179 accelerators and GPUs tracked by Epoch AI
| Name | FP16 FLOP/s | Price |
|---|---|---|
| NVIDIA Vera Rubin Superchip | 4000.0T | — |
| AMD Instinct MI355X | 2516.6T | $30.0K |
| NVIDIA GB300 (Blackwell Ultra) | 2500.0T | $43.0K |
| NVIDIA GB200 | 2500.0T | $37.0K |
| AMD Instinct MI350X | 2309.6T | $25.0K |
| Google TPU v7 Ironwood | 2307.0T | $12.0K |
| NVIDIA B200 | 2250.0T | — |
| NVIDIA B300 (Blackwell Ultra) | 2250.0T | — |
| NVIDIA B100 | 1750.0T | $35.0K |
| Intel Habana Gaudi3 | 1678.0T | — |
| AMD Instinct MI325X | 1307.4T | — |
| AMD Instinct MI300X | 1307.4T | $15.0K |
| Microsoft Maia 200 | 1268.0T | — |
| Biren BR100 | 1024.0T | — |
| NVIDIA GH200 | 989.5T | — |
| NVIDIA H200 SXM | 989.5T | $30.0K |
| NVIDIA GH100 | 989.5T | — |
| NVIDIA H100 SXM5 80GB | 989.4T | $33.6K |
| NVIDIA H800 SXM5 | 989.0T | — |
| AMD Instinct MI300A | 980.6T | — |
| Google TPU v6e Trillium | 918.0T | — |
| Huawei Ascend 920 | 900.0T | — |
| Huawei Ascend 910C | 800.0T | — |
| Maia 100 (M100) | 800.0T | — |
| NVIDIA H100 PCIe | 756.0T | — |
| Amazon Trainium3 | 671.0T | — |
| Amazon Trainium2 | 667.0T | — |
| AMD Radeon Instinct MI308X | 653.7T | — |
| Meta MTIA 300 | 600.0T | — |
| NVIDIA RTX Pro 6000 Blackwell | 503.8T | $13.3K |
| Google TPU v5p | 459.0T | — |
| Intel Habana Gaudi2 | 450.0T | — |
| AMD Radeon Instinct MI250X | 383.0T | $14.5K |
| Tesla D1 Dojo | 362.0T | — |
| NVIDIA GeForce RTX 4090 | 330.0T | $1.6K |
| NVIDIA A100 SXM4 80 GB | 312.0T | — |
| NVIDIA A100 PCIe | 312.0T | — |
| NVIDIA A100 SXM4 40 GB | 312.0T | $15.0K |
| NVIDIA A100 | 312.0T | — |
| NVIDIA A800 PCIe 80 GB | 312.0T | — |
| NVIDIA A800 PCIe 40 GB | 311.8T | — |
| NVIDIA A800 SXM | 311.8T | — |
| Google TPU v4 | 275.0T | — |
| Google TPU v5e | 197.0T | — |
| AWS Inferentia2 | 190.0T | — |
| Amazon Trainium1 | 190.0T | — |
| NVIDIA L40 | 181.0T | — |
| Meta MTIA v2 (MTIA 200) | 177.0T | — |
| Meta MTIA v2 | 177.0T | — |
| NVIDIA GeForce RTX 3090 Ti | 160.0T | $2.0K |
| NVIDIA A40 PCIe | 149.7T | — |
| NVIDIA HGX H20 | 148.0T | $12.0K |
| Google TPU v4i | 138.0T | — |
| NVIDIA Tesla V100S PCIe 32 GB | 130.0T | — |
| NVIDIA Tesla V100 SXM2 16 GB | 125.0T | — |
| NVIDIA Tesla V100 DGXS 16 GB | 125.0T | — |
| NVIDIA Tesla V100 DGXS 32 GB | 125.0T | — |
| NVIDIA Tesla V100 SXM3 32 GB | 125.0T | — |
| NVIDIA Tesla V100 SXM2 32 GB | 125.0T | $11.5K |
| NVIDIA Tesla V100 SXM2 | 125.0T | $11.5K |
| NVIDIA V100 | 125.0T | — |
| Google TPU v3 | 123.0T | — |
| NVIDIA L4 | 121.0T | — |
| NVIDIA L20 PCle | 119.5T | $4.3K |
| NVIDIA GeForce RTX 2080 Ti 11GB | 113.8T | $999 |
| NVIDIA Tesla V100 PCIe 16 GB | 112.0T | $10.7K |
| NVIDIA Tesla V100 PCIe 32 GB | 112.0T | $11.5K |
| MTT S4000 | 100.0T | — |
| NVIDIA L2 PCle | 96.5T | — |
| Meta MTIA v1 | 51.2T | — |
| Meta MTIA v1 (MTIA 100) | 51.2T | — |
| Google TPU v2 | 46.0T | — |
| NVIDIA VR200 | — | — |
| NVIDIA GeForce RTX 3080 | — | $699 |
| Huawei Ascend 910A | — | — |
| Huawei Ascend 910A Pro | — | — |
| Baidu Kunlun I | — | — |
| NVIDIA GeForce RTX 3090 | — | $1.5K |
| Huawei Ascend 910A Premium | — | — |
| NVIDIA Quadro RTX 4000 | — | $899 |
| Baidu Kunlun K200 | — | — |
| Huawei Ascend 910 | — | — |
| Huawei Ascend 310 | — | — |
| Huawei Ascend 910 (320 TFLOPS) | — | — |
