Minimum requirement: an NVIDIA GPU
Required
BallTrack AI runs a neural network on every single frame to find the ball and the players. That detection
runs on CUDA, so an NVIDIA GPU is required, not
optional — without one the app has nothing to run on.
There is no usable fallback: a CPU / Apple Silicon code path
exists, but it is orders of magnitude too slow to
process a full match, and it is not offered as an option.
VRAM is what sets your speed. The app reads your
card’s memory and picks the number of detection
workers for you — more VRAM means more parallel
workers and a faster run. Each worker is a separate process holding
its own copy of the detection model on the GPU, so VRAM is what caps
how many run at once. There is only one detection
model, chosen for you: no model picker, no worker slider,
nothing to tune by hand.
-
16 GB VRAM
Supported minimum — the app runs 2 detection workers
-
24 GB VRAM
The validated tier — all 5 workers (the reference RTX 3090)
-
Below 16 GB
Experimental — the app runs a single worker; unvalidated
16 GB is the supported floor. The pipeline and
every measured number on these pages were validated on a
24 GB card (an RTX 3090); 16 GB cards
run with a reduced worker count that the app sets automatically.
Cards below 16 GB are treated as experimental —
the app will still start and run a single worker, but that tier is
unvalidated. Cards older than the RTX 20 series, and
non-NVIDIA GPUs, are not supported at all. See the
GPU performance table for
the worker ladder.
Honest caveat: only the 24 GB tier has actually
been measured — five workers plus the parent process using about
11 GB on a 24 GB card. 16 GB is the supported floor,
in the app’s own automatic cap and on paper, but
24 GB is what has been run end to end.
FullField Stitch above has no such
requirement — it stitches on any OpenCL-capable GPU, NVIDIA or
not, and its planned Apple Silicon build needs no NVIDIA
hardware at all. The hard NVIDIA requirement applies to BallTrack
AI only.