显卡算力参数对比
文章目录
- 显卡算力参数对比
- A 显卡参数查询
- B 显卡性能对比：
综合看：T4最具性价比
A 显卡参数查询
查询网址：https://www.techpowerup.com/gpu-specs/ ，以下列出部分：
| Product Name | GPU Chip | Released | Bus | Memory | GPU clock | Memory clock | Shaders / TMUs / ROPs |
|---|---|---|---|---|---|---|---|
| GeForce RTX 5090 | GB102 | 2025 | PCIe 5.0 x16 | 32 GB, GDDR7, 512 bit | 2235 MHz | 2500 MHz | 21760 / 680 / 192 |
| L40 | AD102 | Oct 13th, 2022 | PCIe 4.0 x16 | 48 GB, GDDR6, 384 bit | 735 MHz | 2250 MHz | 18176 / 568 / 192 |
| L20 | AD102 | Nov 16th, 2023 | PCIe 4.0 x16 | 48 GB, GDDR6, 384 bit | 1440 MHz | 2250 MHz | 11776 / 368 / 128 |
| GeForce RTX 4090 | AD102 | Sep 20th, 2022 | PCIe 4.0 x16 | 24 GB, GDDR6X, 384 bit | 2235 MHz | 1313 MHz | 16384 / 512 / 176 |
| GeForce RTX 4090 D | AD102 | Dec 28th, 2023 | PCIe 4.0 x16 | 24 GB, GDDR6X, 384 bit | 2280 MHz | 1313 MHz | 14592 / 456 / 176 |
| GeForce RTX 4090 Ti | AD102 | Never Released | PCIe 4.0 x16 | 24 GB, GDDR6X, 384 bit | 2325 MHz | 1500 MHz | 18176 / 568 / 192 |
| GeForce RTX 4070 | AD104 | Apr 12th, 2023 | PCIe 4.0 x16 | 12 GB, GDDR6X, 192 bit | 1920 MHz | 1313 MHz | 5888 / 184 / 64 |
| GeForce RTX 4070 SUPER | AD104 | Jan 8th, 2024 | PCIe 4.0 x16 | 12 GB, GDDR6X, 192 bit | 1980 MHz | 1313 MHz | 7168 / 224 / 80 |
| GeForce RTX 4070 Ti | AD104 | Jan 3rd, 2023 | PCIe 4.0 x16 | 12 GB, GDDR6X, 192 bit | 2310 MHz | 1313 MHz | 7680 / 240 / 80 |
| GeForce RTX 4070 Ti SUPER | AD103 | Jan 8th, 2024 | PCIe 4.0 x16 | 16 GB, GDDR6X, 256 bit | 2340 MHz | 1313 MHz | 8448 / 264 / 96 |
| GeForce RTX 4060 | AD107 | May 18th, 2023 | PCIe 4.0 x8 | 8 GB, GDDR6, 128 bit | 1830 MHz | 2125 MHz | 3072 / 96 / 48 |
| GeForce RTX 4060 Ti 8 GB | AD106 | May 18th, 2023 | PCIe 4.0 x8 | 8 GB, GDDR6, 128 bit | 2310 MHz | 2250 MHz | 4352 / 136 / 48 |
| GeForce RTX 3090 | GA102 | Sep 1st, 2020 | PCIe 4.0 x16 | 24 GB, GDDR6X, 384 bit | 1395 MHz | 1219 MHz | 10496 / 328 / 112 |
| GeForce RTX 2080 Ti | TU102 | Sep 20th, 2018 | PCIe 3.0 x16 | 11 GB, GDDR6, 352 bit | 1350 MHz | 1750 MHz | 4352 / 272 / 88 |
| GeForce GTX 1080 Ti | GP102 | Mar 10th, 2017 | PCIe 3.0 x16 | 11 GB, GDDR5X, 352 bit | 1481 MHz | 1376 MHz | 3584 / 224 / 88 |
| GeForce GTX 1070 | GP104 | Jun 10th, 2016 | PCIe 3.0 x16 | 8 GB, GDDR5, 256 bit | 1506 MHz | 2002 MHz | 1920 / 120 / 64 |
| GeForce GTX 1060 6 GB | GP106 | Jul 19th, 2016 | PCIe 3.0 x16 | 6 GB, GDDR5, 192 bit | 1506 MHz | 2002 MHz | 1280 / 80 / 48 |
