An official website of the United States government
The AI boom is bringing a plethora of domain-specific architectures for Neural Network computations. Google’s Tensor Processing Unit (TPU), a Deep Neural Network (DNN) accelerator, has replaced the CPUs/GPUs in its data centers, claiming more than 15X rate of inference. However, the unprecedented growth in DNN workloads with the widespread use of AI services projects an increasing energy consumption of TPU based data centers. In this work, we parametrize the extreme hardware underutilization in TPU systolic array and propose UPTPU: an intelligent, dataflow adaptive power-gating paradigm to provide a staggering 3.5X - 6.5X energy efficiency to TPU for different input batch sizes.
more »
« less
Implementing a Timing Error-Resilient and Energy-Efficient Near-Threshold Hardware Accelerator for Deep Neural Network Inference
Increasing processing requirements in the Artificial Intelligence (AI) realm has led to the emergence of domain-specific architectures for Deep Neural Network (DNN) applications. Tensor Processing Unit (TPU), a DNN accelerator by Google, has emerged as a front runner outclassing its contemporaries, CPUs and GPUs, in performance by 15×–30×. TPUs have been deployed in Google data centers to cater to the performance demands. However, a TPU’s performance enhancement is accompanied by a mammoth power consumption. In the pursuit of lowering the energy utilization, this paper proposes PREDITOR—a low-power TPU operating in the Near-Threshold Computing (NTC) realm. PREDITOR uses mathematical analysis to mitigate the undetectable timing errors by boosting the voltage of the selective multiplier-and-accumulator units at specific intervals to enhance the performance of the NTC TPU, thereby ensuring a high inference accuracy at low voltage. PREDITOR offers up to 3×–5× improved performance in comparison to the leading-edge error mitigation schemes with a minor loss in accuracy.
more »
« less
- Award ID(s):
- 2106237
- PAR ID:
- 10347902
- Date Published:
- Journal Name:
- Journal of Low Power Electronics and Applications
- Volume:
- 12
- Issue:
- 2
- ISSN:
- 2079-9268
- Page Range / eLocation ID:
- 32
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
In this work, we propose a novel approach for the real-time estimation of chip-level spatial power maps for commercial Google Coral M.2 TPU chips based on a machine-learning technique for the first time. The new method can enable the development of more robust runtime power and thermal control schemes to take advantage of spatial power information such as hot spots that are otherwise not available. Different from the existing commercial multi-core processors in which real-time performance-related utilization information is available, the TPU from Google does not have such information. To mitigate this problem, we propose to use features that are related to the workloads of running different deep neural networks (DNN) such as the hyperparameters of DNN and TPU resource information generated by the TPU compiler. The new approach involves the offline acquisition of accurate spatial and temporal temperature maps captured from an external infrared thermal imaging camera under nominal working conditions of a chip. To build the dynamic power density map model, we apply generative adversarial networks (GAN) based on the workload-related features. Our study shows that the estimated total powers match the manufacturer's total power measurements extremely well. Experimental results further show that the predictions of power maps are quite accurate, with the RMSE of only 4.98\rm mW/mm^2, or 2.6\% of the full-scale error. The speed of deploying the proposed approach on an Intel Core i7-10710U is as fast as 6.9ms, which is suitable for real-time estimation.more » « less
-
The use of machine learning techniques has significantly increased the physics discovery potential of neutrino telescopes. In the upcoming years, we are expecting upgrades of currently existing detectors and new telescopes with novel experimental hardware, yielding more statistics as well as more complicated data signals. This calls for an upgrade on the software side needed to handle this more complex data in a more efficient way. Specifically, we seek low power and fast software methods to achieve real-time signal processing, where current machine learning methods are too expensive to be deployed in the resource-constrained regions where these experiments are located. We present the first attempt at and a proof-of-concept for enabling machine learning methods to be deployed in-detector for water/ice neutrino telescopes via quantization and deployment on Google Edge Tensor Processing Units (TPUs). We design a recursive neural network with a residual convolutional embedding and adapt a quantization process to deploy the algorithm on a Google Edge TPU. This algorithm can achieve similar reconstruction accuracy compared with traditional GPU-based machine learning solutions while requiring the same amount of power compared with CPU-based regression solutions, combining the high accuracy and low power advantages and enabling real-time in-detector machine learning in even the most power-restricted environments.more » « less
-
Approximate computing is a promising way to improve the power efficiency of deep learning. While recent work proposes new arithmetic circuits (adders and multipliers) that consume substantially less power at the cost of computation errors, these approximate circuits decrease the end-to-end accuracy of common models. We present AutoApprox, a framework to automatically generate approximate low-power deep learning accelerators without any accuracy loss. AutoApprox generates a wide range of approximate ASIC accelerators with a TPUv3 systolic-array template. AutoApprox uses a learned router to assign each DNN layer to an approximate systolic array from a bank of arrays with varying approximation levels. By tailoring this routing for a specific neural network architecture, we discover circuit designs without the accuracy penalty from prior methods. Moreover, AutoApprox optimizes for the end-to-end performance, power and area of the the whole chip and PE mapping rather than simply measuring the performance of the arithmetic units in iso-lation. To our knowledge, our work is the first to demonstrate the effectiveness of custom-tailored approximate circuits in delivering significant chip-level energy savings with zero accuracy loss on a large-scale dataset such as ImageNet. AutoApprox synthesizes a novel approximate accelerator based on the TPU that reduces end-to-end power consumption by 3.2% and area by 5.2% at a sub-10nm process with no degradation in ImageNet validation top-1 and top-5 accuracy.more » « less
-
Rapid growth in Deep Neural Network (DNN) workloads has increased the energy footprint of the Artificial Intelligence (AI) computing realm. For optimum energy efficiency, we propose operating a DNN hardware in the Low-Power Computing (LPC) region. However, operating at LPC causes increased delay sensitivity to Process Variation (PV). Delay faults are an intriguing consequence of PV. In this paper, we demonstrate the vulnerability of DNNs to delay variations, substantially lowering the prediction accuracy. To overcome delay faults, we present STRIVE—a post-fabrication fault detection and reactive error reduction technique. We also introduce a time-borrow correction technique to ensure error-free DNN computation.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.3390/jlpea12020032
