Galaxy: a high-performance energy-efficient multi-chip architecture using photonic interconnects
Galaxy: a high-performance energy-efficient multi-chip architecture using photonic interconnects
复制标题
DOI：
10.1145/2597652.2597664
复制
发表时间：
2014-06
期刊：
影响因子：
--
通讯作者：
Y. Demir;Yan Pan;Seokwoo Song;N. Hardavellas;John Kim;G. Memik
中科院分区：
文献类型：
--
作者：
Y. Demir;Yan Pan;Seokwoo Song;N. Hardavellas;John Kim;G. Memik
现代半导体技术的可伸缩性趋势导致越来越密集的多核芯片。不幸的是，面积，功率，芯片外带宽和产量的物理限制将单芯片设计限制为相对较少的核心，除了缩放量表变得不切实际。多芯片设计克服了这些约束，并且可以通过传统的单芯片体系结构达到无法实现的尺度。但是，为了提供相称的性能，多芯片架构需要带宽，延迟和能量消耗的跨芯片互连，远远超出了电信号的范围。我们提出了Galaxy，这是一种通过光纤连接多个较小的芯片来构建多核“虚拟芯片”的架构。纤维的光学损失较低，可以灵活地放置芯片，并提供更简单的包装，功率和热量要求。同时，光信号传导的低潜伏期和高带宽密度保持核心的紧密耦合，从而使虚拟芯片可以匹配不受面积，功率和带宽限制的单个芯片的性能。我们的结果表明，Galaxy在具有电气或光子互连的最佳单芯片替代方案上达到2.2倍的加速度（最大3.4倍），而较小的能量 - 播放产品（最大6.8倍）。我们表明，与带有硅波导的宏观芯片相比，Galaxy尺度达到4K核，并在激光量下降6倍时达到2.5倍的速度。
The scalability trends of modern semiconductor technology lead to increasingly dense multicore chips. Unfortunately, physical limitations in area, power, off-chip bandwidth, and yield constrain single-chip designs to a relatively small number of cores, beyond which scaling becomes impractical. Multi-chip designs overcome these constraints, and can reach scales impossible to realize with conventional single-chip architectures. However, to deliver commensurate performance, multi-chip architectures require a cross-chip interconnect with bandwidth, latency, and energy consumption well beyond the reach of electrical signaling. We propose Galaxy, an architecture that enables the construction of a many-core "virtual chip" by connecting multiple smaller chiplets through optical fibers. The low optical loss of fibers allows the flexible placement of chiplets, and offers simpler packaging, power, and heat requirements. At the same time, the low latency and high bandwidth density of optical signaling maintain the tight coupling of cores, allowing the virtual chip to match the performance of a single chip that is not subject to area, power, and bandwidth limitations. Our results indicate that Galaxy attains speedup of 2.2x over the best single-chip alternatives with electrical or photonic interconnects (3.4x maximum), and 2.6x smaller energy-delay product (6.8x maximum). We show that Galaxy scales to 4K cores and attains 2.5x speedup at 6x lower laser power compared to a Macrochip with silicon waveguides.