XPU Catalog
Browse all AI compute accelerators with verified specs and benchmarks
Found 72 XPUs
| Vendor↑ | Model↕ | Architecture↕ | Form Factor↕ | VRAM (GB)↕ | TFLOPs↕ | TDP (W)↕ | Launch Date↕ | Actions |
|---|---|---|---|---|---|---|---|---|
| Alibaba | Hanguang 800 | XuanTie | — | — | — | 150 | Sep 2019 | Details |
| AMD | MI100 | CDNA | PCIe | 32 | 23.1 | 300 | Nov 2020 | Details |
| AMD | MI210 | CDNA 2 | PCIe | 64 | 181 | 300 | Jan 2022 | Details |
| AMD | MI250X | CDNA 2 | OAM | 128 | 383 | 560 | Nov 2021 | Details |
| AMD | MI300A | CDNA 3 | OAM | 128 | 980.6 | 760 | Dec 2023 | Details |
| AMD | MI300X | CDNA 3 | OAM | 192 | 1,307 | 750 | Dec 2023 | Details |
| AMD | MI325X | CDNA 3.5 | OAM | 256 | 1,400 | 800 | Oct 2024 | Details |
| AMD | MI350X | CDNA 4 | OAM | 288 | 2,100 | 850 | Jun 2025 | Details |
| AMD | Radeon PRO W7900 | RDNA 3 | PCIe | 48 | 122 | 295 | Apr 2023 | Details |
| AMD | Radeon RX 7900 XT | RDNA 3 | PCIe | 20 | 104 | 315 | Dec 2022 | Details |
| AMD | Radeon RX 7900 XTX | RDNA 3 | PCIe | 24 | 122 | 355 | Dec 2022 | Details |
| AWS | Inferentia2 | Inferentia Gen2 | — | 32 | 190 | 150 | Nov 2022 | Details |
| AWS | Trainium | Inferentia/Trainium | — | 32 | 190 | 200 | Nov 2021 | Details |
| AWS | Trainium2 | Trainium Gen2 | — | — | 680 | 300 | Nov 2023 | Details |
| Baidu | Kunlun II | Kunlun Core | — | 32 | — | 200 | Aug 2021 | Details |
| Biren Technology | BR100 | Biren GPU | — | 64 | — | 550 | Aug 2022 | Details |
| Cambricon | MLU370 | MLU Core | — | 48 | 256 | 300 | Nov 2022 | Details |
| Cerebras | WSE-3 | Wafer Scale Engine | — | 44 | — | 23000 | Mar 2024 | Details |
| Enflame Technology | CloudBlazer T20 | GCU (AI Graphics Unit) | — | 32 | — | 250 | Nov 2021 | Details |
| Etched | Sohu | Transformer ASIC | PCIe | 144 | 10,000 | 700 | Jun 2024 | Details |
| FuriosaAI | Warboy | FuriosaAI NPU | — | 32 | — | 80 | Oct 2021 | Details |
| TPU v4 | TPU v4 | Mezzanine | — | 275 | 300 | May 2021 | Details | |
| TPU v5e | TPU v5 | Mezzanine | — | 197 | 250 | Aug 2023 | Details | |
| TPU v5p | TPU v5 | Mezzanine | — | 459 | 400 | Aug 2023 | Details | |
| TPU v6e (Trillium) | Trillium | Custom | 32 | 918 | 200 | May 2024 | Details | |
| Graphcore | Bow IPU | Bow-2000 | — | 3.6 | — | 200 | Mar 2022 | Details |
| Graphcore | IPU-M2000 | Colossus Mk2 | — | 3.6 | — | 180 | Jul 2020 | Details |
| Groq | LPU Inference Engine | TSP (Tensor Streaming Processor) | — | 230 | — | 300 | Feb 2024 | Details |
| Huawei | Ascend 910B | Da Vinci | — | 64 | — | 400 | Aug 2023 | Details |
| Iluvatar CoreX | BI-V150 | Iluvatar GPU | — | 48 | 300 | 350 | Apr 2023 | Details |
| Intel | Data Center GPU Max 1100 | Ponte Vecchio | OAM | 48 | 177 | 300 | Jan 2023 | Details |
| Intel | Data Center GPU Max 1550 | Ponte Vecchio | OAM | 128 | 419 | 600 | Jan 2023 | Details |
| Intel Habana | Gaudi 2 | Gaudi Gen2 | OAM | 96 | 432 | 600 | May 2022 | Details |
| Intel Habana | Gaudi 3 | Gaudi Gen3 | OAM | 128 | 1,835 | 900 | Apr 2024 | Details |
| Meta | MTIA v1 | Meta Training & Inference Accelerator | — | 128 | — | 400 | May 2023 | Details |
| Microsoft | Maia 100 | Maia | — | 64 | 700 | 500 | Nov 2023 | Details |
