Overview As a GPU Performance Architect in RTG, you help shape high-performance graphics and compute hardware. You collaborate with a global, cross-functional team to analyze workloads, identify bottlenecks, and explore innovative PPA improvements. You will contribute to GPU cores and SoCs, while advancing analytics tools and simulation methods. This role offers impact on next-gen GPUs and AI/ML workloads and a chance to work with a world-class GPU team.
Responsibilities- Analyze workloads and competitive scenarios for game, AI, and ML applications
- Decompose complex technical problems, outline viable solutions, and drive team progress
- Collaborate with architects to identify bottlenecks in graphics cores and SoCs
- Evaluate existing and emerging graphics/compute paradigms and algorithms
- Develop and run RTL and high-level simulations to estimate gains
- Propose innovative PPA improvement ideas and validate them
- Lead initiatives to enhance RTG analysis tools
- Coordinate with engineers and managers across multiple sites
Key requirements- Experience in architecture, performance analysis, and AI/ML/graphics/compute algorithms
- Proficiency in kernel development for LLMs using PyTorch, Triton, CUDA, HIP
- Strong C/C++ and scripting (Python) skills
- Knowledge of graphics/compute APIs (DirectX, Vulkan, CUDA, HIP)
- Understanding of GPU architecture and compilers
- Experience with graphics/GPU profiling tools is a plus (PIX, RenderDoc, AMD tools)
- Undergrad degree in a related field (Electrical Engineering, Computer Architecture, or Computer Science)
- Collaborative mindset
- Mentoring and team-building
- Strong analytical and problem-solving abilities
- GPU architecture
- Performance analysis
- AI/ML/graphics/compute algorithms
