Quasi-Static Voltage Scaling for Energy Minimization With Time Constraints

Alexandru, Andrei; Eles, Petru; Jovanovic, Olivera; Schmitz, Marcus; Ogniewski, Jens; Peng, Zebo

doi:10.1109/tvlsi.2009.2030199

Cited by 21 publications

(7 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, power consumption becomes the critical limitation in an increasing number of applications [1,2,3], and one of the most effective ways to reduce power consumption is lowing the power supply voltage [4,5,6,7]. Hence, voltage scaling technique has been widely adopted in low-power design [8,9,10,11]. Aggressive voltage scaling into the sub-threshold region would provide better energy efficiency.…”

Section: Introductionmentioning

confidence: 99%

An ultra-low leakage energy efficient level shifter with wide conversion range

You

Jia

Tang

et al. 2019

IEICE Electron. Express

View full text Add to dashboard Cite

An ultra-low leakage energy efficient level shifter that can convert extremely low input voltage into the supply voltage level is presented in this paper. In order to reduce the leakage power dissipation, the super-cutoff mechanism and MTCMOS technique are utilized in the proposed structure. At the same time, a positive feedback circuit is inserted to avoid the loss of performance. Post-layout simulation results in a 55-nm MTCMOS process demonstrate that for the voltage level conversion from 0.3 V to 1.2 V, the proposed level shifter exhibits a propagation delay of 70.77 ns and an energy per transition of 89.55 fJ for input frequency of 1 MHz. Meanwhile, the static power of the proposed level shifter is as low as 27.82 pW. The proposed level shifter only occupies 7.79 um 2 , which demonstrates prominent area efficiency.

show abstract

Section: Introductionmentioning

confidence: 99%

An ultra-low leakage energy efficient level shifter with wide conversion range

You

Jia

Tang

et al. 2019

IEICE Electron. Express

View full text Add to dashboard Cite

show abstract

“…This gives a grand total of 3072 processing cores in recent GPU chips [1] enforcing energy efficiency as a primary concern. Earlier work has suggested supply voltage overscaling (VOS) [2] to reduce energy consumption. However, reducing the operating voltage of a core beyond a critical point leads to the so-called "path walls" [3].…”

Section: Introductionmentioning

confidence: 99%

Associative Memristive Memory for Approximate Computing in GPUs

Ghofrani

Rahimi

Lastras-Montaño

et al. 2016

IEEE J. Emerg. Sel. Topics Circuits Syst.

View full text Add to dashboard Cite

Using associative memories to enable computing-with-memory is a promising approach to improve energy efficiency. Associative memories can be tightly coupled with processing elements to restore and later recall function responses for a subset of input values. This approach avoids the actual function execution on the processing element to save on energy. The challenge, however, is to reduce the energy consumption of associative memory modules themselves. Here we address the challenge of designing ultra-low-power associative memories. We use memristive parts for memory implementation and demonstrate the energy saving potential of integrating associative memristive memory (AMM) into graphics processing units (GPUs). To reduce the energy consumption of AMM modules, we leverage approximate computing which benefits from application-level tolerance to errors: We employ voltage overscaling on AMM modules which deliberately relaxes its searching criteria to approximately match stored patterns within a 2 bit Hamming distance of the search pattern. This introduces some errors to the computation that are tolerable for target applications. We further reduce the energy consumption by employing purely resistive crossbar architectures for AMM modules. To evaluate the proposed architecture, we integrate AMM modules with floating point units in an AMD Southern Islands GPU and run four image processing kernels on an AMM-integrated GPU. Our experimental results show that employing AMM modules reduces energy consumption of running these kernels by 23%-45%, compared to a baseline GPU without AMM. The image processing kernels tolerate errors resulting from approximate search operations, maintaining an acceptable image quality, i.e., a PSNR above 30 dB.

show abstract

“…frequencies are calculated such that they guarantee that tasks will not violate their dead- Quasi−static Alg. proposed in [2] Greedy [4] Task−splitting [5] Greedy [4] Task−splitting [5] Hypothetical ( . It can be shown that running the voltage scaling algorithm for all possible start times t s of a task results in a frequency function f (t s ) that is convex [2].…”

Section: Online Algorithmmentioning

confidence: 99%

“…proposed in [2] Greedy [4] Task−splitting [5] Greedy [4] Task−splitting [5] Hypothetical ( . It can be shown that running the voltage scaling algorithm for all possible start times t s of a task results in a frequency function f (t s ) that is convex [2]. Thus any linear approximation of the frequency will result in frequency values higher or equal to the optimal one, and consequently no deadline will be violated.…”

Section: Online Algorithmmentioning

confidence: 99%