# Forecasting GPU Performance for Deep Learning Training and Inference

arXiv: https://arxiv.org/abs/2407.13853 · HTML: https://mlsanity.com/paper/2407.13853/

## Key facts

- Cited by 15 papers in the map (data through 14 Oct 2026).
- Builds on 12 of its 52 references that are in the map.
- First posted on arXiv 18 Jul 2024.

## Builds on (most cited first, 12 of 12)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://mlsanity.com/paper/1409.1556/) (2014)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model](https://mlsanity.com/paper/2201.11990/) (2022)
- [TPU v4: An Optically Reconfigurable Supercomputer for Machine Learning with Hardware Support for Embeddings](https://mlsanity.com/paper/2304.01433/) (2023)
- [Memory-Efficient Pipeline-Parallel DNN Training](https://mlsanity.com/paper/2006.09503/) (2020)
- [TopoOpt: Co-optimizing Network Topology and Parallelization Strategy for Distributed Training Jobs](https://mlsanity.com/paper/2202.00433/) (2022)
- [Daydream: Accurately Estimating the Efficacy of Optimizations for DNN Training](https://mlsanity.com/paper/2006.03318/) (2020)
- [A better measure of relative prediction accuracy for model selection and model estimation](https://mlsanity.com/paper/2105.05249/) (2021)
- [MAD Max Beyond Single-Node: Enabling Large Machine Learning Model Acceleration on Distributed Systems](https://mlsanity.com/paper/2310.02784/) (2023)

## Built on by (most cited first, 15 of 15)

- [Forecasting LLM Inference Performance via Hardware-Agnostic Analytical Modeling](https://mlsanity.com/paper/2508.00904/) (2025)
- [Revati: Transparent GPU-Free Time-Warp Emulation for LLM Serving](https://mlsanity.com/paper/2601.00397/) (2026)
- [Boosting LLM Serving through Spatial-Temporal GPU Resource Sharing](https://mlsanity.com/paper/2504.19516/) (2025)
- [Can Large Language Models Predict Parallel Code Performance?](https://mlsanity.com/paper/2505.03988/) (2025)
- [KernelSight-LM: A Kernel-Level LLM Inference Simulator](https://mlsanity.com/paper/2606.28565/) (2026)
- [Maya: Optimizing Deep Learning Training Workloads using GPU Runtime Emulation](https://mlsanity.com/paper/2503.20191/) (2025)
- [FASER: Fine-Grained Phase Management for Speculative Decoding in Dynamic LLM Serving](https://mlsanity.com/paper/2604.20503/) (2026)
- [Rethinking Dynamic Networks and Heterogeneous Computing with Automatic Parallelization](https://mlsanity.com/paper/2506.02787/) (2025)
- [Training Time Prediction for Mixed Precision-based Distributed Training](https://mlsanity.com/paper/2604.16145/) (2026)
- [Strait: Perceiving Priority and Interference in ML Inference Serving](https://mlsanity.com/paper/2604.28175/) (2026)
- [EnergyLens: Predictive Energy-Aware Exploration for Multi-GPU LLM Inference Optimization](https://mlsanity.com/paper/2605.14249/) (2026)
- [Leaky Language Models: Stealing Architecture and Inference Optimizations via Per-Token Timing](https://mlsanity.com/paper/2607.20723/) (2026)
- [Gleam: Adaptive Network-Efficient CUDA API Remoting for Cross-Device GPU Sharing over LANs](https://mlsanity.com/paper/2607.23115/) (2026)
- [CARB: A Characterization-Guided Framework for CNN Inference Cost Prediction and Deployment Screening](https://mlsanity.com/paper/2608.10506/) (2026)
- [Federated 3D Gaussian Splatting for Large-Scale Scene Reconstruction at Wireless Edge](https://mlsanity.com/paper/2609.32177/) (2026)

## Abstract (arXiv)

Deep learning kernels exhibit predictable memory accesses and compute patterns, making GPUs' parallel architecture well-suited for their execution. Software and runtime systems for GPUs are optimized to better utilize the stream multiprocessors, on-chip cache, and off-chip high-bandwidth memory. As deep learning models and GPUs evolve, access to newer GPUs is often limited, raising questions about the performance of new model architectures on existing GPUs, existing models on new GPUs, and new model architectures on new GPUs. To address these questions, we introduce NeuSight, a framework to predict the performance of various deep learning models, for both training and inference, on unseen GPUs without requiring actual execution. The framework leverages both GPU hardware behavior and software library optimizations to estimate end-to-end performance. Previous work uses regression models that capture linear trends or multilayer perceptrons to predict the overall latency of deep learning kernels on GPUs. These approaches suffer from higher error percentages when forecasting performance on unseen models and new GPUs. Instead, NeuSight decomposes the prediction problem into smaller problems, bounding the prediction through fundamental performance laws. NeuSight decomposes a single deep learning kernel prediction into smaller working sets called tiles, which are executed independently on the GPU. Tile-granularity predictions are determined using a machine learning approach and aggregated to estimate end-to-end latency. NeuSight outperforms prior work across various deep learning workloads and the latest GPUs. It reduces the percentage error from 121.4% and 30.8% to 2.3% in predicting the latency of GPT3 model for training and inference on H100, compared to state-of-the-art prior work, where both GPT3 and H100 were not used to train the framework.