Memory Performance Architect
San Jose, CA, United States
At Cadence, we hire and develop leaders and innovators who want to make an impact on the world of technology. Memory Architect Intern
Location US (San Jose/Austin)
Summary
We are looking for an intern to help build performance models, perform architectural tradeoff analysis and enable data driven design decisions for our next generation DDR memory controller architectures that can meet todays complex SoC and workload requirements. An architecture foundation and hardware modelling experience (C++/Python) is desired.
Responsibilities
Develop cycle-level C++/SystemC performance models and analysis of hardware features, applications, benchmarks, and use cases
Analyze memory subsystem-level architectural trade-offs (throughput, latency, power, hardware cost) across different scenarios
Develop synthetic memory traffic/traces that are representative of real-world SoC components (CPU, GPU, DSP, NoC, etc)
Develop scripts to automate generation of various performance metrics and statistics post RTL simulation that helps identify performance bottlenecks
Correlate performance models to match RTL across configurations and traffic conditions
Work with Memory Architects to understand feature requirements and evaluate architectural choices and tradeoffs
Required Skills
Currently pursuing a MS in Electrical/Computer Engineering or equivalent
Strong coding skills in C++/SystemC, python and similar programming languages
Basic understanding of performance principles, e.g. Amdahls Law, Queuing Theory, throughput/latency tradeoffs
Additional Skills
Experience in architecture analysis and hardware modeling (functional or performance) and simulator development skills
Understanding RTL, analyzing waveforms
Understanding of memory protocols and timing DDR4, DDR5, LP4, LP5
Experience with industry standard or academic performance simulators at any level Memory, SoC, CPU models
Experience using performance benchmarks
Concepts related to Quality of Service (QoS) and how memory controller can tradeoff performance and QoS
Were doing work that matters. Help us solve what others cant.