VCK190 Evaluation Board

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 B8192F DPU core with batch=3 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 1333 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	55.6
2	centerpoint_0_pt centerpoint_1_pt	2560x40x4	54	111.2	225.3
3	densebox_320_320	320x320	0.49	1017.2	2166.2
4	densebox_640_360	360x640	1.1	499.5	1027
5	efficientnet-b0_tf2	224x224	0.36	543.1	715.6
6	efficientNet-edgetpu-L_tf	300x300	19.36	232.5	279.5
7	efficientNet-edgetpu-M_tf	240x240	7.34	519	700.5
8	efficientNet-edgetpu-S_tf	224x224	4.72	697.9	1015
9	ENet_cityscapes_pt	512x1024	8.6	21.1	48
10	face_landmark	96x72	0.14	5261.6	10043.2
11	face-quality	80x60	0.06	7945.7	19532.7
12	face-quality_pt	80x60	0.06	7803.9	18385.5
13	facerec_resnet20	112x96	3.5	1963.2	2551.9
14	facerec-resnet20_mixed_pt	112x96	3.5	1964.2	2551.9
15	facerec_resnet64	112x96	11	1086.7	1244
16	facereid-large_pt	96x96	0.5	5044.6	12373.3
17	facereid-small_pt	80x80	0.09	7183.9	17007.1
18	fpn	256x512	8.9	82.8	189.9
19	FPN_Res18_Medical_segmentation	320x320	45.3	66.7	188.9
20	FPN-resnet18_covid19-seg_pt	352x352	22.7	366.3	577
21	FPN-resnet18_Endov	240x320	13.75	102.6	231.9
22	hourglass-pe_mpii	256x256	10.2	73.1	131.1
23	inception_resnet_v2_tf	299x299	26.4	251.4	306.8
24	inception_v1	224x224	3.2	981.3	1749.4
25	inception_v1_tf	224x224	3	986.8	1770.3
26	inception_v2	224x224	4	780.7	1243.8
27	inception_v2_tf	224x224	3.88	558.2	746.6
28	inception_v3	299x299	11.4	424.7	611.5
29	inception_v3_pt	299x299	5.7	424.3	610.9
30	inception_v3_tf	299x299	11.5	423.8	610.7
31	inception_v3_tf2	299x299	11.5	440.7	647.1
32	inception_v4	299x299	24.5	243.6	295.4
33	inception_v4_2016_09_09_tf	299x299	24.6	243.5	295.2
34	medical_seg_cell_tf2	128x128	5.3	1158.5	2032.4
35	MLPerf_resnet50_v1.5_tf	224x224	8.19	841.1	1345.3
36	mlperf_ssd_resnet34_tf	1200x1200	433	10.6	23.1
37	mobilenet_1_0_224_tf2	224x224	1.1	1383.7	3575.9
38	mobilenet_edge_0_75_tf	224x224	0.62	1302.8	3111.5
39	mobilenet_edge_1_0_tf	224x224	0.99	1228.7	2719.8
40	mobilenet_v1_0_25_128_tf	128x128	0.027	3627	7948.5
41	mobilenet_v1_0_5_160_tf	160x160	0.15	2580.2	6211.5
42	mobilenet_v1_1_0_224_tf	224x224	1.1	1381.2	3578.8
43	mobilenet_v2	224x224	0.6	1257.8	2851.2
44	mobilenet_v2_1_0_224_tf	224x224	0.6	1255.1	2828
45	mobilenet_v2_1_4_224_tf	224x224	1.2	1042.7	1975.3
46	mobilenet_v2_cityscapes_tf	1024x2048	132.74	4.5	11.6
47	MT-resnet18_mixed_pt	512x320	13.65	120.4	264.3
48	multi_task	288x512	14.8	145.9	302.3
49	multi_task_v3_pt	320x512	25.44	71.5	184.9
50	openpose_pruned_0_3	368x368	49.9	22.1	42.9
51	personreid-res18_pt	176x80	1.1	2748.5	4888.3
52	personreid-res50_pt	256x128	5.4	1127	1823.7
53	plate_detection	320x320	0.49	1190.7	2489.3
