In-Datacenter Performance Analysis of a Tensor Processing Unit
The DOI matches its Crossref record; 52 of the 54 references it cites were found in a registry or an index.
Abstract
Many architects believe that major improvements in cost-energy-performance must now come from domain-specific hardware. This paper evaluates a custom ASIC---called a Tensor Processing Unit (TPU) --- deployed in datacenters since 2015 that accelerates the inference phase of neural networks (NN). The heart of the TPU is a 65,536 8-bit MAC matrix multiply unit that offers a peak throughput of 92 TeraOps/second (TOPS) and a large (28 MiB) software-managed on-chip memory. The TPU's deterministic execution model is a better match to the 99th-percentile response-time requirement of our NN applications than are the time-varying optimizations of CPUs and GPUs that help average throughput more than guaranteed latency. The lack of such features helps explain why, despite having myriad MACs and a big memory, the TPU is relatively small and low power. We compare the TPU to a server-class Intel Haswell CPU and an Nvidia K80 GPU, which are contemporaries deployed in the same datacenters. Our workload, written in the high-level TensorFlow framework, uses production NN applications (MLPs, CNNs, and LSTMs) that represent 95% of our datacenters' NN inference demand. Despite low utilization for some applications, the TPU is on average about 15X -- 30X faster than its contemporary GPU or CPU, with TOPS/Watt about 30X -- 80X higher. Moreover, using the CPU's GDDR5 memory in the TPU would triple achieved TOPS and raise TOPS/Watt to nearly 70X the GPU and 200X the CPU.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“In the past few years, some companies and universities have presented application specific integrated circuits (ASICs) based on the complementary metal oxide semiconductor (CMOS) technology that are capable to compute and store information in the same unit.”
From Hardware implementation of memristor-based artificial neural networks (2024) · cites this paper as [19]
In the citing paper
“Additional optimization techniques, such as model compression [165] and hardware acceleration using graphics processing units (GPUs) or tensor processing units (TPUs) [166], further enhance efficiency.”
From Multimodal Large Language Models in Health Care: Applications, Challenges, and Future Outlook. (2024) · cites this paper as [166]
In the citing paper
“Tensor Processing Units are a class of proprietary AI accelerator hardware developed by Google, and used in their internal computing projects and employed in Google Cloud datacenters [17].”
From The rising costs of training frontier AI models (2024) · cites this paper as [17]
In the citing paper
“Resource-efficient computing devices that require less natural resources (energy, water, raw materials) to build and use can be seen as an obvious gain in terms of environmental impacts, and different hardware platforms are being developed towards this end as an alternative to GPUs and CPUs, which represent the current status quo of AI computing hardware [61, 62].”
From From Efficiency Gains to Rebound Effects: The Problem of Jevons' Paradox in AI's Polarized Environmental Debate (2025) · cites this paper as [62]
Citations
Cites
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
Going deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380
Mastering the game of Go with deep neural networks and tree search
Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., van den Driessche, G., Schrittwieser, J., Antonoglou, I., Panneershelvam, V., Lanctot, M., et al. (2016) · Nature · doi:10.1038/nature16961 · Cited by 15,612
Identity mappings in deep residual networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_38 · Cited by 10,741
Identity Mappings in Deep Residual Networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · arXiv · doi:10.48550/arxiv.1603.05027 · Cited by 10,665
TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
Abadi, M., Agarwal, A., Barham, P., Brevdo, E., Chen, Z., Citro, C., Corrado, G. S., Davis, A., Dean, J., Devin, M., et al. (2016) · arXiv · doi:10.48550/arxiv.1603.04467 · Cited by 9,408
Self-citationComputer Architecture: A Quantitative Approach
Hennessy, J. L., & Patterson, D. A. (1989) · UCLA - Biblioteca de Ciencias y Tecnologia Felix Morales Bueno · Cited by 9,375
Self-citationGoogle's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
Wu, Y., Schuster, M., Chen, Z., Le, Q. V., Norouzi, M., Macherey, W., Krikun, M., Cao, Y., Gao, Q., Macherey, K., et al. (2016) · arXiv · doi:10.48550/arxiv.1609.08144 · Cited by 5,319
Self-citationWilliams, S., Waterman, A., & Patterson, D. A. (2009) · Communications of the ACM · doi:10.1145/1498765.1498785 · Cited by 2,798
Self-citation
Research works printed but not linked: 10
- Asanović, K. 1998. Asanović, K., Beck, Johnson, J., Wawrzynek, J., Kingsbury, B. and Morgan, N., November 1998. Training Neural Networks with Spert-II. Chapter 11 in Parallel Architectures for Artificial Networks: Paradigms and Implementations, N. Sundararajan and P. Saratchandran (Eds.), IEEE Computer Society Press, ISBN 0-8186-8399-6. https://people.eecs.berkeley.edu/~krste/papers/annbook.pdf
- Han S.; Pool J.; Tran J.; and Dally W. 2015. Learning both weights and connections for efficient neural networks. In Advances in Neural Information Processing Systems. · doi:10.5555/2969239.2969366
- Ross, J., Phelps, A., 2015. Computing Convolutions Using a Neural Network Processor, Patent Application No. 62/164,902.
- Ross, J., 2015. Prefetching Weights for a Neural Network Processor, Patent Application No. 62/164,981.
- Ross, J., Thorson, G., 2015. Rotating Data for Neural Network Computations, Patent Application No. 62/164,908.
- Schurman, E. and Brutlag, J., 2009, June. The user and business impact of server delays, additional bytes, and HTTP chunking in web search. In Velocity Web Performance and Operations Conference.
- 10.5555/800048.801719 · doi:10.5555/800048.801719
- Steinberg, D., 2015. Full-Chip Simulations, Keys to Success. Proc. Synopsys Users Group (SNUG) Silicon Valley 2015.
