Accelerator Choice Is Not Enough: AlphaFold2 Inference on Cloud TPUs
cs.DC, cs.LG, cs.PF
Submitted: 2026-09-28
Updated: 2026-09-29
Code: https://github.com/lorenzopazienza/alphafold-tpu-benchmark
Sources
- Duet Benchmarking: Improving Measurement Accuracy in the Cloud
- FastFold: Reducing AlphaFold Training Time from 11 Days to 67 Hours
- CCL-Bench 1.0: A Trace-Based Benchmark for LLM Infrastructure
- Performance Benchmarking of Tensor Trains for accelerated Quantum-Inspired Homogenization on TPU, GPU and CPU architectures
- TPU v4: An Optically Reconfigurable Supercomputer for Machine Learning with Hardware Support for Embeddings
- Fine-Tuning and Serving Gemma 4 31B on Google Cloud TPU: A Technical Comparison with GPU Baselines
- APACE: AlphaFold2 and advanced computing as a service for accelerated discovery in biophysics
- DrJAX: Scalable and Differentiable MapReduce Primitives in JAX
- Compiler-First State Space Duality and Portable $O(1)$ Autoregressive Caching for Inference
- JAXBench: Benchmarking Autonomous TPU Kernel Optimization
- HelixFold: An Efficient Implementation of AlphaFold2 using PaddlePaddle
- GSPMD: General and Scalable Parallelization for ML Computation Graphs
- ParaFold: Paralleling AlphaFold for Large-Scale Predictions
- ScaleFold: Reducing AlphaFold Initial Training Time to 10 Hours
Related papers
- iScheduler: Reinforcement Learning-Driven Continual Optimization for Large-Scale Resource Investment Problems
- SAMM: Sharded Automated Market Maker
- InferScale: GPU-Native KV Injection for Personalized LLM Serving
- Vigil: Accountable Liveness against Selective Silence
- Steelhead: Interleaving Partially Synchronous and Asynchronous Commit Rules on a Shared DAG
- Democratizing MoE inference on commodity GPUs with CoMoE