CUDA
Nvidia's parallel computing platform and API for general-purpose programming on Nvidia GPUs.
Commenters describe CUDA as the GPU compute platform with no real alternative123, one that made ML fun and is easier to work with than Vulkan for rendering.45 The most common caveat is that it ties you to Nvidia hardware, leaving AMD users struggling678, and one commenter warns that Nvidia documents poorly which hardware supports which features.9
Why people recommend it
No real alternative · 3
Commenters say nothing from AMD, Apple or Google matches CUDA yet, though rivals are chipping away at its edge.
“there is literally no alternative to CUDA.”
“you should just buy an NVIDIA card, CUDA is just that good”
Show 1 more
“CUDA still has an edge but every release like this chips away at that.”
Compute power · 2
It unlocks far more power than CPUs for ML and other suited workloads.
“there is literally no alternative to CUDA.”
“man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun”
Usability · 2
One rendering developer switched from Vulkan because writing a fast rasterizer in CUDA was easier.
“Easier to write a fast software rasterizer in cuda, than to draw a single triangle in Vulkan”
“man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun”
Broad card support · 1
CUDA works across every card Nvidia makes, not just a select few.
“A stack that can be used for EVERY card they produce, like CUDA, not just a select few.”
Caveats and dissent
Nvidia lock-in · 3
CUDA requires an Nvidia GPU, and AMD owners describe the experience as a struggle.
“man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun”
“you should just buy an NVIDIA card, CUDA is just that good”
Show 1 more
“A stack that can be used for EVERY card they produce, like CUDA, not just a select few.”
Hardware documentation · 1
Nvidia poorly documents feature support per GPU, so users rely on third-party tables.
“Nvidia is bad at documenting which hardware does what things, and CUDA users often have to use third party tables”
Learning curve · 1
One commenter says it is a tool for skilled users and not super easy to use.
“there is literally no alternative to CUDA.”
Often compared with
Who recommends it in their own words
Threads where it came up
- Google open-sources TorchTPU, a native PyTorch back end for TPUs 1 mention
- CUDA for AMD on Windows 1 mention
- Nvidia's Risky Business 1 mention
- Can AMD Break the CUDA Moat? AMD Advancing AI 2026 1 mention
- Learn OpenGL, extensive tutorial resource for learning Modern OpenGL 1 mention
- Alternative(s) to run CUDA on non-Nvidia hardware 1 mention
Data: cuda.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 8
“CUDA still has an edge but every release like this chips away at that.”
“man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun”
“there is literally no alternative to CUDA.”
“Nvidia says CUDA Tiles says CUDA Tiles MLIR is open source.”
“you should just buy an NVIDIA card, CUDA is just that good”
“Nvidia is bad at documenting which hardware does what things, and CUDA users often have to use third party tables”
“Easier to write a fast software rasterizer in cuda, than to draw a single triangle in Vulkan”
“A stack that can be used for EVERY card they produce, like CUDA, not just a select few.”