- Thorson, G., Clark, C., Luu, D., 2015. Vector Computation Unit in a Neural Network Processor, Patent Application No. 62/165,022.
- Young, C., 2015. Batch Processing in a Neural Network Processor, Patent Application No. 62/165,020.
Cited by
Howard, A., Sandler, M., Chen, B., Wang, W., Chen, L.-C., Tan, M., Chu, G., Vasudevan, V., Zhu, Y., Pang, R., et al. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00140 · Cited by 9,708
“We train our models using synchronous training setup on 4x4 TPU Pod [24] using standard tensorflow RMSPropOptimizer with 0.”
MethodInfluentialSelf-citationMastering the game of Go without human knowledge
Silver, D., Schrittwieser, J., Simonyan, K., Antonoglou, I., Huang, A., Guez, A., Hubert, T., Baker, L., Lai, M., Bolton, A., et al. (2017) · Nature · doi:10.1038/nature24270 · Cited by 9,084
A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play
Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Graepel, T., et al. (2018) · Science · doi:10.1126/science.aar6404 · Cited by 3,562
Grandmaster level in StarCraft II using multi-agent reinforcement learning
Vinyals, O., Babuschkin, I., Czarnecki, W. M., Mathieu, M., Dudzik, A., Chung, J., Choi, D. H., Powell, R., Ewalds, T., Georgiev, P., et al. (2019) · Nature · doi:10.1038/s41586-019-1724-z · Cited by 3,455
On the Opportunities and Risks of Foundation Models
Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M. S., Bohg, J., Bosselut, A., Brunskill, E., et al. (2021) · arXiv · doi:10.48550/arxiv.2108.07258 · Cited by 2,676
Self-citationProtTrans: Toward Understanding the Language of Life Through Self-Supervised Learning
Elnaggar, A., Heinzinger, M., Dallago, C., Rehawi, G., Wang, Y., Jones, L., Gibbs, T., Feher, T., Angerer, C., Steinegger, M., et al. (2022) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2021.3095381 · Cited by 2,329
In-memory computing with resistive switching devices
Ielmini, D., & Wong, H.-S. P. (2018) · Nature Electronics · doi:10.1038/s41928-018-0092-2 · Cited by 2,298
“For instance, the tensor processing unit (TPU) has been recently developed for accelerating the multiply-accumulate (MAC) operation, which constitutes the major workload in the inference phase of neural networks in data centres for image and speech recognition 6 .”
BackgroundMemory devices and applications for in-memory computing
Sebastian, A., Le Gallo, M., Khaddam-Aljameh, R., & Eleftheriou, E. (2020) · Nature Nanotechnology · doi:10.1038/s41565-020-0655-z · Cited by 2,289
“The current approaches, such as the use of hundreds of processors in parallel (for example, graphics processing units 3 ) or application-specific processors 4,5 that are custom designed for specific applications, are not likely to fully overcome the challenge of data movement.”
Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields
Barron, J. T., Mildenhall, B., Verbin, D., Srinivasan, P. P., & Hedman, P. (2022) · 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52688.2022.00539 · Cited by 1,935
“For all NeRF-like models, we report train times from a TPU v2 with 32 cores [18], as well as model size (the train times and model sizes of SVS, Deep Blending, and Point-Based Neural Rendering are not presented, as this comparison would not be particularly meaningful).”
MethodMip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields
Barron, J. T., Mildenhall, B., Tancik, M., Hedman, P., Martin-Brualla, R., & Srinivasan, P. P. (2021) · 2021 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00580 · Cited by 1,921
Fully hardware-implemented memristor convolutional neural network
Yao, P., Wu, H., Gao, B., Tang, J., Zhang, Q., Zhang, W., Yang, J. J., & Qian, H. (2020) · Nature · doi:10.1038/s41586-020-1942-4 · Cited by 2,179
“Such demand has led to the redesign of conventional computing systems to operate CNNs with higher performance and lower power consumption, ranging from general application platforms, such as graphics-processing units (GPUs) 13 , to application-specific accelerators 14,15 .”
BackgroundDigital twin: values, challenges and enablers from a modeling perspective
Rasheed, A., San, O., & Kvamsdal, T. (2020) · IEEE Access · doi:10.1109/access.2020.2970143 · Cited by 1,952
“, TPUs [462]) for not only graphics rendering but also scientific computing applications.”
BackgroundDeep Learning in Mobile and Wireless Networking: A Survey
Zhang, C., Patras, P., & Haddadi, H. (2019) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2019.2904897 · Cited by 1,768
Parallel convolutional processing using an integrated photonic tensor core
Feldmann, J., Youngblood, N., Karpov, M., Gehring, H., Li, X., Stappers, M., Le Gallo, M., Fu, X., Lukashchuk, A., Raja, A. S., et al. (2021) · Nature · doi:10.1038/s41586-020-03070-1 · Cited by 1,900
“Although various custom silicon computing hardware—that is, field-programmable gate arrays (FPGAs) 10 , ASICs 11 and graphics processing units (GPUs) 12 —have been developed to improve computational throughput and efficiency, they still depend on the same underlying electronic components, which are fundamentally limit…”
BackgroundMemristive crossbar arrays for brain-inspired computing
Xia, Q., & Yang, J. J. (2019) · Nature Materials · doi:10.1038/s41563-019-0291-x · Cited by 1,800
Photonics for artificial intelligence and neuromorphic computing
Shastri, B. J., Tait, A. N., Ferreira de Lima, T., Pernice, W. H. P., Bhaskaran, H., Wright, C. D., & Prucnal, P. R. (2021) · Nature Photonics · doi:10.1038/s41566-020-00754-y · Cited by 1,756
“2 Machine learning acceleration: vector-matrix multiplications [14], deep learning inference [15], ultrafast or online learning [16]”
BackgroundA State-of-the-Art Survey on Deep Learning Theory and Architectures
Alom, M. Z., Taha, T. M., Yakopcic, C., Westberg, S., Sidike, P., Nasrin, M. S., Hasan, M., Van Essen, B. C., Awwal, A. A. S., & Asari, V. K. (2019) · Electronics · doi:10.3390/electronics8030292 · Cited by 1,620
Convergence of Edge Computing and Deep Learning: A Comprehensive Survey
Wang, X., Han, Y., Leung, V. C. M., Niyato, D., Yan, X., & Chen, X. (2020) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2020.2970550 · Cited by 1,498
Deep Learning for IoT Big Data and Streaming Analytics: A Survey
Mohammadi, M., Al-Fuqaha, A., Sorour, S., & Guizani, M. (2018) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2018.2844341 · Cited by 1,471
A Survey on Federated Learning Systems: Vision, Hype and Reality for Data Privacy and Protection
Li, Q., Wen, Z., Wu, Z., Hu, S., Wang, N., Li, Y., Liu, X., & He, B. (2023) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2021.3124599 · Cited by 1,407
DeepXplore: Automated Whitebox Testing of Deep Learning Systems
Pei, K., Cao, Y., Yang, J., & Jana, S. (2017) · Proceedings of the 26th Symposium on Operating Systems Principles · doi:10.1145/3132747.3132785 · Cited by 1,217
“In recent years, DNNs have surpassed human performance in many application domains due to increasing availability of large datasets [20, 38, 47], specialized hardware [34, 50], and efficient training algorithms [31, 39, 66, 71].”