54	plate_num	96x288	1.75	448.3	919.1
55	pmg_pt	224x224	2.28	1176.8	2059.2
56	pointpainting- pointpainting_nuscenes_40000_64_0_pt pointpainting_nuscenes_40000_64_1_pt	40000x64x16	112	3.5	6.5
57	pointpillars_kitti_12000_0_pt pointpillars_kitti_12000_1_pt	12000x100x4	10.8	34.8	53.9
58	pointpillars_nuscenes_40000_64_0_pt pointpillars_nuscenes_40000_64_1_pt	40000x64x5	108	7.5	17.2
59	rcan_pruned_tf	360x640	86.95	47.9	61.7
60	refinedet_baseline	480x360	123	119.6	149.9
61	RefineDet-Medical_EDD_tf	320x320	9.8	416.7	919.8
62	refinedet_pruned_0_8	360x480	25	267.4	465.5
63	refinedet_pruned_0_92	360x480	10.1	339.8	702.4
64	refinedet_pruned_0_96	360x480	5.1	405.7	855.7
65	refinedet_VOC_tf	320x320	81.9	81	186.6
66	reid	80x160	0.95	2672.4	5056.3
67	resnet18	224x224	3.7	1281.1	2952.3
68	resnet50	224x224	7.7	850.8	1381.6
69	resnet50_pt	224x224	4.1	839.1	1344.3
70	resnet50_tf2	224x224	7.7	850.3	1379.4
71	resnet_v1_101_tf	224x224	14.4	607.6	834.4
72	resnet_v1_152_tf	224x224	21.8	457.3	574.8
73	resnet_v1_50_tf	224x224	7	905.9	1520.7
74	resnet_v2_101_tf	299x299	26.78	276.2	344.8
75	resnet_v2_152_tf	299x299	40.47	210.5	247.9
76	resnet_v2_50_tf	299x299	13.1	401.2	563.6
77	retinaface	360x640	1.11	335.1	752.4
78	SA_gate_pt	360x360	178	8.2	14
79	salsanext_pt	64x2048	20.4	11.0	22.2
80	salsanext_v2_pt	64x2048	32	9.3	21.5
81	SemanticFPN_cityscapes_pt	256x512	10	88.6	191
82	SemanticFPN_Mobilenetv2_pt	512x1024	5.4	21.8	48.3
83	semantic_seg_citys_tf2	512x1024	54	17.6	45.2
84	sp_net	128x224	0.55	1813.1	4026.6
85	squeezenet	227x227	0.76	2406.1	4165.6
86	squeezenet_pt	224x224	0.82	807.5	1965.6
87	ssd_adas_pruned_0_95	360x480	6.3	382.2	820.6
88	ssd_inception_v2_coco_tf	300x300	9.6	191.5	386.2
89	ssdlite_mobilenet_v2_coco_tf	300x300	1.5	334.1	504.2
90	ssd_mobilenet_v1_coco_tf	300x300	2.5	359.8	517.9
91	ssd_mobilenet_v2	360x480	6.6	74.9	167
92	ssd_mobilenet_v2_coco_tf	300x300	3.8	312.9	510.5
93	ssd_pedestrian_pruned_0_97	360x360	5.9	317.7	709
94	ssd_resnet_50_fpn_coco_tf	640x640	178.4	10	11.8
95	ssd_traffic_pruned_0_9	360x480	11.6	281.5	668.8
96	tiny_yolov3_vmss	416x416	5.46	521.5	1279.6
97	unet_chaos-CT_pt	512x512	23.3	83.4	220
98	vgg_16_tf	224x224	31	318.8	371.9
99	vgg_19_tf	224x224	39.3	286	328.3
100	vpgnet_pruned_0_99	480x640	2.5	269.7	678.9
101	yolov2_voc	448x448	34	278.8	460.2
102	yolov2_voc_pruned_0_66	448x448	11.6	385.9	847
103	yolov2_voc_pruned_0_71	448x448	9.9	406.6	943.7
104	yolov2_voc_pruned_0_77	448x448	7.8	424.8	1048.9
105	yolov3_adas_pruned_0_9	256x512	5.5	406.4	967.8
106	yolov3_bdd	288x512	53.7	150.4	200.1
107	yolov3_voc	416x416	65.4	156.1	200.9
108	yolov3_voc_tf	416x416	65.6	156.8	201.5
109	yolov4_leaky_spp_m	416x416	60.1	118.5	167.7
110	yolov4_leaky_spp_m_pruned_0_36	416x416	38.2	129.6	190.2