From Microarchitecture to Silicon: Building High-Performance CPUs across Intel and Qualcomm
DOI:
https://doi.org/10.64137/3107-9458/ICCSEMTI26-107Keywords:
CPU Microarchitecture, High-Performance Computing, Silicon Design, Intel Architecture, Qualcomm Snapdragon, Power–Performance Optimization, Front-End And Back-End DesignAbstract
This paper outlines the complete cycle of creating efficient central processing units (CPUs), starting from initial microarchitectural ideas to actual manufacturable silicon devices. It does so by focusing on two main industry paradigms: the Intel x86 ecosystem and Qualcomm's ARM-based architectures. In the scenario where modern computing workloads are increasingly demanding higher performance with limited power and thermal budgets, CPU architects are compelled to make a very complicated balancing act among instruction-set philosophy, pipeline depth, memory hierarchy, power management, and physical implementation. Thus, fundamental decisions on architectural trade-offs, validation methods, and the physical outcomes of the chips are discussed. The major problems addressed revolve around keeping the ramping up of the performance at par with the power density limits, designing scalable products for different market segments, handling the complexity of the design, and ensuring the chips can be produced at a high yield in advanced process nodes. The article puts forward a vendor-neutral methodology for design and verification, which is a combination of microarchitecture, power mgmt, and physical design constraints initially optimized together and then supported by iterative modeling and silicon-aware verification. A side-by-side comparison that takes a look into the CPU development stories of Intel and Qualcomm is provided through a case study where the focus is on the design decisions, the performance-per-watt metrics, and the post-silicon validation activities, including bring-up days. The paper finds that while the architectural philosophies are different, the converging practices—such as workload-driven designs, aggressive power optimizations, and close couplings of hardware/software co-design—are the common ingredients for the success in both ecosystems. Subsequently, the paper articulates the implications for the future of CPU architectures.
References
[1] Boggs, Darrell, et al. "The Microarchitecture of the Intel Pentium 4 Processor on 90nm Technology." Intel Technology Journal 8.1 (2004).
[2] Reed, Daniel, Dennis Gannon, and Jack Dongarra. "Reinventing high performance computing: challenges and opportunities." arXiv preprint arXiv:2203.02544 (2022).
[3] Gera, Prasun, et al. "Performance characterisation and simulation of Intel's integrated GPU architecture." 2018 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS). IEEE, 2018.
[4] Henry, Glenn, et al. "High-performance deep-learning coprocessor integrated into x86 soc with server-class cpus industrial product." 2020 ACM/IEEE 47th Annual International Symposium on Computer Architecture (ISCA). IEEE, 2020.
[5] Nikolić, Goran, et al. "Fifty years of microprocessor evolution: from single CPU to multicore and manycore systems." Facta universitatis-series: Electronics and Energetics 35.2 (2022): 155-186.
[6] Yahya, Jawad Haj, et al. "DarkGates: A Hybrid Power-Gating Architecture to Mitigate the Performance Impact of Dark-Silicon in High Performance Processors." 2022 IEEE International Symposium on High-Performance Computer Architecture (HPCA). IEEE, 2022.
[7] Halpern, Matthew, Yuhao Zhu, and Vijay Janapa Reddi. "Mobile CPU's rise to power: Quantifying the impact of generational mobile CPU design trends on performance, energy, and user satisfaction." 2016 IEEE International Symposium on High Performance Computer Architecture (HPCA). IEEE, 2016.
[8] Smith, Brad. "ARM and Intel battle over the mobile chip's future." Computer 41.5 (2008): 15-18.
[9] Kurth, Andreas, et al. "An open-source platform for high-performance non-coherent on-chip communication." IEEE Transactions on Computers 71.8 (2021): 1794-1809.
[10] Kalyanam, Vijay Kiran, et al. "A Proactive System for Voltage-Droop Mitigation in a 7-nm Hexagon™ Processor." IEEE Journal of Solid-State Circuits 56.4 (2020): 1166-1175.
[11] McKeown, Michael, et al. "Power and Energy Characterization of an Open Source 25-Core Manycore Processor." HPCA. 2018.
[12] Jiang, Jianyu, et al. "CRONUS: Fault-isolated, secure and high-performance heterogeneous computing for trusted execution environment." 2022 55th IEEE/ACM International Symposium on Microarchitecture (MICRO). IEEE, 2022.
[13] Garg, Rahul, and Laurie Hendren. "A portable and high-performance general matrix-multiply (GEMM) library for GPUs and single-chip CPU/GPU systems." 2014 22nd Euromicro International Conference on Parallel, Distributed, and Network-Based Processing. IEEE, 2014.
[14] Teodoro, George, et al. "Comparative performance analysis of intel xeon phi, gpu, and cpu." arXiv preprint arXiv:1311.0378 (2013).
[15] Cabrera, Anthony, et al. "Toward performance portable programming for heterogeneous systems on a chip: A case study with qualcomm snapdragon soc." 2021 IEEE High Performance Extreme Computing Conference (HPEC). IEEE, 2021.
