Intel · posted · checked
GPU Performance Engineer
$142k – $232kHybrid · FolsomSenior
Who can apply
Hybrid · Folsom
Location: US, California, Folsom
Job Details:
Job Description:
The Role and Impact
As part of Intel’s GPU Performance team, the successful candidate will contribute to the architecture, performance analysis, and optimization of Intel’s next-generation Graphics Processing Units (GPUs). The role will focus on understanding and improving the performance of modern graphics/compute workloads, with the goal of delivering best-in-class performance and efficiency on current and future GPU architectures.
The candidate will perform detailed analysis of real-world and synthetic workloads to identify performance bottlenecks across the GPU hardware and software stack. This includes understanding application and shader behavior, graphics API usage, GPU pipeline utilization, memory and cache behavior, scheduling, and interactions between software and the underlying hardware architecture.
A key aspect of the role is translating workload-level performance observations into actionable hardware and software improvements. The candidate will work across architectural models, simulation environments, performance counters, traces, RTL, drivers, and application workloads to determine the root causes of performance issues and identify optimization opportunities.
The role requires close collaboration with GPU architects, RTL designers, driver and compiler engineers, software developers, validation teams, and application/workload experts in a highly multidisciplinary environment. The candidate will help connect what applications are doing at the API and shader level with how the underlying GPU architecture executes those workloads.
Key Responsibilities:
Analyze the performance of modern 3D/Compute workloads, games, benchmarks, ML/AI Kernels, and rendering applications on current and future GPU architectures.
Develop a deep understanding of workload behavior across the graphics pipeline, including geometry processing, rasterization, shader execution, texture operations, memory accesses, caching, synchronization, and render-target operations.
Identify GPU performance bottlenecks using architectural models, simulation environments, performance counters, workload traces, and other performance-analysis infrastructure.
Perform detailed shader analysis, including examining shader instruction sequences, execution characteristics, occupancy, latency, throughput, register usage, memory traffic, and other factors influencing GPU performance.
Investigate performance issues across both hardware and software boundaries, distinguishing architectural limitations from RTL implementation issues, driver behavior, compiler behavior, or workload characteristics.