BackgroundDeep learning for cellular image analysis
Moen, E., Bannon, D., Kudo, T., Graf, W., Covert, M., & Van Valen, D. (2019) · Nature Methods · doi:10.1038/s41592-019-0403-1 · Cited by 1,225
“Third, they provide an easy gateway for specialized hardware such as graphical processing units (GPUs) and tensor processing units 65–67 .”
BackgroundEyeriss v2: A Flexible Accelerator for Emerging Deep Neural Networks on Mobile Devices
Chen, Y.-H., Yang, T.-J., Emer, J. S., & Sze, V. (2019) · IEEE Journal on Emerging and Selected Topics in Circuits and Systems · doi:10.1109/jetcas.2019.2910232 · Cited by 1,106
“Two common DNN accelerator designs: (a) Spatial accumulation array [29]–[32]: iacts are reused vertically and psums are accumulated horizontally.”
BackgroundMethodNeuro-Inspired Computing With Emerging Nonvolatile Memorys
Yu, S. (2018) · Proceedings of the IEEE · doi:10.1109/jproc.2018.2790840 · Cited by 1,170
TVM: an automated end-to-end optimizing compiler for deep learning
(2018) · arXiv (Cornell University) · doi:10.5555/3291168.3291211 · Cited by 1,170
“ch can typically be decomposed into tensor operators like matrix-matrix multiplication or 1D convolution. These natural decompositions have led to the recent trend of adding tensor compute primitives [1, 12, 21]. These new primitives create both opportunities and challenges for schedule-based compilation; while using…”
BackgroundMethodInfluentialEquivalent-accuracy accelerated neural-network training using analogue memory
Ambrogio, S., Narayanan, P., Tsai, H., Shelby, R. M., Boybat, I., di Nolfo, C., Sidler, S., Giordano, M., Bodini, M., Farinha, N. C. P., et al. (2018) · Nature · doi:10.1038/s41586-018-0180-5 · Cited by 1,151
Unified rational protein engineering with sequence-based deep representation learning
Alley, E. C., Khimulya, G., Biswas, S., AlQuraishi, M., & Church, G. M. (2019) · Nature Methods · doi:10.1038/s41592-019-0598-1 · Cited by 1,131
Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm
Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Graepel, T., et al. (2017) · arXiv · doi:10.48550/arxiv.1712.01815 · Cited by 1,102
“Training proceeded for 700,000 steps (mini-batches of size 4,096) starting from randomly initialised parameters, using 5,000 first-generation TPUs (16) to generate self-play games and 64 second-generation TPUs to train the neural networks; training lasted for approximately 9 hours in chess, 12 hours in shogi and 34 ho…”
MethodInfluentialTowards artificial general intelligence with hybrid Tianjic chip architecture
Pei, J., Deng, L., Song, S., Zhao, M., Zhang, Y., Wu, S., Wang, G., Zou, Z., Wu, Z., He, W., et al. (2019) · Nature · doi:10.1038/s41586-019-1424-8 · Cited by 1,082
Pruning and quantization for deep neural network acceleration: A survey
Liang, T., Glossner, J., Wang, L., Shi, S., & Zhang, X. (2021) · Neurocomputing · doi:10.1016/j.neucom.2021.07.045 · Cited by 943
United States Data Center Energy Usage Report
Lawrence Berkeley National Laboratory (LBNL), B., Shehabi, A., USDOE, Smith, S., Sartor, D., Brown, R., Herrlin, M., Koomey, J., Masanet, E., Horner, N., et al. (2016) · doi:10.2172/1372902 · Cited by 1,040
Big Transfer (BiT): General Visual Representation Learning
Kolesnikov, A., Beyer, L., Zhai, X., Puigcerver, J., Yung, J., Gelly, S., & Houlsby, N. (2020) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-58558-7_29 · Cited by 1,039
“ochs with an initial learning rate of 0.03, with 5000 linear warmup steps and decay the learning rate after 10, 23, 30 and 37 epochs. We use a global batch size of 4096 and train on a Cloud TPUv3-512 [22], resulting in 8 images per chip. For optimization with large batch sizes we employ recipes from [9]. In particular…”
MethodInfluentialBig Transfer (BiT): General Visual Representation Learning
Kolesnikov, A., Beyer, L., Zhai, X., Puigcerver, J., Yung, J., Gelly, S., & Houlsby, N. (2019) · arXiv · doi:10.48550/arxiv.1912.11370 · Cited by 1,039
“ochs with an initial learning rate of 0.03, with 5000 linear warmup steps and decay the learning rate after 10, 23, 30 and 37 epochs. We use a global batch size of 4096 and train on a Cloud TPUv3-512 [22], resulting in 8 images per chip. For optimization with large batch sizes we employ recipes from [9]. In particular…”
MethodInfluentialEvolution Strategies as a Scalable Alternative to Reinforcement Learning
Salimans, T., Ho, J., Chen, X., Sidor, S., & Sutskever, I. (2017) · arXiv · doi:10.48550/arxiv.1703.03864 · Cited by 1,027
“Similarly, specialized hardware for neural network inference, such as TPUs [Jouppi et al., 2017], can be used directly when performing optimization using ES, while their limited memory usually makes backpropagation impossible.”
