Debian Codegraph @

results/NVIDIA_CUDA/NVIDIA_A100-SXM4-40GB.log

raw
1 Platform: NVIDIA CUDA
2 Device: NVIDIA A100-SXM4-40GB
3 Driver version : 515.48.07 (Linux x64)
4 Compute units : 108
5 Clock frequency : 1410 MHz
6
7 Global memory bandwidth (GBPS)
8 float : 1292.96
9 float2 : 1377.07
10 float4 : 1419.22
11 float8 : 1443.80
12 float16 : 1464.44
13
14 Single-precision compute (GFLOPS)
15 float : 19352.89
16 float2 : 19386.64
17 float4 : 19367.97
18 float8 : 19285.78
19 float16 : 19110.61
20
21 No half precision support! Skipped
22
23 Double-precision compute (GFLOPS)
24 double : 9723.09
25 double2 : 9707.90
26 double4 : 9686.22
27 double8 : 9636.75
28 double16 : 9547.99
29
30 Integer compute (GIOPS)
31 int : 19278.83
32 int2 : 19335.19
33 int4 : 19273.49
34 int8 : 19357.90
35 int16 : 19346.59
36
37 Integer compute Fast 24bit (GIOPS)
38 int : 19289.10
39 int2 : 19293.80
40 int4 : 19278.51
41 int8 : 19233.96
42 int16 : 19053.93
43
44 Transfer bandwidth (GBPS)
45 enqueueWriteBuffer : 14.54
46 enqueueReadBuffer : 13.04
47 enqueueWriteBuffer non-blocking : 7.79
48 enqueueReadBuffer non-blocking : 7.21
49 enqueueMapBuffer(for read) : 20.03
50 memcpy from mapped ptr : 20.65
51 enqueueUnmap(after write) : 26.77
52 memcpy to mapped ptr : 20.86
53
54 Kernel launch latency : 5.76 us
55
56 Device: NVIDIA A100-SXM4-40GB
57 Driver version : 515.48.07 (Linux x64)
58 Compute units : 108
59 Clock frequency : 1410 MHz
60
61 Global memory bandwidth (GBPS)
62 float : 1293.90
63 float2 : 1377.21
64 float4 : 1419.86
65 float8 : 1443.43
66 float16 : 1464.50
67
68 Single-precision compute (GFLOPS)
69 float : 19349.87
70 float2 : 19381.94
71 float4 : 19362.20
72 float8 : 19280.28
73 float16 : 19105.41
74
75 No half precision support! Skipped
76
77 Double-precision compute (GFLOPS)
78 double : 9716.58
79 double2 : 9701.24
80 double4 : 9680.46
81 double8 : 9631.46
82 double16 : 9543.48
83
84 Integer compute (GIOPS)
85 int : 19275.95
86 int2 : 19324.45
87 int4 : 19268.48
88 int8 : 19351.87
89 int16 : 19343.15
90
91 Integer compute Fast 24bit (GIOPS)
92 int : 19283.22
93 int2 : 19287.28
94 int4 : 19272.75
95 int8 : 19230.13
96 int16 : 19047.88
97
98 Transfer bandwidth (GBPS)
99 enqueueWriteBuffer : 14.50
100 enqueueReadBuffer : 13.05
101 enqueueWriteBuffer non-blocking : 7.71
102 enqueueReadBuffer non-blocking : 7.27
103 enqueueMapBuffer(for read) : 19.83
104 memcpy from mapped ptr : 19.54
105 enqueueUnmap(after write) : 26.77
106 memcpy to mapped ptr : 20.55
107
108 Kernel launch latency : 5.65 us
109
110 Device: NVIDIA A100-SXM4-40GB
111 Driver version : 515.48.07 (Linux x64)
112 Compute units : 108
113 Clock frequency : 1410 MHz
114
115 Global memory bandwidth (GBPS)
116 float : 1304.11
117 float2 : 1376.87
118 float4 : 1419.82
119 float8 : 1444.07
120 float16 : 1465.06
121
122 Single-precision compute (GFLOPS)
123 float : 19350.41
124 float2 : 19382.11
125 float4 : 19363.12
126 float8 : 19281.61
127 float16 : 19108.25
128
129 No half precision support! Skipped
130
131 Double-precision compute (GFLOPS)
132 double : 9719.24
133 double2 : 9704.38
134 double4 : 9682.93
135 double8 : 9633.92
136 double16 : 9544.74
137
138 Integer compute (GIOPS)
139 int : 19277.98
140 int2 : 19332.19
141 int4 : 19269.01
142 int8 : 19352.73
143 int16 : 19343.15
144
145 Integer compute Fast 24bit (GIOPS)
146 int : 19283.32
147 int2 : 19288.03
148 int4 : 19273.28
149 int8 : 19231.30
150 int16 : 19048.40
151
152 Transfer bandwidth (GBPS)
153 enqueueWriteBuffer : 14.37
154 enqueueReadBuffer : 13.13
155 enqueueWriteBuffer non-blocking : 7.50
156 enqueueReadBuffer non-blocking : 6.90
157 enqueueMapBuffer(for read) : 19.81
158 memcpy from mapped ptr : 20.73
159 enqueueUnmap(after write) : 26.77
160 memcpy to mapped ptr : 20.62
161
162 Kernel launch latency : 5.75 us
163
164 Device: NVIDIA A100-SXM4-40GB
165 Driver version : 515.48.07 (Linux x64)
166 Compute units : 108
167 Clock frequency : 1410 MHz
168
169 Global memory bandwidth (GBPS)
170 float : 1303.89
171 float2 : 1376.82
172 float4 : 1419.15
173 float8 : 1444.89
174 float16 : 1465.04
175
176 Single-precision compute (GFLOPS)
177 float : 19339.44
178 float2 : 19388.10
179 float4 : 19371.42
180 float8 : 19289.58
181 float16 : 19115.54
182
183 No half precision support! Skipped
184
185 Double-precision compute (GFLOPS)
186 double : 9724.83
187 double2 : 9710.39
188 double4 : 9689.79
189 double8 : 9641.13
190 double16 : 9552.76
191
192 Integer compute (GIOPS)
193 int : 19285.03
194 int2 : 19313.19
195 int4 : 19286.42
196 int8 : 19361.56
197 int16 : 19347.78
198
199 Integer compute Fast 24bit (GIOPS)
200 int : 19292.73
201 int2 : 19297.12
202 int4 : 19282.58
203 int8 : 19238.22
204 int16 : 19056.33
205
206 Transfer bandwidth (GBPS)
207 enqueueWriteBuffer : 14.48
208 enqueueReadBuffer : 13.16
209 enqueueWriteBuffer non-blocking : 7.18
210 enqueueReadBuffer non-blocking : 6.98
211 enqueueMapBuffer(for read) : 19.99
212 memcpy from mapped ptr : 19.35
213 enqueueUnmap(after write) : 26.77
214 memcpy to mapped ptr : 20.63
215
216 Kernel launch latency : 5.70 us