| GeForce GTX 1050 Ti | GP107 | Oct 25th, 2016 | PCIe 3.0 x16 | 4 GB, GDDR5, 128 bit | 1291 MHz | 1752 MHz | 768 / 48 / 32 |
| GeForce GTX 1660 Ti | TU116 | Feb 22nd, 2019 | PCIe 3.0 x16 | 6 GB, GDDR6, 192 bit | 1500 MHz | 1500 MHz | 1536 / 96 / 48 |
| — | |||||||
| Tesla T4 | TU104 | Sep 13th, 2018 | PCIe 3.0 x16 | 16 GB, GDDR6, 256 bit | 585 MHz | 1250 MHz | 2560 / 160 / 64 |
| Tesla V100 PCIe 32 GB | GV100 | Mar 27th, 2018 | PCIe 3.0 x16 | 32 GB, HBM2, 4096 bit | 1230 MHz | 876 MHz | 5120 / 320 / 128 |
| A100 PCIe 80 GB | GA100 | Jun 28th, 2021 | PCIe 4.0 x16 | 80 GB, HBM2e, 5120 bit | 1065 MHz | 1512 MHz | 6912 / 432 / 160 |
| A800 PCIe 80 GB | GA100 | Nov 8th, 2022 | PCIe 4.0 x16 | 80 GB, HBM2e, 5120 bit | 1065 MHz | 1512 MHz | 6912 / 432 / 160 |
| H800 PCIe 80 GB | GH100 | Mar 21st, 2023 | PCIe 5.0 x16 | 80 GB, HBM2e, 5120 bit | 1095 MHz | 1593 MHz | 14592 / 456 / 24 |
| H100 PCIe 96 GB | GH100 | Mar 21st, 2023 | PCIe 5.0 x16 | 96 GB, HBM3, 5120 bit | 1665 MHz | 1313 MHz | 16896 / 528 / 24 |
B 显卡性能对比：
| 标志 | 显卡 | Graphics Processor | Cores | Memory Size (GB) | Bus Width (bit) | FP16 (half) | FP32 (float) | FP64 (double) | 备注 |
|---|---|---|---|---|---|---|---|---|---|
| 1 | NVIDIA H100 PCIe 96 GB | GH100 | 16896 | 96 | 5120 | 248.3 TFLOPS (4:1) | 62.08 TFLOPS | 31.04 TFLOPS (1:2) | |
| NVIDIA H100 SXM5 96 GB | GH100 | 16896 | 96 | 5120 | 248.3 TFLOPS (4:1) | 62.08 TFLOPS | 31.04 TFLOPS (1:2) | ||
| 1 | NVIDIA H800 PCIe 80 GB | GH100 | 14592 | 80 | 5120 | 204.9 TFLOPS (4:1) | 51.22 TFLOPS | 25.61 TFLOPS (1:2) | |
| NVIDIA H800 SXM5 | GH100 | 16896 | 80 | 5120 | 237.2 TFLOPS (4:1) | 59.30 TFLOPS | 29.65 TFLOPS (1:2) | ||
| 1 | NVIDIA A100 PCIe 80 GB | GA100 | 6912 | 80 | 5120 | 77.97 TFLOPS (4:1) | 19.49 TFLOPS | 9.746 TFLOPS (1:2) | |
| 1 | NVIDIA A800 PCIe 80 GB | GA100 | 6912 | 80 | 5120 | 77.97 TFLOPS (4:1) | 19.49 TFLOPS | 9.746 TFLOPS (1:2) | |
| 1 | NVIDIA Tesla V100 PCIe 32 GB | GV100 | 5120 | 32 | 4096 | 28.26 TFLOPS (2:1) | 14.13 TFLOPS | 7.066 TFLOPS (1:2) | |
| 1 | NVIDIA Tesla T4 | TU104 | 2560 | 16 | 256 | 65.13 TFLOPS (8:1) | 8.141 TFLOPS | 254.4 GFLOPS (1:32) | |
| 1 | NVIDIA L40 | AD102 | 18176 | 48 | 384 | 90.52 TFLOPS (1:1) | 90.52 TFLOPS | 1,414 GFLOPS (1:64) | |
| 1 | NVIDIA L20 | AD102 | 11776 | 48 | 384 | 59.35 TFLOPS (1:1) | 59.35 TFLOPS | 927.4 GFLOPS (1:64) | |
| 1 | NVIDIA GeForce RTX 4090 | AD102 | 16384 | 24 | 384 | 82.58 TFLOPS (1:1) | 82.58 TFLOPS | 1,290 GFLOPS (1:64) | |