MethodMachine Learning and Deep Learning frameworks and libraries for large-scale data mining: a survey
Nguyen, G., Dlugolinsky, S., Bobák, M., Tran, V., López García, Á., Heredia, I., Malík, P., & Hluchý, L. (2019) · Artificial Intelligence Review · doi:10.1007/s10462-018-09679-z · Cited by 864
“Recently, there is a rise of new types of accelerators such as cryptographic and AI accelerators [like Google’s TPUs (Jouppi et al. 2017) or Intel’s Neural Processors (Kloss 2017)], which belong to new hardware architecture generation optimized for ML, DL and other AI workloads (Kobielus 2018; Tang 2018).”
BackgroundModel Compression and Hardware Acceleration for Neural Networks: A Comprehensive Survey
Deng, B. L., Li, G., Han, S., Shi, L., & Xie, Y. (2020) · Proceedings of the IEEE · doi:10.1109/jproc.2020.2976475 · Cited by 996
“Conventional DNN accelerators solely optimize the hardware architecture to improve the compute parallelism and reduce memory accesses [8], [9].”
BackgroundMethodInfluentialStand-Alone Self-Attention in Vision Models
Ramachandran, P., Parmar, N., Vaswani, A., Bello, I., Levskaya, A., & Shlens, J. (2019) · arXiv · doi:10.48550/arxiv.1906.05909 · Cited by 977
A Survey on Distributed Machine Learning
Verbraeken, J., Wolting, M., Katzy, J., Kloppenburg, J., Verbelen, T., & Rellermeyer, J. S. (2021) · ACM Computing Surveys · doi:10.1145/3377454 · Cited by 872
“When running benchmarks, Jouppi et al. [80] found that the performance per watt of a TPU can approach 200x that of a traditional system.”
BackgroundA compute-in-memory chip based on resistive random-access memory
Wan, W., Kubendran, R., Schaefer, C., Eryilmaz, S. B., Zhang, W., Wu, D., Deiss, S., Raina, P., Qian, H., Gao, B., et al. (2022) · Nature · doi:10.1038/s41586-022-04992-8 · Cited by 943
Efficient and self-adaptive in-situ learning in multilayer memristor neural networks
Li, C., Belkin, D., Li, Y., Yan, P., Hu, M., Ge, N., Jiang, H., Montgomery, E., Lin, P., Wang, Z., et al. (2018) · Nature Communications · doi:10.1038/s41467-018-04484-2 · Cited by 931
Deep Learning for Audio Signal Processing
Purwins, H., Li, B., Virtanen, T., Schluter, J., Chang, S.-Y., & Sainath, T. (2019) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2019.2908700 · Cited by 855
“Instead, processors optimized for matrix operations are commonly used, mostly general-purpose graphics processing units (GPGPUs) [150] and application-specific integrated circuits such as the proprietary tensor processing units (TPUs) [8].”
MethodSchneider, G. (2018) · Nature Reviews Drug Discovery · doi:10.1038/nrd.2017.232 · Cited by 910
Federated Learning for Internet of Things: Recent Advances, Taxonomy, and Open Challenges
Khan, L. U., Saad, W., Han, Z., Hossain, E., & Hong, C. S. (2021) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2021.3090430 · Cited by 892
Dive into Deep LearningPreprint
Zhang, A., Lipton, Z. C., Mu, L., & Smola, A. J. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.11342 · Cited by 999
Demystifying Parallel and Distributed Deep Learning: An In-depth Concurrency Analysis
Ben-Nun, T., & Hoefler, T. (2020) · ACM Computing Surveys · doi:10.1145/3320060 · Cited by 577
“We see this trend with the NVIDIA Tensor Cores [188], the Tensor Processing Unit [129], other ASICs and FPGAs [35, 187], and even neuromorphic computing [4].”
MethodBourtoule, L., Chandrasekaran, V., Choquette-Choo, C. A., Jia, H., Travers, A., Zhang, B., Lie, D., & Papernot, N. (2021) · 2021 IEEE Symposium on Security and Privacy (SP) · doi:10.1109/sp40001.2021.00019 · Cited by 810
“We chose to study neural networks because they almost always generate the largest computational costs and require investments in dedicated accelerators [20], [21]; we also discuss other ML techniques in § IV.”
MethodA new golden age for computer architecture
Hennessy, J. L., & Patterson, D. A. (2019) · Communications of the ACM · doi:10.1145/3282307 · Cited by 690
“As an example DSA, consider the Google TPU v1, which was designed to accelerate neural net inference.(17,18) The TPU has been in production since 2015 and powers applications ranging from search queries to language translation to image recognition to AlphaGo and AlphaZero, the DeepMind programs for playing Go and Ches…”
MethodSelf-citationAnsel, J., Yang, E., He, H., Gimelshein, N., Jain, A., Voznesensky, M., Bao, B., Bell, P., Berard, D., Burovski, E., et al. (2024) · Proceedings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2 · doi:10.1145/3620665.3640366 · Cited by 737
Efficient Deep Learning: A Survey on Making Deep Learning Models Smaller, Faster, and Better
Menghani, G. (2023) · ACM Computing Surveys · doi:10.1145/3578938 · Cited by 605
“The first iteration of the TPU was designed for inference with 8-bit integers, and was being used in Google for a year prior to their announcement in 2016 [86].”
