Skip to content

Commit e618946

Browse files
suisiyuansuisiyuan
committed
fix vendor_zoo nvidia cards.
1 parent 0dbf19b commit e618946

File tree

7 files changed

+93
-40
lines changed

7 files changed

+93
-40
lines changed

‎vendor_zoo/NVIDIA/A10-PCIe.json

+11-4
Original file line numberDiff line numberDiff line change
@@ -38,20 +38,23 @@
3838
},
3939
"标量参数": {
4040
"标量精度": null,
41-
"INT8标量算力(TFLOPS)": null,
41+
"INT8标量算力(TOPS)": null,
4242
"FP16标量算力(TFLOPS)": null,
4343
"FP32标量算力(TFLOPS)": null
4444
},
4545
"向量参数": {
4646
"向量精度": " FP32",
47-
"INT8量算力(TFLOPS)": null,
47+
"INT8量算力(TOPS)": null,
4848
"FP16向量算力(TFLOPS)": null,
4949
"FP32向量算力(TFLOPS)": 31.2
5050
},
5151
"张量参数": {
5252
"张量精度": "TF32, FP16, BF16, INT8, INT4",
53-
"INT8张量算力(TFLOPS)": 250.0,
53+
"INT4张量算力(TOPS)": 500.0,
54+
"INT8张量算力(TOPS)": 250.0,
55+
"BF16张量算力(TFLOPS)": 125.0,
5456
"FP16张量算力(TFLOPS)": 125.0,
57+
"TF32张量算力(TFLOPS)": 62.5,
5558
"FP32张量算力(TFLOPS)": null
5659
}
5760
},
@@ -61,5 +64,9 @@
6164
"RDMA协议": null,
6265
"下行带宽(GB/s)": null,
6366
"上行带宽(GB/s)": null
64-
}
67+
},
68+
"数据源": [
69+
"https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/a10/pdf/datasheet-new/nvidia-a10-datasheet.pdf",
70+
"https://www.techpowerup.com/gpu-specs/a10-pcie.c3793"
71+
]
6572
}

‎vendor_zoo/NVIDIA/A100-PCIe.json

+14-6
Original file line numberDiff line numberDiff line change
@@ -38,21 +38,25 @@
3838
},
3939
"标量参数": {
4040
"标量精度": null,
41-
"INT8标量算力(TFLOPS)": null,
41+
"INT8标量算力(TOPS)": null,
4242
"FP16标量算力(TFLOPS)": null,
4343
"FP32标量算力(TFLOPS)": null
4444
},
4545
"向量参数": {
4646
"向量精度": "FP64, FP32, TF32",
47-
"INT8量算力(TFLOPS)": null,
47+
"INT8量算力(TOPS)": null,
4848
"FP16向量算力(TFLOPS)": null,
49-
"FP32向量算力(TFLOPS)": 19.5
49+
"FP32向量算力(TFLOPS)": 19.5,
50+
"FP64向量算力(TFLOPS)": 9.7
5051
},
5152
"张量参数": {
5253
"张量精度": "FP64, FP32, TF32, FP16, BF16, INT8",
53-
"INT8张量算力(TFLOPS)": 624.0,
54+
"INT8张量算力(TOPS)": 624.0,
55+
"BF16张量算力(TFLOPS)": 312.0,
5456
"FP16张量算力(TFLOPS)": 312.0,
55-
"FP32张量算力(TFLOPS)": 156.0
57+
"TF32张量算力(TFLOPS)": 156.0,
58+
"FP32张量算力(TFLOPS)": null,
59+
"FP64张量算力(TFLOPS)": 19.5
5660
}
5761
},
5862
"卡间通信参数": {
@@ -61,5 +65,9 @@
6165
"RDMA协议": null,
6266
"下行带宽(GB/s)": null,
6367
"上行带宽(GB/s)": null
64-
}
68+
},
69+
"数据源": [
70+
"https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/a100/pdf/nvidia-a100-datasheet-nvidia-us-2188504-web.pdf",
71+
"https://www.techpowerup.com/gpu-specs/a100-pcie-80-gb.c3821"
72+
]
6573
}

