A GPU SYSTEMS CLUB FOR DE ANZA STUDENTS
Explore GPU hardware and the kernels beneath PyTorch, from consumer cards to datacenter chips.
01 / GET TOGETHER
Our schedule is waiting on advisor availability. Confirmed meetings will appear here.
A simple tour of consumer GPU architecture and the NVIDIA H100 datacenter GPU. We’ll look at how cores and memory work together, then unpack the H100’s streaming multiprocessors, Tensor Cores, and HBM memory.
Get the details on DiscordWe’ll add meetings when our advisor’s availability is confirmed. Nothing to RSVP to just yet.
02 / OPEN A FEW TABS
Ideas for future sessions. Open a topic for the project we have in mind.
Follow a tensor operation from PyTorch to the GPU kernel that does the work. Inspect the threads, memory reads, and data transfers behind it, then time the operation on a CPU and GPU.
What we’d make: a small timing experiment and an annotated CUDA or Triton kernel.
PROJECT IDEAStart with a pixel’s coordinates and turn them into color. Build a moving gradient, draw a shape with a distance function, and change a shader to see exactly what each line does.
What we’d make: a small interactive shader in the browser, with controls for color, motion, and shape.
FUTURE DIRECTIONExplore what a program actually reads and writes. Use a small C or C++ example to inspect arrays, pointers, memory layout, and caches, then connect those ideas to GPU memory access.
What we’d make: a benchmark that compares two ways to walk through the same data and explains the difference.
FUTURE DIRECTION03 / YOU CAN START HERE
A GPU systems club, not a general AI club.
We explore GPU architecture, memory, kernels, and graphics. PyTorch gives us a starting point to trace what runs underneath a tensor operation; teaching its API or building AI apps isn’t the goal.
Inspired by GPU MODE, with short explanations and small experiments for De Anza students.
For your first meeting
No GPU or CUDA experience needed. Bring a laptop if you have one; basic Python helps with later code sessions.
COME HANG OUT
Get meeting updates, ask GPU questions, and share what you’re building.
Join the Discord