Senior Compute Kernel Architect, GPU Power

What you'll do

• Design CUDA kernels to enhance GPU power consumption by targeting worst-case current draw.
• Validate PDN assumptions with hardware architects using di/dt specifications to identify weak points.
• Develop a library of power stress microbenchmarks for various GPU functional units.
• Analyze trade-offs between kernel throughput, power efficiency, and voltage stability for future GPUs.
• Ensure alignment of power stress methodologies from pre-silicon simulation through post-silicon validation.

What we're looking for

• 5+ years experience in GPU kernel development with CUDA programming.
• Strong skills in CUDA and C++ programming, including assembly or PTX level optimization.
• Proficiency with GPU performance profiling tools like Nsight Compute and nvprof.
• Deep understanding of GPU architecture and PDN design principles.
• Experience writing power stress microbenchmarks for worst-case scenarios.
• Direct experience with post-silicon power characterization using lab equipment.
• Knowledge of DVFS, AVFS features and their interaction with kernel behavior.