{"slug": "cuda", "name": "CUDA", "category": "dev-libs", "category_name": "Libraries and frameworks", "description": "Nvidia's parallel computing platform and API for general-purpose programming on Nvidia GPUs.", "recommenders": 5, "commenters": 8, "warners": 1, "score": 4.047, "rising": 1.02, "first": "2025-11-15", "latest": "2026-10-20", "why": "Commenters describe CUDA as the GPU compute platform with no real alternative, one that made ML fun and is easier to work with than Vulkan for rendering.", "why_is_quote": false, "has_verdict": true, "sentiment": {"recommend": 5, "neutral": 0, "compare": 2, "warn": 1}, "spark": [1, 0, 0, 0, 0, 0, 0, 0, 2, 1, 1, 0], "listed": true, "loved": ["No real alternative", "Compute power"], "caveat": "Nvidia lock-in", "short_quote": "you should just buy an NVIDIA card, CUDA is just that good", "icon": true, "domain": null, "github_repo": null, "official_url": "https://developer.nvidia.com/cuda-toolkit", "pricing": "free", "aliases": ["CUDA"], "verdict": [{"text": "Commenters describe CUDA as the GPU compute platform with no real alternative", "cites": [{"id": 788242464, "item_id": 49265154, "url": "https://news.ycombinator.com/item?id=49265154", "by": "kevstev", "date": "2026-08-11", "stance": "recommend", "quote": "there is literally no alternative to CUDA.", "excerpt": "I think OP is overstating it a bit tbh. So Nvidia has the market for the hardware, which helps, but there is literally no alternative to CUDA. Nvidia keeps it a scalpel for skilled users, its not super easy to use, but unlocks orders of more magnitudes of power for the use cases it excels at vs CPUs. I don't have access to anything like it in the Apple ecosystem. AMD has had years to try and counter it, but just has ", "context": "", "use_case": "GPU computing", "switched_from": "", "story": "Nvidia's Risky Business"}, {"id": 784394544, "item_id": 49024659, "url": "https://news.ycombinator.com/item?id=49024659", "by": "cyber_kinetist", "date": "2026-07-23", "stance": "recommend", "quote": "you should just buy an NVIDIA card, CUDA is just that good", "excerpt": "The one and only Holy Bible of Graphics Programming. If you're starting to learn computer graphics, just study through the entire site and do the examples one by one. It doesn't matter one bit that it uses a slightly outdated API called OpenGL - you're supposed to learn how to render things first, not about some weird obscure hardware / driver details! After you've learned it, you can start learning CUDA if you want ", "context": "", "use_case": "GPU compute", "switched_from": "", "story": "Learn OpenGL, extensive tutorial resource for learning Modern OpenGL"}, {"id": 802743408, "item_id": 50171463, "url": "https://news.ycombinator.com/item?id=50171463", "by": "jacquesm", "date": "2026-10-20", "stance": "compare", "quote": "CUDA still has an edge but every release like this chips away at that.", "excerpt": "Those are the two main interfaces: the inference hardware and the endpoints. Endpoints are pretty much standardized to the point that you can switch them out in seconds. CUDA still has an edge but every release like this chips away at that. Even so, TPUs are unobtanium so while I appreciate Google's effort here I also recognize that it is not something that will be useful to me. In that sense AMD is still better posi", "context": "", "use_case": "inference hardware stack", "switched_from": "", "story": "Google open-sources TorchTPU, a native PyTorch back end for TPUs"}]}, {"text": ", one that made ML fun and is easier to work with than Vulkan for rendering.", "cites": [{"id": 794997360, "item_id": 49687335, "url": "https://news.ycombinator.com/item?id=49687335", "by": "Nurysso", "date": "2026-09-13", "stance": "recommend", "quote": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun", "excerpt": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun and on amd its a war especially on rdna 2 cards which i have. hopefully one day we will be able to properly translate cuda for its amd counter parts", "context": "RDNA 2 AMD GPU owner", "use_case": "machine learning", "switched_from": "", "story": "CUDA for AMD on Windows"}, {"id": 782471296, "item_id": 48904456, "url": "https://news.ycombinator.com/item?id=48904456", "by": "mschuetz", "date": "2026-07-14", "stance": "recommend", "quote": "Easier to write a fast software rasterizer in cuda, than to draw a single triangle in Vulkan", "excerpt": "> unimportant implementation detail that should be assessed on performance alone rather than usability I disagree. Usability is important, or you