‎vendor_zoo/NVIDIA/A100-SXM4.json

+15-7
Original file line numberDiff line numberDiff line change
@@ -38,21 +38,25 @@
3838
},
3939
"标量参数": {
4040
"标量精度": null,
41-
"INT8标量算力(TFLOPS)": null,
41+
"INT8标量算力(TOPS)": null,
4242
"FP16标量算力(TFLOPS)": null,
4343
"FP32标量算力(TFLOPS)": null
4444
},
4545
"向量参数": {
4646
"向量精度": "FP64, FP32, TF32",
47-
"INT8量算力(TFLOPS)": null,
47+
"INT8量算力(TOPS)": null,
4848
"FP16向量算力(TFLOPS)": null,
49-
"FP32向量算力(TFLOPS)": 19.5
49+
"FP32向量算力(TFLOPS)": 19.5,
50+
"FP64向量算力(TFLOPS)": 9.7
5051
},
5152
"张量参数": {
52-
"张量精度": "FP64, FP16, BF16, INT8",
53-
"INT8张量算力(TFLOPS)": 624.0,
53+
"张量精度": "FP64, FP32, TF32, FP16, BF16, INT8",
54+
"INT8张量算力(TOPS)": 624.0,
55+
"BF16张量算力(TFLOPS)": 312.0,
5456
"FP16张量算力(TFLOPS)": 312.0,
55-
"FP32张量算力(TFLOPS)": 156.0
57+
"TF32张量算力(TFLOPS)": 156.0,
58+
"FP32张量算力(TFLOPS)": null,
59+
"FP64张量算力(TFLOPS)": 19.5
5660
}
5761
},
5862
"卡间通信参数": {
@@ -61,5 +65,9 @@
6165
"RDMA协议": null,
6266
"下行带宽(GB/s)": null,
6367
"上行带宽(GB/s)": null
64-
}
68+
},
69+
"数据源": [
70+
"https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/a100/pdf/nvidia-a100-datasheet-nvidia-us-2188504-web.pdf",
71+
"https://www.techpowerup.com/gpu-specs/a100-sxm4-80-gb.c3746"
72+
]
6573
}

‎vendor_zoo/NVIDIA/A30-PCIe.json

+16-7
Original file line numberDiff line numberDiff line change
@@ -16,7 +16,7 @@
1616
"内存": {
1717
"内存类型": "HBM2e",
1818
"内存容量(GB)": 24,
19-
"内存带宽(GB/s)": 1223
19+
"内存带宽(GB/s)": 933.1
2020
},
2121
"一级缓存": {
2222
"缓存类型": "Cache(x56)",
@@ -38,21 +38,26 @@
3838
},
3939
"标量参数": {
4040
"标量精度": null,
41-
"INT8标量算力(TFLOPS)": null,
41+
"INT8标量算力(TOPS)": null,
4242
"FP16标量算力(TFLOPS)": null,
4343
"FP32标量算力(TFLOPS)": null
4444
},
4545
"向量参数": {
4646
"向量精度": "FP64, FP32",
47-
"INT8量算力(TFLOPS)": null,
47+
"INT8量算力(TOPS)": null,
4848
"FP16向量算力(TFLOPS)": null,
49-
"FP32向量算力(TFLOPS)": 10.3
49+
"FP32向量算力(TFLOPS)": 10.3,
50+
"FP64向量算力(TFLOPS)": 5.2
5051
},
5152
"张量参数": {
5253
"张量精度": "TF32, FP16, BF16, INT8, INT4",
53-
"INT8张量算力(TFLOPS)": 330.0,
54+
"INT4张量算力(TOPS)": 661.0,
55+
"INT8张量算力(TOPS)": 330.0,
56+
"BF16张量算力(TFLOPS)": 165.0,
5457
"FP16张量算力(TFLOPS)": 165.0,
55-
"FP32张量算力(TFLOPS)": 82.0
58+
"TF32张量算力(TFLOPS)": 82.0,
59+
"FP32张量算力(TFLOPS)": null,
60+
"FP64张量算力(TFLOPS)": 10.3
5661
}
5762
},
5863
"卡间通信参数": {
@@ -61,5 +66,9 @@
6166
"RDMA协议": null,
6267
"下行带宽(GB/s)": null,
6368
"上行带宽(GB/s)": null
64-
}
69+
},
70+
"数据源": [
71+
"https://www.nvidia.com/content/dam/en-zz/Solutions/data-center/products/a30-gpu/pdf/a30-datasheet.pdf",
72+
"https://www.techpowerup.com/gpu-specs/a30-pcie.c3792"
73+
]
6574
}