MethodSimVLM: Simple Visual Language Model Pretraining with Weak Supervision
Wang, Z., Yu, J., Yu, A. W., Dai, Z., Tsvetkov, Y., & Cao, Y. J. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2108.10904 · Cited by 327
RegNeRF: Regularizing Neural Radiance Fields for View Synthesis from Sparse Inputs
Niemeyer, M., Barron, J. T., Mildenhall, B., Sajjadi, M. S. M., Geiger, A., & Radwan, N. (2022) · 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52688.2022.00540 · Cited by 609
Perceiver IO: A General Architecture for Structured Inputs & Outputs
Jaegle, A., Borgeaud, S., Alayrac, J.-B., Doersch, C., Ionescu, C., Ding, X. D., Koppula, S., Zoran, D., Brock, A., Shelhamer, E., et al. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.14795 · Cited by 198
“We use a larger batch size of 8192 and train on 256 TPUs.”
MethodInfluentialLearning Data Augmentation Strategies for Object Detection
Zoph, B., Cubuk, E. D., Ghiasi, G., Lin, T.-Y., Shlens, J., & Le, Q. V. (2020) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-58583-9_34 · Cited by 570
Gan, Y., Zhang, Y., Cheng, D., Shetty, A., Rathi, P., Katarki, N., Bruno, A., Hu, J., Ritchken, B., Jackson, B., et al. (2019) · Proceedings of the Twenty-Fourth International Conference on Architectural Support for Programming Languages and Operating Systems · doi:10.1145/3297858.3304013 · Cited by 665
“This includes analyzing the cycle breakdown in modern servers, examining whether big or small cores are preferable [25, 35, 41, 42, 46– 48], determining the pressure microservices put on instruction caches [37, 52], and exploring the potential they have for hardware acceleration [24, 27, 38, 49, 71].”
BackgroundSelf-citationTimeloop: A Systematic Approach to DNN Accelerator Evaluation
Parashar, A., Raina, P., Shao, Y. S., Chen, Y.-H., Ying, V. A., Mukkara, A., Venkatesan, R., Khailany, B., Keckler, S. W., & Emer, J. (2019) · 2019 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS) · doi:10.1109/ispass.2019.00042 · Cited by 571
“To provide high performance and energy efficiency to this class of compute-intensive algorithms, many DNN hardware accelerators have been proposed [1], [2], [5], [6], [12], [13], [15], [18], [25], [29], [33], [35], [39].”
BackgroundDeep neural networks for the evaluation and design of photonic devices
Jiang, J., Chen, M., & Fan, J. A. (2020) · Nature Reviews Materials · doi:10.1038/s41578-020-00260-1 · Cited by 551
Dillon, J. V., Langmore, I., Tran, D., Brevdo, E., Vasudevan, S., Moore, D., Patton, B., Alemi, A., Hoffman, M., & Saurous, R. A. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1711.10604 · Cited by 226
“…the ecosystem, TensorFlow Distributions inherits and maintains integration with TensorFlow graph operations, automatic differentiation, idiomatic batching and vectorization, device-specific kernel optimizations and XLA, and accelerator support for CPUs, GPUs, and tensor processing units (TPUs) [25].”
BackgroundMethodInfluentialThe future of computing beyond Moore’s Law
Shalf, J. (2020) · Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences · doi:10.1098/rsta.2019.0061 · Cited by 526
“Examples of this trend exist in smartphone technologies, which contain dozens of specialized accelerators co-located on the same chip; in hardware deployed in massive data centres, such as Google’s Tensor Processing Unit (TPU), which accelerates the Tensorflow programming framework for ML tasks; in field-programmable ga…”
BackgroundInfluentialEstimation of energy consumption in machine learning
García-Martín, E., Rodrigues, C. F., Riley, G., & Grahn, H. (2019) · Journal of Parallel and Distributed Computing · doi:10.1016/j.jpdc.2019.07.007 · Cited by 558
Photonic matrix multiplication lights up photonic accelerator and beyond
Zhou, H., Dong, J., Cheng, J., Dong, W., Huang, C., Shen, Y., Zhang, Q., Gu, M., Qian, C., Chen, H., et al. (2022) · Light: Science & Applications · doi:10.1038/s41377-022-00717-8 · Cited by 563
Streaming End-to-end Speech Recognition for Mobile Devices
He, Y., Sainath, T. N., Prabhavalkar, R., McGraw, I., Alvarez, R., Zhao, D., Rybach, D., Kannan, A., Wu, Y., Pang, R., et al. (2019) · ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2019.8682336 · Cited by 601
Chang, J., Sitzmann, V., Dun, X., Heidrich, W., & Wetzstein, G. (2018) · Scientific Reports · doi:10.1038/s41598-018-30619-y · Cited by 564
“ − | | Φ ΔΦ e minimize PSF { } , (7) i opt 2 F 2”
Vowel recognition with four coupled spin-torque nano-oscillators
Romera, M., Talatchian, P., Tsunegi, S., Abreu Araujo, F., Cros, V., Bortolotti, P., Trastoy, J., Yakushiji, K., Fukushima, A., Kubota, H., et al. (2018) · Nature · doi:10.1038/s41586-018-0632-y · Cited by 572
CondConv: Conditionally Parameterized Convolutions for Efficient Inference
Yang, B., Bender, G., Le, Q. V., & Ngiam, J. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1904.04971 · Cited by 208
“Deploying the best-performing models for inference can consume significant datacenter capacity [19] and are often not feasible for applications with strict latency constraints.”
