I noticed that version 1.0.0 of the cuTENSOR library requires GPUs with compute capability 7, while earlier versions (at least 0.2, I didn’t check all of them) ran fine with compute capability 6.1. We have a cluster with 72 GPUs with 6.1, and the cuTENSOR library is extremely useful for our simulations (we actually use it through the TensorOperations.jl package of the Julia programming language). Is there any possibility of supporting GPUs with compute capability 6.1 again in future versions of cuTENSOR?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Cutensor errors on compute capability 6.1 card (Quadro P6000) | 1 | 45 | June 3, 2026 | |
| Just Released: cuTENSOR V1.5 | 0 | 268 | June 21, 2022 | |
| Is cuTENSOR supported on the Jetson AGX Xavier? | 1 | 678 | August 10, 2022 | |
| cuTENSOR 2.0: A Comprehensive Guide for Accelerating Tensor Computations | 0 | 352 | March 9, 2024 | |
| Programming Distributed Multi-GPU Tensor Operations with cuTENSOR v1.4 | 0 | 423 | November 29, 2021 | |
| cuTENSOR v1.3.0 Now Available: Up to 2x Performance | 0 | 389 | April 28, 2021 | |
| understanding the CUDA CUDA Toolkit Number and the CUDA toolkit version | 5 | 2849 | September 26, 2016 | |
| cuTENSOR 2.0: Applications and Performance | 0 | 305 | March 9, 2024 | |
| CUDA compatibility with 1050Ti | 2 | 4209 | March 27, 2017 | |
| Compatibility problems | 2 | 388 | May 5, 2025 |