‎vendor_zoo/NVIDIA/H100-PCIe.json

+14-6
Original file line numberDiff line numberDiff line change
@@ -38,21 +38,25 @@
3838
},
3939
"标量参数": {
4040
"标量精度": null,
41-
"INT8标量算力(TFLOPS)": null,
41+
"INT8标量算力(TOPS)": null,
4242
"FP16标量算力(TFLOPS)": null,
4343
"FP32标量算力(TFLOPS)": null
4444
},
4545
"向量参数": {
4646
"向量精度": "FP64, FP32",
47-
"INT8量算力(TFLOPS)": null,
47+
"INT8量算力(TOPS)": null,
4848
"FP16向量算力(TFLOPS)": null,
49-
"FP32向量算力(TFLOPS)": 51.0
49+
"FP32向量算力(TFLOPS)": 51.0,
50+
"FP64向量算力(TFLOPS)": 26.0
5051
},
5152
"张量参数": {
5253
"张量精度": "FP64, TF32, FP16, BF16, FP8, INT8",
53-
"INT8张量算力(TFLOPS)": 3026.0,
54+
"INT8张量算力(TOPS)": 3026.0,
55+
"FP8张量算力(TFLOPS)": 3026.0,
56+
"BF16张量算力(TFLOPS)": 1513.0,
5457
"FP16张量算力(TFLOPS)": 1513.0,
55-
"FP32张量算力(TFLOPS)": 756.0
58+
"TF32张量算力(TFLOPS)": 756.0,
59+
"FP64张量算力(TFLOPS)": 51.0
5660
}
5761
},
5862
"卡间通信参数": {
@@ -61,5 +65,9 @@
6165
"RDMA协议": null,
6266
"下行带宽(GB/s)": null,
6367
"上行带宽(GB/s)": null
64-
}
68+
},
69+
"数据源": [
70+
"https://resources.nvidia.com/en-us-tensor-core/nvidia-tensor-core-gpu-datasheet?ncid=no-ncid",
71+
"https://www.techpowerup.com/gpu-specs/h100-pcie-80-gb.c3899"
72+
]
6573
}

‎vendor_zoo/NVIDIA/H100-SXM5.json

+14-6
Original file line numberDiff line numberDiff line change
@@ -38,21 +38,25 @@
3838
},
3939
"标量参数": {
4040
"标量精度": null,
41-
"INT8标量算力(TFLOPS)": null,
41+
"INT8标量算力(TOPS)": null,
4242
"FP16标量算力(TFLOPS)": null,
4343
"FP32标量算力(TFLOPS)": null
4444
},
4545
"向量参数": {
4646
"向量精度": "FP64, FP32",
47-
"INT8量算力(TFLOPS)": null,
47+
"INT8量算力(TOPS)": null,
4848
"FP16向量算力(TFLOPS)": null,
49-
"FP32向量算力(TFLOPS)": 67.0
49+
"FP32向量算力(TFLOPS)": 67.0,
50+
"FP64向量算力(TFLOPS)": 34.0
5051
},
5152
"张量参数": {
5253
"张量精度": "FP64, TF32, FP16, BF16, FP8, INT8",
53-
"INT8张量算力(TFLOPS)": 3958.0,
54+
"INT8张量算力(TOPS)": 3958.0,
55+
"FP8张量算力(TFLOPS)": 3958.0,
56+
"BF16张量算力(TFLOPS)": 1979.0,
5457
"FP16张量算力(TFLOPS)": 1979.0,
55-
"FP32张量算力(TFLOPS)": 989.0
58+
"TF32张量算力(TFLOPS)": 989.0,
59+
"FP64张量算力(TFLOPS)": 67.0
5660
}
5761
},
5862
"卡间通信参数": {
@@ -61,5 +65,9 @@
6165
"RDMA协议": null,
6266
"下行带宽(GB/s)": null,
6367
"上行带宽(GB/s)": null
64-
}
68+
},
69+
"数据源": [
70+
"https://resources.nvidia.com/en-us-tensor-core/nvidia-tensor-core-gpu-datasheet?ncid=no-ncid",
71+
"https://www.techpowerup.com/gpu-specs/h100-sxm5-80-gb.c3900"
72+
]
6573
}

‎vendor_zoo/NVIDIA/T4.json

+9-4
Original file line numberDiff line numberDiff line change
@@ -38,19 +38,20 @@
3838
},
3939
"标量参数": {
4040
"标量精度": null,
41-
"INT8标量算力(TFLOPS)": null,
41+
"INT8标量算力(TOPS)": null,
4242
"FP16标量算力(TFLOPS)": null,
4343
"FP32标量算力(TFLOPS)": null
4444
},
4545
"向量参数": {
4646
"向量精度": "FP32",
47-
"INT8量算力(TFLOPS)": null,
47+
"INT8量算力(TOPS)": null,
4848
"FP16向量算力(TFLOPS)": null,
4949
"FP32向量算力(TFLOPS)": 8.1
5050
},
5151
"张量参数": {
5252
"张量精度": "FP16, INT8, INT4",
53-
"INT8张量算力(TFLOPS)": 130.0,
53+
"INT4张量算力(TOPS)": 260.0,
54+
"INT8张量算力(TOPS)": 130.0,
5455
"FP16张量算力(TFLOPS)": 65.0,
5556
"FP32张量算力(TFLOPS)": null
5657
}
@@ -61,5 +62,9 @@
6162
"RDMA协议": null,
6263
"下行带宽(GB/s)": null,
6364
"上行带宽(GB/s)": null
64-
}
65+
},
66+
"数据源": [
67+
"https://www.nvidia.com/en-us/data-center/tesla-t4/",
68+
"https://www.techpowerup.com/gpu-specs/tesla-t4.c3316"
69+
]
6570
}

0 commit comments

Comments
 (0)