Aligning artificial intelligence with climate change mitigation
Kaack, L. H., Donti, P. L., Strubell, E., Kamiya, G., Creutzig, F., & Rolnick, D. (2022) · Nature Climate Change · doi:10.1038/s41558-022-01377-7 · Cited by 570
Hardware implementation of memristor-based artificial neural networks
Aguirre, F., Sebastian, A., Le Gallo, M., Song, W., Wang, T., Yang, J. J., Lu, W., Chang, M.-F., Ielmini, D., Yang, Y., et al. (2024) · Nature Communications · doi:10.1038/s41467-024-45670-9 · Cited by 564
“Acomprehensivelist of these ASICs comprising those such as the Google TPU 19 , Amazon inferentia 20 , Tesla NPU 21 , etc., are summarized in ref.”
A Configurable Cloud-Scale DNN Processor for Real-Time AI
Fowers, J., Ovtcharov, K., Papamichael, M., Massengill, T., Liu, M., Lo, D., Alkalay, S., Haselman, M., Adams, L., Ghandi, M., et al. (2018) · 2018 ACM/IEEE 45th Annual International Symposium on Computer Architecture (ISCA) · doi:10.1109/isca.2018.00012 · Cited by 520
“The TPU1, a commercially deployed cloud NPU [10] most closely aligns to L3 using a systolic two-dimensional matrix-matrix multiplication array.”
BackgroundMethodOpportunities and Challenges for Machine Learning in Materials Science
Morgan, D., & Jacobs, R. (2020) · Annual Review of Materials Research · doi:10.1146/annurev-matsci-070218-010015 · Cited by 549
Bit Fusion: Bit-Level Dynamically Composable Architecture for Accelerating Deep Neural Network
Sharma, H., Park, J., Suda, N., Lai, L., Chau, B., Kim, J. K., Chandra, V., & Esmaeilzadeh, H. (2018) · 2018 ACM/IEEE 45th Annual International Symposium on Computer Architecture (ISCA) · doi:10.1109/isca.2018.00069 · Cited by 543
There’s plenty of room at the Top: What will drive computer performance after Moore’s law?
Leiserson, C. E., Thompson, N. C., Emer, J. S., Kuszmaul, B. C., Lampson, B. W., Sanchez, D., & Schardl, T. B. (2020) · Science · doi:10.1126/science.aam9744 · Cited by 469
“Google has developed a tensor-processing unit (TPU) ( 52 ) specifically designed for deep learning, embracing special-purpose processing and eschewing the broader functionality of GPUs.”
BackgroundFINN- R: An End-to-End Deep-Learning Framework for Fast Exploration of Quantized Neural Networks
Blott, M., Preußer, T. B., Fraser, N. J., Gambardella, G., O’brien, K., Umuroglu, Y., Leeser, M., & Vissers, K. (2018) · ACM Transactions on Reconfigurable Technology and Systems · doi:10.1145/3242897 · Cited by 446
“However, a modest direct quantization as to 8-bit fixedpoint quantities has little or no impact on accuracy [27] but already leads to significant resource savings.”
BackgroundInfluentialSIGMA: A Sparse and Irregular GEMM Accelerator with Flexible Interconnects for DNN Training
Qin, E., Samajdar, A., Kwon, H., Nadella, V., Srinivasan, S., Das, D., Kaul, B., & Krishna, T. (2020) · 2020 IEEE International Symposium on High Performance Computer Architecture (HPCA) · doi:10.1109/hpca47549.2020.00015 · Cited by 487
“An alternate term for this dataflow is weight-stationary , and is used by the Google TPU [23], [37].”
BackgroundMethodMetNet: A Neural Weather Model for Precipitation Forecasting
Sønderby, C. K., Espeholt, L., Heek, J., Dehghani, M., Oliver, A., Salimans, T., Agrawal, S., Hickey, J. J., & Kalchbrenner, N. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2003.12140 · Cited by 291
“In the field of machine learning deep neural networks (DNN) have seen remarkable progress in recent years due to increased amounts of available data, better model architectures and ease of implementation on powerful specialized hardware such as GPUs and TPUs (Jouppi et al., 2017).”
BackgroundAccurate deep neural network inference using computational phase-change memory
Joshi, V., Le Gallo, M., Haefeli, S., Boybat, I., Nandakumar, S. R., Piveteau, C., Dazzi, M., Rajendran, B., Sebastian, A., & Eleftheriou, E. (2020) · Nature Communications · doi:10.1038/s41467-020-16108-9 · Cited by 490
“Significant progress in this direction has been made with the introduction of specialized hardware for inference operating at reduced digital precision (4 to 8-bit), such as Google’s tensor processing unit (TPU) 1 and low-power graphical processing units (GPUs) such as NVIDIA T4 2 .”
MethodThe Best of Both Worlds: Combining Recent Advances in Neural Machine Translation
Chen, M. X., Firat, O., Bapna, A., Johnson, M., Macherey, W., Foster, G., Jones, L., Schuster, M., Shazeer, N., Parmar, N., et al. (2018) · Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/p18-1008 · Cited by 418
“such as GPUs and Tensor Processing Units (TPUs) (Jouppi et al., 2017).”
BackgroundMethodMachine Learning at Facebook: Understanding Inference at the Edge
Wu, C.-J., Brooks, D., Chen, K., Chen, D., Choudhury, S., Dukhan, M., Hazelwood, K., Isaac, E., Jia, Y., Jia, B., et al. (2019) · 2019 IEEE International Symposium on High Performance Computer Architecture (HPCA) · doi:10.1109/hpca.2019.00048 · Cited by 439
“Recent work on hardware for machine learning and efficient training and inference has substantially advanced the state-of-the-art [41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66].”
BackgroundIn-Memory Computing: Advances and Prospects
Verma, N., Jia, H., Valavi, H., Tang, Y., Ozatay, M., Chen, L.-Y., Zhang, B., & Deaville, P. (2019) · IEEE Solid-State Circuits Magazine · doi:10.1109/mssc.2019.2922889 · Cited by 473
“Here, processing engines (PEs) are arranged in a 2D array, matching the dataflow of MVM [5], [6].”
