The DOI matches its Crossref record; 23 of the 23 references it cites were found in a registry or an index.
Evaluating Modern GPU Interconnect: PCIe, NVLink, NV-SLI, NVSwitch and GPUDirect
Open accessAbstract
High performance multi-GPU computing becomes an inevitable trend due to the ever-increasing demand on computation capability in emerging domains such as deep learning, big data and planet-scale simulations. However, the lack of deep understanding on how modern GPUs can be connected and the real impact of state-of-the-art interconnect technology on multi-GPU application performance become a hurdle. In this paper, we fill the gap by conducting a thorough evaluation on five latest types of modern GPU interconnects: PCIe, NVLink-V1, NVLink-V2, NVLink-SLI and NVSwitch, from six high-end servers and HPC platforms: NVIDIA P100-DGX-1, V100-DGX-1, DGX-2, OLCF's SummitDev and Summit supercomputers, as well as an SLI-linked system with two NVIDIA Turing RTX-2080 GPUs. Based on the empirical evaluation, we have observed four new types of GPU communication network NUMA effects: three are triggered by NVLink's topology, connectivity and routing, while one is caused by PCIe chipset design issue. These observations indicate that, for an application running in a multi-GPU node, choosing the right GPU combination can impose considerable impact on GPU communication efficiency, as well as the application's overall performance. Our evaluation can be leveraged in building practical multi-GPU performance models, which are vital for GPU task allocation, scheduling and migration in a shared environment (e.g., AI cloud and HPC centers), as well as communication-oriented performance tuning.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
A Proof for the Queuing Formula: L = λ W
Little, J. D. C. (1961) · Operations Research · doi:10.1287/opre.9.3.383 · Cited by 2,777
Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour
Goyal, P., Dollár, P., Girshick, R., Noordhuis, P., Wesolowski, L., Kyrola, A., Tulloch, A., Jia, Y., & He, K. (2017) · arXiv · doi:10.48550/arxiv.1706.02677 · Cited by 2,294 · cited 1×
“This is due to the ever-increasing demand of computation capability from emerging domains such as machine learning, big data and planet-scale simulations [1], [2].”
BackgroundUltra-Performance Pascal GPU and NVLink Interconnect
Foley, D., & Danskin, J. (2017) · IEEE Micro · doi:10.1109/mm.2017.37 · Cited by 224
Fuhrer, O., Chadha, T., Hoefler, T., Kwasniewski, G., Lapillonne, X., Leutwyler, D., Lüthi, D., Osuna, C., Schär, C., Schulthess, T. C., et al. (2018) · Geoscientific Model Development · doi:10.5194/gmd-11-1665-2018 · Cited by 182
Massively Distributed SGD: ImageNet/ResNet-50 Training in a FlashPreprint
Mikami, H., Suganuma, H., U‐Chupala, P., Tanaka, Y., & Kageyama, Y. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.05233 · Cited by 138
MVAPICH2-GPU: optimized GPU to GPU communication for InfiniBand clusters
Wang, H., Potluri, S., Luo, M., Singh, A. K., Sur, S., & Panda, D. K. D. (2011) · Computer Science - Research and Development · doi:10.1007/s00450-011-0171-3 · Cited by 124
Tartan: Evaluating Modern GPU Interconnect via a Multi-GPU Benchmark Suite
Li, A., Song, S. L., Chen, J., Liu, X., Tallent, N. R., & BARKER, K. J. (2018) · IEEE International Symposium on Workload Characterization · doi:10.1109/iiswc.2018.8573483 · Cited by 68
Self-citationEvaluation of Deep Learning Frameworks Over Different HPC Architectures
Shams, S., Platania, R., Lee, K., & Park, S. (2017) · IEEE International Conference on Distributed Computing Systems (ICDCS) · doi:10.1109/icdcs.2017.259 · Cited by 67
ImageNet/ResNet-50 Training in 224 Seconds.Preprint
Mikami, H., Suganuma, H., U‐Chupala, P., Tanaka, Y., & Kageyama, Y. (2018) · arXiv (Cornell University) · Cited by 58
Ben‐Nun, T., Levy, E., Barak, A., & Rubin, E. (2015) · International Conference for High Performance Computing, Networking, Storage and Analysis (SC) · doi:10.1145/2807591.2807611 · Cited by 52
