Pith. sign in

REVIEW 9 cited by

TPU v4: An Optically Reconfigurable Supercomputer for Machine Learning with Hardware Support for Embeddings

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2304.01433 v3 pith:VGAJPDPU submitted 2023-04-04 cs.AR cs.AIcs.LGcs.PF

Norman P. Jouppi , George Kurian , Sheng Li , Peter Ma , Rahul Nagarajan , Lifeng Nai , Nishant Patil , Suvinay Subramanian
show 6 more authors
Andy Swing Brian Towles Cliff Young Xiang Zhou Zongwei Zhou David Patterson
This is my paper · ORCID
classification cs.ARcs.AIcs.LGcs.PF
keywords powerfastermodelslesssupercomputerembeddingsgooglelearning
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original abstract

In response to innovations in machine learning (ML) models, production workloads changed radically and rapidly. TPU v4 is the fifth Google domain specific architecture (DSA) and its third supercomputer for such ML models. Optical circuit switches (OCSes) dynamically reconfigure its interconnect topology to improve scale, availability, utilization, modularity, deployment, security, power, and performance; users can pick a twisted 3D torus topology if desired. Much cheaper, lower power, and faster than Infiniband, OCSes and underlying optical components are <5% of system cost and <3% of system power. Each TPU v4 includes SparseCores, dataflow processors that accelerate models that rely on embeddings by 5x-7x yet use only 5% of die area and power. Deployed since 2020, TPU v4 outperforms TPU v3 by 2.1x and improves performance/Watt by 2.7x. The TPU v4 supercomputer is 4x larger at 4096 chips and thus ~10x faster overall, which along with OCS flexibility helps large language models. For similar sized systems, it is ~4.3x-4.5x faster than the Graphcore IPU Bow and is 1.2x-1.7x faster and uses 1.3x-1.9x less power than the Nvidia A100. TPU v4s inside the energy-optimized warehouse scale computers of Google Cloud use ~3x less energy and produce ~20x less CO2e than contemporary DSAs in a typical on-premise data center.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 9 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. OpenAlex reports about 19 citations worldwide. Full citation record

  1. Atleus: Accelerating Transformers on the Edge Enabled by 3D Heterogeneous Manycore Architectures

    cs.AR 2025-01 conditional novelty 7.0 of 10

    A 3D heterogeneous ReRAM-plus-systolic-array accelerator that claims up to 56x speedup and 64.5x energy efficiency over GPUs for transformer fine-tuning and inference.

  2. Machine Learning Fleet Efficiency: Analyzing and Optimizing Large-Scale Google TPU Systems with ML Productivity Goodput

    cs.LG 2025-02 conditional novelty 6.0 of 10

    Google proposes ML Productivity Goodput, a product of scheduling, runtime, and program goodputs, as a fleet-level metric for identifying and tracking efficiency improvements in large ML accelerator fleets.

  3. Towards Resource-Efficient Compound AI Systems

    cs.DC 2025-01 conditional novelty 6.0 of 10

    A declarative workflow system with an adaptive runtime can cut compound AI workflow completion time and energy use by auto-selecting resources and parallelism.

  4. HalleluBERT: Let Every Token That Has Meaning Bear Its Weight

    cs.CL 2025-10 conditional novelty 5.0 of 10

    HalleluBERT, a Hebrew-only RoBERTa encoder family trained from scratch at scale, reports the highest unweighted mean scores on BMC, NEMO, and SMCD benchmarks, but without statistical significance testing.

  5. SindBERT, the Sailor: Charting the Seas of Turkish NLP

    cs.CL 2025-10 conditional novelty 5.0 of 10

    SindBERT releases Turkish RoBERTa base/large models trained on 312GB of text; they match existing models, with the large variant best on two of four tasks and little scaling gain.

  6. Toward Lifelong-Sustainable Electronic-Photonic AI Systems via Extreme Efficiency, Reconfigurability, and Robustness

    physics.optics 2025-09 conditional novelty 4.0 of 10

    Photonic-electronic AI accelerators can amortize embodied carbon across longer, more flexible lifetimes if co-designed for efficiency and reconfigurability.

  7. TriADA: Massively Parallel Trilinear Matrix-by-Tensor Multiply-Add Algorithm and Device Architecture for the Acceleration of 3D Discrete Transformations

    cs.DC 2025-06 conditional novelty 4.0 of 10

    The paper proposes a triple-stage outer-product algorithm and an isomorphic 3D mesh architecture that computes separable 3D orthogonal transforms in N1+N2+N3 time steps.

  8. Fast Link Recovery via PTP-synchronized Nanosecond Optical Switching

    eess.SY 2024-12 reject novelty 4.0 of 10

    PTP-synchronized optical switches achieve roughly 62 ns jitter and restore a failed link in 2.7 ms, with scheduled recovery also available.

  9. A Survey on Efficiency Optimization Techniques for DNN-based Video Analytics: Process Systems, Algorithms, and Applications

    cs.CV 2025-07 conditional novelty 3.0 of 10

    A bottom-up survey of efficiency optimization techniques for DNN-based video analytics, spanning storage, computing, algorithms, and applications.

Pith tools