MethodTensor Comprehensions: Framework-Agnostic High-Performance Machine Learning Abstractions
Vasilache, N., Zinenko, O., Theodoridis, T., Goyal, P., DeVito, Z., Moses, W. S., Verdoolaege, S., Adams, A., & Cohen, A. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1802.04730 · Cited by 238
“improve user experience and revenue, based on individual preferences and history—the landscape is heavily shifting to custom hardware where the most serious efficiency gains lie, both in data centers [41, 65] and on mobile devices. Still, these models need to be trained before being deployed. Training usually involves…”
BackgroundFraga-Lamas, P., Lopes, S. I., & Fernández-Caramés, T. M. (2021) · Sensors · doi:10.3390/s21175745 · Cited by 405
“In [124], Jouppi et al. compare the performance of several processors used by Google cloud-based systems on inference tasks when running various types of NNs.”
BackgroundScaling for edge inference of deep neural networks
Xu, X., Ding, Y., Hu, S. X., Niemier, M., Cong, J., Hu, Y., & Shi, Y. (2018) · Nature Electronics · doi:10.1038/s41928-018-0059-3 · Cited by 462
Recent advances in convolutional neural network acceleration
Zhang, Q., Zhang, M., Chen, T., Sun, Z., Ma, Y., & Yu, B. (2019) · Neurocomputing · doi:10.1016/j.neucom.2018.09.038 · Cited by 434
Edge learning using a fully integrated neuro-inspired memristor chip
Zhang, W., Yao, P., Gao, B., Liu, Q., Wu, D., Zhang, Q., Li, Y., Qin, Q., Li, J., Zhu, Z., et al. (2023) · Science · doi:10.1126/science.ade3483 · Cited by 437
FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
Dao, T., Fu, D. Y., Ermon, S., Rudra, A., & Ré, C. (2022) · arXiv · doi:10.48550/arxiv.2205.14135 · Cited by 1,214
“[46] Norman P Jouppi, Cliff Young, Nishant Patil, David Patterson, Gaurav Agrawal, Raminder Bajwa, Sarah Bates, Suresh Bhatia, Nan Boden, Al Borchers, et al.”
BackgroundSimba: Scaling Deep-Learning Inference with Multi-Chip-Module-Based Architecture
Shao, Y. S., Clemons, J., Venkatesan, R., Zimmer, B., Fojtik, M., Jiang, N., Keller, B., Klinefelter, A., Pinckney, N., Raina, P., et al. (2019) · Proceedings of the 52nd Annual IEEE/ACM International Symposium on Microarchitecture · doi:10.1145/3352460.3358302 · Cited by 406
“For example, the first generation of the Tensor Processing Unit (TPU) delivers 92 TOPS [39] and is designed for inference applications.”
BackgroundMethodInfluentialEfficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review
Shuvo, M. M. H., Islam, S. K., Cheng, J., & Morshed, B. I. (2023) · Proceedings of the IEEE · doi:10.1109/jproc.2022.3226481 · Cited by 436
“Tianjic [324] and Simba [325] are two spatial architectures that utilized the concept of distributed weights [326] and weight exchange among PEs [319].”
BackgroundResultProtTrans: Towards Cracking the Language of Life’s Code Through Self-Supervised Learning
Elnaggar, A., Heinzinger, M., Dallago, C., Rehawi, G., Wang, Y., Jones, L., Gibbs, T., Feher, T., Angerer, C., Steinegger, M., et al. (2020) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2020.07.12.199554 · Cited by 354
SpAtten: Efficient Sparse Attention Architecture with Cascade Token and Head Pruning
Wang, H., Zhang, Z., & Han, S. (2021) · 2021 IEEE International Symposium on High-Performance Computer Architecture (HPCA) · doi:10.1109/hpca51647.2021.00018 · Cited by 427
PUMA: A Programmable Ultra-efficient Memristor-based Accelerator for Machine Learning Inference
Ankit, A., Hajj, I. E., Chalamalasetti, S. R., Ndu, G., Foltin, M., Williams, R. S., Faraboschi, P., Hwu, W.-M. W., Strachan, J. P., Roy, K., et al. (2019) · Proceedings of the Twenty-Fourth International Conference on Architectural Support for Programming Languages and Operating Systems · doi:10.1145/3297858.3304049 · Cited by 424
A Manifesto for Future Generation Cloud Computing: Research Directions for the Next Decade
Buyya, R., Srirama, S. N., Casale, G., Calheiros, R., Simmhan, Y., Varghese, B., Gelenbe, E., Javadi, B., Vaquero, L. M., Netto, M. A. S., et al. (2019) · ACM Computing Surveys · doi:10.1145/3241737 · Cited by 362
“, Amazon is offering GPUs, Google has been deploying Tensor Processing Units (TPUs) [99] and Microsoft is deploying FPGAs in the Azure Cloud [129].”
MethodLarge-Scale Optical Neural Networks Based on Photoelectric Multiplication
Hamerly, R., Bernstein, L., Sludds, A., Soljačić, M., & Englund, D. (2019) · Physical Review X · doi:10.1103/physrevx.9.021032 · Cited by 388
“However, the trend in deep learning has been toward lower-precision arithmetic, which is faster and more energy-efficient [17].”
BackgroundMethodInfluentialCompute-in-Memory Chips for Deep Learning: Recent Trends and Prospects
Yu, S., Jiang, H., Huang, S., Peng, X., & Lu, A. (2021) · IEEE Circuits and Systems Magazine · doi:10.1109/mcas.2021.3092533 · Cited by 391
Learned Step Size Quantization
Esser, S. K., McKinstry, J. L., Bablani, D., Appuswamy, R., & Modha, D. S. (2019) · arXiv · doi:10.48550/arxiv.1902.08153 · Cited by 415
“The advantages of using such algorithms to create networks for low precision hardware has been demonstrated in several deployed systems (Esser et al., 2016; Jouppi et al., 2017; Qiu et al., 2016).”