| Moore Threads | MTT S80 | Chunxiao | — | 32 | — | 300 | Mar 2022 | Details |
| NVIDIA | A10 | Ampere | PCIe | 24 | 125 | 150 | Apr 2021 | Details |
| NVIDIA | A100 SXM | Ampere | SXM | 80 | 312 | 400 | May 2020 | Details |
| NVIDIA | A40 | Ampere | PCIe | 48 | 150 | 300 | Oct 2020 | Details |
| NVIDIA | B200 | Blackwell | SXM | 192 | 2,250 | 1000 | Mar 2024 | Details |
| NVIDIA | GB200 NVL72 | Blackwell | Custom | 13824 | 360,000 | 120000 | Mar 2024 | Details |
| NVIDIA | GB200 Superchip | Blackwell | SXM | 384 | 5,000 | 2700 | Mar 2024 | Details |
| NVIDIA | GeForce RTX 4060 Ti | Ada Lovelace | PCIe | 8 | 44.2 | 160 | May 2023 | Details |
| NVIDIA | GeForce RTX 4070 | Ada Lovelace | PCIe | 12 | 58.2 | 200 | Apr 2023 | Details |
| NVIDIA | GeForce RTX 4070 Super | Ada Lovelace | PCIe | 12 | 71 | 220 | Jan 2024 | Details |
| NVIDIA | GeForce RTX 4070 Ti | Ada Lovelace | PCIe | 12 | 80.2 | 285 | Jan 2023 | Details |
| NVIDIA | GeForce RTX 4070 Ti Super | Ada Lovelace | PCIe | 16 | 88.2 | 285 | Jan 2024 | Details |
| NVIDIA | GeForce RTX 4080 | Ada Lovelace | PCIe | 16 | 97.5 | 320 | Nov 2022 | Details |
| NVIDIA | GeForce RTX 4080 Super | Ada Lovelace | PCIe | 16 | 104.4 | 320 | Jan 2024 | Details |
| NVIDIA | GeForce RTX 4090 | Ada Lovelace | PCIe | 24 | 165.2 | 450 | Oct 2022 | Details |
| NVIDIA | GeForce RTX 5070 | Blackwell | PCIe | 12 | 61.6 | 250 | Feb 2025 | Details |
| NVIDIA | GeForce RTX 5070 Ti | Blackwell | PCIe | 16 | 88 | 300 | Feb 2025 | Details |
| NVIDIA | GeForce RTX 5080 | Blackwell | PCIe | 16 | 112.6 | 360 | Jan 2025 | Details |
| NVIDIA | GeForce RTX 5090 | Blackwell | PCIe | 32 | 209.5 | 575 | Jan 2025 | Details |
| NVIDIA | H100 PCIe | Hopper | PCIe | 80 | 1,513 | 350 | Sep 2022 | Details |
| NVIDIA | H100 SXM | Hopper | SXM | 80 | 1,979 | 700 | Sep 2022 | Details |
| NVIDIA | H200 PCIe | Hopper | PCIe | 141 | 1,513 | 350 | Mar 2024 | Details |
| NVIDIA | H200 SXM | Hopper | SXM | 141 | 1,979 | 700 | Nov 2023 | Details |
| NVIDIA | L4 | Ada Lovelace | PCIe | 24 | 121 | 72 | Mar 2023 | Details |
| NVIDIA | L40S | Ada Lovelace | PCIe | 48 | 733 | 350 | Oct 2023 | Details |
| NVIDIA | RTX 4000 Ada Generation | Ada Lovelace | PCIe | 20 | 53.4 | 130 | Apr 2023 | Details |
| NVIDIA | RTX 5000 Ada Generation | Ada Lovelace | PCIe | 32 | 130.6 | 250 | Aug 2023 | Details |
| NVIDIA | RTX 6000 Ada Generation | Ada Lovelace | PCIe | 48 | 182.2 | 300 | Dec 2022 | Details |
| NVIDIA | RTX PRO 6000 Blackwell Max-Q | Blackwell | PCIe | 96 | 125 | 300 | Mar 2025 | Details |
| NVIDIA | RTX PRO 6000 Blackwell Server Edition | Blackwell | PCIe | 96 | 250 | 600 | Mar 2025 | Details |
| NVIDIA | RTX PRO 6000 Blackwell Workstation Edition | Blackwell | PCIe | 96 | 250 | 600 | Mar 2025 | Details |
| Qualcomm | Cloud AI 100 | Qualcomm AI | PCIe | 16 | 50 | 75 | Sep 2020 | Details |
| Rebellions | ATOM | Rebellions NPU | — | — | — | 200 | Jun 2023 | Details |
| SambaNova | SN40L | Reconfigurable Dataflow Unit (RDU) | — | 640 | — | 700 | Jan 2024 | Details |
| Tenstorrent | Grayskull | Tensix Core | — | 8 | 200 | 65 | Mar 2021 | Details |
| Tenstorrent | Wormhole | Tensix Core | — | 24 | 364 | 160 | Oct 2023 | Details |
72
Total XPUs
25
Vendors
114
Benchmarks
39
Price Points