Accelerating Frontier MoE Training with 3D Integrated Optics

• URL: http://arxiv.org/abs/2510.15893v1
• Date: Tue, 09 Sep 2025 00:41:42 GMT
• Title: Accelerating Frontier MoE Training with 3D Integrated Optics
• Authors: Mikhail Bernadskiy, Peter Carson, Thomas Graham, Taylor Groves, Ho John Lee, Eric Yeh,
• Abstract summary: 3D-stacked optics and logic offers a transformative, power-efficient scale-up solution for connecting hundreds of GPU packages.<n>We show that the substantial increases in bandwidth and radix enabled by 3D CPO allow for an 8X increase in scale-up capability.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The unabated growth in AI workload demands is driving the need for concerted advances in compute, memory, and interconnect performance. As traditional semiconductor scaling slows, high-speed interconnects have emerged as the new scaling engine, enabling the creation of larger logical GPUs by linking many GPUs into a single, low-latency, high-bandwidth compute domain. While initial scale-up fabrics leveraged copper interconnects for their power and cost advantages, the maximum reach of passive electrical interconnects (approximately 1 meter) effectively limits the scale-up domain to within a single rack. The advent of 3D-stacked optics and logic offers a transformative, power-efficient scale-up solution for connecting hundreds of GPU packages (thousands of GPUs) across multiple data center racks. This work explores the design tradeoffs of scale-up technologies and demonstrates how frontier LLMs necessitate novel photonic solutions to achieve aggressive power and performance targets. We model the benefits of 3D CPO (Passage) enabled GPUs and switches within the scale-up domain when training Frontier Mixture of Experts (MoE) models exceeding one trillion parameters. Our results show that the substantial increases in bandwidth and radix enabled by 3D CPO allow for an 8X increase in scale-up capability. This affords new opportunities for multi-dimensional parallelism within the scale-up domain and results in a 2.7X reduction in time-to-train, unlocking unprecedented model scaling.

Related papers

• Scalable Generative Game Engine: Breaking the Resolution Wall via Hardware-Algorithm Co-Design [17.941176878609337]
  We bridge the gap between generative models and high-resolution neural simulations by introducing a scalable textitHardware-Algorithm Co-Design framework.<n>Our system delivers fluid 26.4 FPS and 48.3 FPS respectively, with an amortized effective latency of 2.7 ms.
  arXiv  Detail & Related papers  (2026-01-31T08:52:51Z)
• Distributed Equivariant Graph Neural Networks for Large-Scale Electronic Structure Prediction [76.62155593340763]
  Equivariant Graph Neural Networks (eGNNs) trained on density-functional theory (DFT) data can potentially perform electronic structure prediction at unprecedented scales.<n>However, the graph representations required for this task tend to be densely connected.<n>We present a distributed eGNN implementation which leverages direct GPU communication and introduce a partitioning strategy of the input graph.
  arXiv  Detail & Related papers  (2025-07-04T23:53:47Z)
• MegaScale-Infer: Serving Mixture-of-Experts at Scale with Disaggregated Expert Parallelism [26.923312725688735]
  Mixture-of-Experts (MoE) showcases tremendous potential to scale large language models (LLMs) with enhanced performance and reduced computational complexity.<n>We present MegaScale-Infer, an efficient and cost-effective system for serving large-scale MoE models.
  arXiv  Detail & Related papers  (2025-04-03T04:20:44Z)
• Transolver++: An Accurate Neural Solver for PDEs on Million-Scale Geometries [67.63077028746191]
  Transolver++ is a highly parallel and efficient neural solver that can solve PDEs on million-scale geometries.<n>Transolver++ increases the single- GPU input capacity to million-scale points for the first time.<n>It achieves over 20% performance gain in million-scale high-fidelity industrial simulations.
  arXiv  Detail & Related papers  (2025-02-04T15:33:50Z)
• Scaling Large Language Model Training on Frontier with Low-Bandwidth Partitioning [2.685330831042324]
  We propose a collection of communication and optimization strategies for ZeRO++ to reduce communication costs and improve memory utilization.<n>For a 20B GPT model, we observe a 1.71x increase in TFLOPS per GPU when compared with ZeRO++ up to 384 GCDs and a scaling efficiency of 0.94 for up to 384 GCDs.
  arXiv  Detail & Related papers  (2025-01-08T04:19:57Z)
• MoE-Lightning: High-Throughput MoE Inference on Memory-constrained GPUs [55.95879347182669]
  MoE architecture is renowned for its ability to increase model capacity without a proportional increase in inference cost. MoE-Lightning introduces a novel CPU-GPU-I/O pipelining schedule, CGOPipe, with paged weights to achieve high resource utilization. MoE-Lightning can achieve up to 10.3x higher throughput than state-of-the-art offloading-enabled LLM inference systems for Mixtral 8x7B on a single T4 GPU (16GB)
  arXiv  Detail & Related papers  (2024-11-18T01:06:12Z)
• FusionLLM: A Decentralized LLM Training System on Geo-distributed GPUs with Adaptive Compression [55.992528247880685]
  Decentralized training faces significant challenges regarding system design and efficiency. We present FusionLLM, a decentralized training system designed and implemented for training large deep neural networks (DNNs) We show that our system and method can achieve 1.45 - 9.39x speedup compared to baseline methods while ensuring convergence.
  arXiv  Detail & Related papers  (2024-10-16T16:13:19Z)
• ExtremeMETA: High-speed Lightweight Image Segmentation Model by Remodeling Multi-channel Metamaterial Imagers [8.976310466890805]
  We propose a large kernel lightweight segmentation model, ExtremeMETA, based on the ExtremeC3Net. Results show that the optimized efficient design improved segmentation performance from 92.45 to 95.97 on mIoU while reducing computational FLOPs from 461.07 MMacs to 166.03 MMacs.
  arXiv  Detail & Related papers  (2024-05-27T18:03:37Z)
• GLEAM: Greedy Learning for Large-Scale Accelerated MRI Reconstruction [50.248694764703714]
  Unrolled neural networks have recently achieved state-of-the-art accelerated MRI reconstruction. These networks unroll iterative optimization algorithms by alternating between physics-based consistency and neural-network based regularization. We propose Greedy LEarning for Accelerated MRI reconstruction, an efficient training strategy for high-dimensional imaging settings.
  arXiv  Detail & Related papers  (2022-07-18T06:01:29Z)
• EfficientViT: Multi-Scale Linear Attention for High-Resolution Dense Prediction [67.11722682878722]
  This work presents EfficientViT, a new family of high-resolution vision models with novel multi-scale linear attention. Our multi-scale linear attention achieves the global receptive field and multi-scale learning. EfficientViT delivers remarkable performance gains over previous state-of-the-art models.
  arXiv  Detail & Related papers  (2022-05-29T20:07:23Z)
• Harmony: Overcoming the hurdles of GPU memory capacity to train massive DNN models on commodity servers [13.620650014358413]
  Deep neural networks (DNNs) have grown exponentially in complexity and size over the past decade. One of the main challenges for researchers who might have access to only limited resources is limited memory capacity compared to model size.
  arXiv  Detail & Related papers  (2022-02-02T22:16:27Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.