BackgroundTen Lessons From Three Generations Shaped Google’s TPUv4i : Industrial Product
Jouppi, N. P., Hyun Yoon, D., Ashcraft, M., Gottscho, M., Jablin, T. B., Kurian, G., Laudon, J., Li, S., Ma, P., Ma, X., et al. (2021) · 2021 ACM/IEEE 48th Annual International Symposium on Computer Architecture (ISCA) · doi:10.1109/isca52012.2021.00010 · Cited by 360
“Commercial Domain Specific Architectures (DSAs) for Deep Neural Networks (DNNs) are well established [21].”
InfluentialSelf-citationArtificial Neural Networks for Neuroscientists: A Primer
Yang, G. R., & Wang, X.-J. (2020) · Neuron · doi:10.1016/j.neuron.2020.09.005 · Cited by 376
GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding
Lepikhin, D., Lee, H., Xu, Y., Chen, D., Firat, O., Huang, Y., Krikun, M., Shazeer, N., & Chen, Z. (2020) · arXiv · doi:10.48550/arxiv.2006.16668 · Cited by 412
“More recently, the industry also invested heavily in building more dedicated hardware systems chasing for more cost-effective neural network hardware (Jouppi et al., 2017).”
BackgroundSustainable AI: Environmental Implications, Challenges and Opportunities
Wu, C.-J., Raghavendra, R., Gupta, U., Acun, B., Ardalani, N., Maeng, K., Chang, G., Behram, F. A., Huang, J., Bai, C., et al. (2021) · arXiv · doi:10.48550/arxiv.2111.00364 · Cited by 405
“Existing practices such as the move to domain-specific architectures at cloud scale [68], [69], [70] reduce AI computing’s footprint by consolidating computing resources at scale and by operating the shared infrastructures more environmentally-friendly with carbon free energy6.”
BackgroundLow-bit Quantization of Neural Networks for Efficient Inference
Choukroun, Y., Kravchik, E., Yang, F., & Kisilev, P. (2019) · 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW) · doi:10.1109/iccvw.2019.00363 · Cited by 404
“Such procedures cannot be efficiently implemented on deep learning accelerators with dedicated matrix multiplication units such as in systolic arrays [21].”
BackgroundBenchmarking TPU, GPU, and CPU Platforms for Deep Learning
Wang, Y. E., Wei, G.-Y., & Brooks, D. M. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1907.10701 · Cited by 221
“In this section, we use the roofline model [56] to analyze the computation and memory bandwidth of FCs and CNNs. Roofline models are useful to demonstrate memory and computation bottlenecks [56, 32].”
BackgroundMethodInfluential(2018) · doi:10.1109/ipdpsw.2018.00091 · Cited by 399
“Google designed and developed Tensor Processing Unit (TPU) specifically for inference workloads.”
MethodPoint-Voxel CNN for Efficient 3D Deep Learning
Liu, Z., Tang, H., Lin, Y., & Han, S. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1907.03739 · Cited by 323
“dth (GB/s) (a) Off-chip DRAM accesses take two orders of magnitude more energy than arithmetic operations (640pJ vs. 3pJ [11]), while the bandwidth is two orders of magnitude less (30GB/s vs. 668GB/s [17]). Efficient 3D deep learning should reduce the memory footprint, which is the bottleneck of conventional voxel-base…”
BackgroundSelf-citation
Metadata
- Authors
- Jouppi, N. P., Young, C., Patil, N., Patterson, D., Agrawal, G., Bajwa, R., Bates, S., Bhatia, S., Boden, N., Borchers, A., Boyle, R., Cantin, P.-L., Chao, C., Clark, C., Coriell, J., Daley, M., Dau, M., Dean, J., Gelb, B., Ghaemmaghami, T. V., Gottipati, R., Gulland, W., Hagmann, R., Ho, C. R., Hogberg, D., Hu, J., Hundt, R., Hurt, D., Ibarz, J., Jaffey, A., Jaworski, A., Kaplan, A., Khaitan, H., Killebrew, D., Koch, A., Kumar, N., Lacy, S., Laudon, J., Law, J., Le, D., Leary, C., Liu, Z., Lucke, K., Lundin, A., MacKean, G., Maggiore, A., Mahony, M., Miller, K., Nagarajan, R., Narayanaswami, R., Ni, R., Nix, K., Norrie, T., Omernick, M., Penukonda, N., Phelps, A., Ross, J., Ross, M., Salek, A., Samadiani, E., Severn, C., Sizikov, G., Snelham, M., Souter, J., Steinberg, D., Swing, A., Tan, M., Thorson, G., Tian, B., Toma, H., Tuttle, E., Vasudevan, V., Walter, R., Wang, W., Wilcox, E., & Yoon, D. H.
- Affiliations
- Google (United States) (US)
- ORCID
- Venue
- Proceedings of the 44th Annual International Symposium on Computer Architecture
- Publisher
- ACM
- Published
- 2017-06-24
- Pages
- 1-12
- ISSN
- 1063-6897 (ISSN-L)
- OpenAlex
- W2606722458
- Cited by
- 4,537
- References
- 54
- Citations by year
- 2026: 231 · 2025: 429 · 2024: 409 · 2023: 498 · 2022: 528 · 2021: 741 · 2020: 641 · 2019: 582 · 2018: 396 · 2017: 77
- FWCI
- 329.20
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- Proceedings - International Symposium on Computer Architecture/Proceedings (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Hardware and Architecture › Parallel Computing and Optimization Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Low-power high-performance VLSI design
- Keywords
- tensor processing units, neural network inference acceleration, domain-specific hardware, energy efficiency, deterministic execution, on-chip memory, CPU vs GPU comparison, TensorFlow
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.