224x224 31 505.1 620.5
104 vgg_19_tf 224x224 39.3 459.9 553.9
105 xilinxSR_pt 360x640x3 182.44 12.8 14.5
106 yolov3_coco_416_tf2 416x416 65.9 190.2 288
107 yolov3_voc_tf 416x416 65.6 215.1 287.5
108 yolov4_csp_pt 640x640 121 79.6 119
109 yolov4_leaky_416_tf 416x416 60.3 139.9 211.2
110 yolov4_leaky_512_tf 512x512 91.2 104.2 152.8
111 yolov5_large_pt 640x640 109.6 88.2 150.4
112 yolov5_nano_pt 640x640 4.6 224.3 395.2
113 yolov5s6_pt 640x640 17 44.5 70.7
114 yolov6m_pt 640x640 82.2 24.8 37.4
115 yolox_nano_pt 416x416x3 1 620.1 1266.4