| NVIDIA GeForce RTX 4090 D | AD102 | 14592 | 24 | 384 | 73.54 TFLOPS (1:1) | 73.54 TFLOPS | 1,149 GFLOPS (1:64) | ||
| 1 | NVIDIA GeForce RTX 4090 Ti | AD102 | 18176 | 24 | 384 | 93.24 TFLOPS (1:1) | 93.24 TFLOPS | 1,457 GFLOPS (1:64) | |
| NVIDIA GeForce RTX 4090 Mobile | AD103 | 9728 | 16 | 256 | 32.98 TFLOPS (1:1) | 32.98 TFLOPS | 515.3 GFLOPS (1:64) | ||
| NVIDIA GeForce RTX 4090 Max-Q | AD103 | 9728 | 16 | 256 | 28.31 TFLOPS (1:1) | 28.31 TFLOPS | 442.3 GFLOPS (1:64) | ||
| NVIDIA GeForce RTX 4070 | AD104 | 5888 | 12 | 192 | 29.15 TFLOPS (1:1) | 29.15 TFLOPS | 455.4 GFLOPS (1:64) | ||
| NVIDIA GeForce RTX 4070 SUPER | AD104 | 7168 | 12 | 192 | 35.48 TFLOPS (1:1) | 35.48 TFLOPS | 554.4 GFLOPS (1:64) | ||
| NVIDIA GeForce RTX 4070 Ti | AD104 | 7680 | 12 | 192 | 40.09 TFLOPS (1:1) | 40.09 TFLOPS | 626.4 GFLOPS (1:64) | ||
| 1 | NVIDIA GeForce RTX 4070 Ti SUPER | AD103 | 8448 | 16 | 256 | 44.10 TFLOPS (1:1) | 44.10 TFLOPS | 689.0 GFLOPS (1:64) | |
| NVIDIA GeForce RTX 4060 | AD107 | 3072 | 8 | 128 | 15.11 TFLOPS (1:1) | 15.11 TFLOPS | 236.2 GFLOPS (1:64) | ||
| NVIDIA GeForce RTX 4060 Ti 8 GB | AD106 | 4352 | 8 | 128 | 22.06 TFLOPS (1:1) | 22.06 TFLOPS | 344.8 GFLOPS (1:64) | ||
| NVIDIA GeForce RTX 4060 Ti 16 GB | AD106 | 4352 | 16 | 128 | 22.06 TFLOPS (1:1) | 22.06 TFLOPS | 344.8 GFLOPS (1:64) | ||
| NVIDIA GeForce RTX 3090 | GA102 | 10496 | 24 | 384 | 35.58 TFLOPS (1:1) | 35.58 TFLOPS | 556.0 GFLOPS (1:64) | ||
| 1 | NVIDIA GeForce RTX 3090 Ti | GA102 | 10752 | 24 | 384 | 40.00 TFLOPS (1:1) | 40.00 TFLOPS | 625.0 GFLOPS (1:64) | |
| NVIDIA GeForce RTX 3080 | GA102 | 8704 | 10 | 320 | 29.77 TFLOPS (1:1) | 29.77 TFLOPS | 465.1 GFLOPS (1:64) | ||
| 1 | NVIDIA GeForce RTX 3080 Ti | GA102 | 10240 | 12 | 384 | 34.10 TFLOPS (1:1) | 34.10 TFLOPS | 532.8 GFLOPS (1:64) | |
| NVIDIA GeForce RTX 2080 | TU104 | 2944 | 8 | 256 | 20.14 TFLOPS (2:1) | 10.07 TFLOPS | 314.6 GFLOPS (1:32) | ||
| 1 | NVIDIA GeForce RTX 2080 Ti | TU102 | 4352 | 11 | 352 | 26.90 TFLOPS (2:1) | 13.45 TFLOPS | 420.2 GFLOPS (1:32) | |
| 1 | NVIDIA GeForce GTX 1080 Ti | GP102 | 3584 | 11 | 352 | 177.2 GFLOPS (1:64) | 11.34 TFLOPS | 354.4 GFLOPS (1:32) | |
| NVIDIA GeForce GTX 1070 Ti | GP104 | 2432 | 8 | 256 | 127.9 GFLOPS (1:64) | 8.186 TFLOPS | 255.8 GFLOPS (1:32) | ||
| NVIDIA GeForce GTX 1060 6 GB | GP106 | 1280 | 6 | 192 | 68.36 GFLOPS (1:64) | 4.375 TFLOPS | 136.7 GFLOPS (1:32) | ||
| NVIDIA GeForce GTX 1050 Ti | GP107 | 768 | 4 | 128 | 33.41 GFLOPS (1:64) | 2.138 TFLOPS | 66.82 GFLOPS (1:32) | ||