| Tianshu Zhixin Tiangai 100 (BI-V100) | — | — |
| T-Head Hanguang 800 | — | — |
| Cambricon MLU270-F4 | — | $2.1K |
| NVIDIA Tesla T4 | — | — |
| NVIDIA Quadro P600 | — | $178 |
| Cambricon MLU270-S4 | — | — |
| NVIDIA Quadro RTX 6000 | — | $6.3K |
| NVIDIA Quadro P4000 | — | $815 |
| MT-2000 | — | — |
| NVIDIA Quadro P5000 | — | $2.5K |
| NVIDIA Quadro P6000 | — | $6.0K |
| NVIDIA P40 | — | $5.7K |
| Baidu Kunlun R100 (100W) | — | — |
| Cambricon MLU370-S4/S8 | — | — |
| Cambricon MLU370-X4 | — | — |
| Cambricon MLU370-S8 | — | — |
| Cambricon MLU370-S4 | — | — |
| NVIDIA GeForce GTX TITAN | — | $999 |
| NVIDIA Tesla K20c | — | $3.2K |
| NVIDIA Tesla K20X | — | $7.7K |
| NVIDIA Tesla C2050 | — | — |
| NVIDIA GeForce GTX 580 | — | $499 |
| NVIDIA GeForce GTX 280 | — | $649 |
| NVIDIA Tesla P100 SXM2 | — | — |
| NVIDIA P100 | — | $5.7K |
| NVIDIA Tesla P100 DGXS | — | — |
| NVIDIA M40 | — | $10.8K |
| NVIDIA TESLA M60 | — | — |
| NVIDIA Quadro M4000 | — | — |
| Google TPU v1 | — | — |
| Huawei Ascend 910B Pro | — | — |
| Huawei Ascend 910B1 (v1) | — | — |
| NVIDIA GeForce GTX TITAN X | — | $999 |
| NVIDIA Quadro K1200 | — | — |
| Huawei Ascend 910B1 (v2) | — | — |
| NVIDIA Tesla K80 | — | $5.0K |
| Baidu Kunlun II | — | — |
| NVIDIA GeForce GTX 980 | — | $549 |
| NVIDIA GeForce GTX Titan Black | — | $999 |
| Shensuan-1 | — | — |
| Huawei Ascend 910B2 | — | — |
| NVIDIA Tesla K40s | — | $7.7K |
| NVIDIA Tesla K40t | — | $7.7K |
| NVIDIA Quadro K6000 | — | $5.3K |
| Huawei Ascend 910B3 | — | — |
| Huawei Ascend 910B4 | — | — |
| Huawei Ascend 800T A2 | — | — |
| NVIDIA RTX A4000 | — | $1.0K |
| NVIDIA Tesla P4 | — | — |
| NVIDIA Tesla P100 PCIe 16GB | — | — |
| NVIDIA Tesla P100 PCIe 12GB | — | — |
| NVIDIA GeForce GTX 780 | — | $649 |
| NVIDIA Quadro K4000 | — | $1.3K |
| NVIDIA P100 PCIe 16GB | — | — |
| NVIDIA GeForce GTX 1080 | — | $599 |
| NVIDIA Rubin CPX | — | — |
| Huawei Ascend 950DT | — | — |
| Huawei Ascend 970 | — | — |
| Huawei Ascend 960 | — | — |
| Cambricon Siyuan 690 | — | — |
| NVIDIA VR300 (Rubin Ultra) | — | — |
| MT-3000 | — | — |
| Cambricon MLU370-X8 | — | — |
| Huawei Ascend 910B | — | $16.9K |
| Baidu Kunlun R100 (75W) | — | — |
| Cambricon Siyuan 590 | — | — |
| NVIDIA H100 NVL | — | — |
| Sunway SW26010-Pro | — | — |
| MetaX MXC500 (PCIe) | — | — |
| MetaX MXN100 | — | — |
| NVIDIA GeForce RTX 4070 | — | — |
| Intel Data Center GPU Max 1100 | — | — |
| Intel Data Center GPU Max 1350 | — | — |
| Intel Data Center GPU Max 1550 | — | — |
| AMD Radeon Instinct MI300 | — | — |
| Baidu Kunlun RG800 | — | — |
| MetaX MXC500 (OAM) | — | — |
| Tianshu Zhixin Zhikai 100 | — | — |
| NVIDIA RTX 6000 Ada Generation | — | $6.8K |
| AMD Radeon RX 7900 XTX | — | $999 |
| NVIDIA GeForce RTX 4080 | — | $1.2K |
| NVIDIA Titan V | — | $3.0K |
| PEZY-SC2 | — | — |
| NVIDIA TITAN Xp | — | $1.2K |
| NVIDIA GeForce GTX 1080 Ti | — | $699 |
| Intel Habana GAUDI HL-205 | — | — |
| NVIDIA Quadro RTX 5000 | — | $2.3K |
| NVIDIA Quadro RTX 8000 | — | — |
| Baidu Kunlun R200 (150W) | — | — |
| Baidu Kunlun R200-8F (160W) | — | — |
| NVIDIA A10G | — | — |
| NVIDIA RTX A5000 | — | $2.3K |
| NVIDIA A10 PCIe | — | $2.8K |
| Cambricon MLU290-M5 | — | — |
| NVIDIA RTX A6000 | — | $4.6K |