Xylem: Enhancing Vertical Thermal Conduction in 3D Processor-Memory Stacks

被引:20
作者
Agrawal, Aditya [1 ]
Torrellas, Josep [1 ]
Idgunji, Sachin [2 ]
机构
[1] Univ Illinois, Champaign, IL 61820 USA
[2] Nvidia Corp, Santa Clara, CA USA
来源
50TH ANNUAL IEEE/ACM INTERNATIONAL SYMPOSIUM ON MICROARCHITECTURE (MICRO) | 2017年
关键词
3D chip; processor-memory integration; thermal management;
D O I
10.1145/3123939.3124547
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
In upcoming architectures that stack processor and DRAM dies, temperatures are higher because of the increased transistor density and the high inter-layer thermal resistance. However, past research has underestimated the extent of the thermal bottleneck. Recent experimental work shows that the Die-to-Die (D2D) layers hinder effective heat transfer, likely leading to the capping of core frequencies. To address this problem, in this paper, we first show how to create pillars of high thermal conduction from the processor die to the heat sink. We do this by aligning and shorting dummy D2D mu bumps with thermal TSVs (TTSVs). This lowers processor temperatures substantially. We then improve application performance by boosting the processor frequency until we consume the available thermal headroom. Finally, these aligned and shorted dummy mu bump-TTSV sites create die regions of higher vertical thermal conduction. Hence, we propose to leverage them with three new architectural techniques: conductivity-aware thread placement, frequency boosting, and thread migration. We evaluate our scheme, called Xylem, using simulations of an 8-core processor at 2.4 GHz and 8 DRAM dies on top. mu Bump-TTSV alignment and shorting in a generic and in a customized Xylem design enable an average increase in processor frequency of 400 MHz and 720 MHz, respectively, at an area overhead of 0.63% and 0.81%, and without exceeding acceptable temperatures. This improves average application performance by 11% and 18%, respectively. Moreover, applying Xylem's conductivity-aware techniques enables further gains.
引用
收藏
页码:546 / 559
页数:14
相关论文
共 56 条
[1]  
[Anonymous], IEEE DES AUT C
[2]  
[Anonymous], IEEE INT C VLSI SYST
[3]  
[Anonymous], 2012, IEEE MICRO
[4]  
Athikulwongse K., 2010, IEEE INT C COMP AID
[5]  
Banerjee K., 2001, P IEEE
[6]  
Black B., 2006, IEEE INT S MICR
[7]  
Borkar S., 2011, IEEE DES AUT C
[8]   A Segmented Bloom Filter Algorithm for Efficient Predictors [J].
Breternitz, M. ;
Loh, Gabriel H. ;
Black, Bryan ;
Rupley, Jeffrey ;
Sassone, Peter G. ;
Attrot, Wesley ;
Wu, Youfeng .
20TH INTERNATIONAL SYMPOSIUM ON COMPUTER ARCHITECTURE AND HIGH PERFORMANCE COMPUTING, PROCEEDINGS, 2008, :123-+
[9]  
Chen Erh-Hao, 2013, IEEE EL COMP TECHN C
[10]  
Chen K. N., 2011, IET COMPUTERS DIGITA