Quantifying NUMA and contention effects in multi-GPU systems
Spafford, K. L., Meredith, J. S., & Vetter, J. S. (2011) · doi:10.1145/1964179.1964194 · Cited by 50
Exploring and analyzing the real impact of modern on-package memory on HPC scientific kernels
Li, A., Liu, W., Kristensen, M. R. B., Vinter, B., Wang, H., Hou, K., Márquez, A., & Song, S. L. (2017) · International Conference for High Performance Computing, Networking, Storage and Analysis (SC) · doi:10.1145/3126908.3126931 · Cited by 47
Self-citationWang, L., Ye, J., Zhao, Y., Wu, W., Li, A., Song, S. L., Xu, Z., & Kraska, T. (2018) · ACM SIGPLAN Notices · doi:10.1145/3200691.3178491 · Cited by 41
Self-citationOptimized Broadcast for Deep Learning Workloads on Dense-GPU InfiniBand Clusters
Awan, A. A., Chu, C.-H., Subramoni, H., & Panda, D. K. D. (2018) · European MPI Users' Group Meeting (EuroMPI) · doi:10.1145/3236367.3236381 · Cited by 40
Evaluating Performance Tradeoffs on the Radeon Open Compute Platform
Sun, Y., Mukherjee, S., Baruah, T., Dong, S., Gutierrez, J., Mohan, P., & Kaeli, D. R. (2018) · IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS) · doi:10.1109/ispass.2018.00034 · Cited by 32
Evaluating On-Node GPU Interconnects for Deep Learning Workloads
Tallent, N. R., Gawande, N. A., Siegel, C. M., et al. (2017) · Lecture notes in computer science · doi:10.1007/978-3-319-72971-8_1 · Cited by 30
Self-citationSuperNeurons: Dynamic GPU Memory Management for Training Deep Neural NetworksPreprint
Wang, L., Ye, J., Zhao, Y., Wu, W., Li, A., Song, S. L., Xu, Z., & Kraska, T. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1801.04380 · Cited by 30
Self-citationOptimizing Sparse Tensor Times Matrix on Multi-core and Many-Core Architectures
Li, J., Ma, Y., Yan, C. C., & Vuduc, R. W. (2016) · Irregular Applications: Architectures and Algorithms · doi:10.1109/ia3.2016.010 · Cited by 29
Self-citationOptimizing sparse tensor times matrix on multi-core and many-core architectures
Li, J., Ma, Y., Yan, C. C., & Vuduc, R. W. (2016) · Irregular Applications: Architectures and Algorithms · doi:10.5555/3018843.3018848 · Cited by 25
Self-citationLi, A., Tay, Y. C., Kumar, A. S., et al. (2015) · ACM International Symposium on High-Performance Parallel and Distributed Computing (HPDC) · doi:10.1145/2749246.2749265 · Cited by 20
Self-citationGPU triggered networking for intra-kernel communications
LeBeane, M., Hamidouche, K., Benton, B., Breternitz, M., Reinhardt, S. K., & John, L. K. (2017) · International Conference for High Performance Computing, Networking, Storage and Analysis (SC) · doi:10.1145/3126908.3126950 · Cited by 20
Optimized Broadcast for Deep Learning Workloads on Dense-GPU InfiniBand Clusters: MPI or NCCL?Preprint
Awan, A. A., Chu, C.-H., Subramoni, H., & Panda, D. K. D. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1707.09414 · Cited by 19
X: A Comprehensive Analytic Model for Parallel Machines
Li, A., Song, S. L., Brugel, E., et al. (2016) · Proceedings - IEEE International Parallel and Distributed Processing Symposium · doi:10.1109/ipdps.2016.89 · Cited by 17
Self-citation
Cited by
DFX: A Low-latency Multi-FPGA Appliance for Accelerating Transformer-based Text Generation
Hong, S., Moon, S., Kim, J., Lee, S., Kim, M., Lee, D., & Kim, J.-Y. (2022) · 2022 55th IEEE/ACM International Symposium on Microarchitecture (MICRO) · doi:10.1109/micro56248.2022.00051 · Cited by 107
Cho, S., Han, D., Kim, J., & Kim, J. (2024) · Journal of Energy Storage · doi:10.1016/j.est.2024.111897 · Cited by 17
Metadata
- Authors
- Li, A., Song, S. L., Chen, J., Li, J., Liu, X., Tallent, N. R., & Barker, K. J.
- Venue
- IEEE Transactions on Parallel and Distributed Systems
- Published
- 2020-01-01
- Volume
- 31
- Issue
- 1
- Pages
- 94-110
- ISSN
- 1045-9219 (ISSN-L)
- 1558-2183
- 2161-9883
- OpenAlex
- W2926767350
- Cited by
- 300
- References
- 23
- FWCI
- 27.95
- Type
- journal article
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.