end up with garbage like Vulkan. I'd rather have 99% of the performance for 100% usability, than 100% of the performance for 5% of the usability. Unfortunately when it comes to graphics APIs, the choices are between bad and worse, which is why I switched to doing real time ", "context": "real-time rendering developer", "use_case": "real-time rendering", "switched_from": "Vulkan", "story": "Writing a bindless GPU abstraction layer"}]}, {"text": " The most common caveat is that it ties you to Nvidia hardware, leaving AMD users struggling", "cites": [{"id": 794997360, "item_id": 49687335, "url": "https://news.ycombinator.com/item?id=49687335", "by": "Nurysso", "date": "2026-09-13", "stance": "recommend", "quote": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun", "excerpt": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun and on amd its a war especially on rdna 2 cards which i have. hopefully one day we will be able to properly translate cuda for its amd counter parts", "context": "RDNA 2 AMD GPU owner", "use_case": "machine learning", "switched_from": "", "story": "CUDA for AMD on Windows"}, {"id": 784394544, "item_id": 49024659, "url": "https://news.ycombinator.com/item?id=49024659", "by": "cyber_kinetist", "date": "2026-07-23", "stance": "recommend", "quote": "you should just buy an NVIDIA card, CUDA is just that good", "excerpt": "The one and only Holy Bible of Graphics Programming. If you're starting to learn computer graphics, just study through the entire site and do the examples one by one. It doesn't matter one bit that it uses a slightly outdated API called OpenGL - you're supposed to learn how to render things first, not about some weird obscure hardware / driver details! After you've learned it, you can start learning CUDA if you want ", "context": "", "use_case": "GPU compute", "switched_from": "", "story": "Learn OpenGL, extensive tutorial resource for learning Modern OpenGL"}, {"id": 734999856, "item_id": 45937491, "url": "https://news.ycombinator.com/item?id=45937491", "by": "boxerab", "date": "2025-11-15", "stance": "recommend", "quote": "A stack that can be used for EVERY card they produce, like CUDA, not just a select few.", "excerpt": "This is a great project, but the bigger question is: why isn't AMD doing this themselves? It continues to boggle my mind how much they don't seem to get the importance of a mature software stack when it is so obviously the key to the success of team red. A stack that can be used for EVERY card they produce, like CUDA, not just a select few. I used to believe that AMD the underdog would catch up some day, but I've mor", "context": "", "use_case": "GPU compute stack", "switched_from": "", "story": "AMD GPUs Go Brrr"}]}, {"text": ", and one commenter warns that Nvidia documents poorly which hardware supports which features.", "cites": [{"id": 782479504, "item_id": 48904969, "url": "https://news.ycombinator.com/item?id=48904969", "by": "DiabloD3", "date": "2026-07-14", "stance": "warn", "quote": "Nvidia is bad at documenting which hardware does what things, and CUDA users often have to use third party tables", "excerpt": "I love how people say things like \"extension spaghetti\", as if all other non-standard APIs have the same problem: hardware gets new features that people want to use from that API, API gains extension to use that hardware feature. CUDA is no different, in fact, often worse. Nvidia is bad at documenting which hardware does what things, and CUDA users often have to use third party tables to figure out what hardware can'", "context": "", "use_case": "GPU programming", "switched_from": "", "story": "Alternative(s) to run CUDA on non-Nvidia hardware"}]}], "praise": [{"label": "No real alternative", "summary": "Commenters say nothing from AMD, Apple or Google matches CUDA yet, though rivals are chipping away at its edge.", "quotes": [{"id": 788242464, "item_id": 49265154, "url": "https://news.ycombinator.com/item?id=49265154", "by": "kevstev", "date": "2026-08-11", "stance": "recommend", "quote": "there is literally no alternative to CUDA.", "excerpt": "I think OP is overstating it a bit tbh. So Nvidia has the market for the hardware, which helps, but there is literally no alternative to CUDA. Nvidia keeps it a scalpel for skilled users, its not super easy to use, but unlocks orders of more magnitudes of power for the use cases it excels at vs CPUs. I don't have access to anything like it in the Apple ecosystem. AMD has had years to try and counter it, but just has ", "context": "", "use_case": "GPU computing", "switched_from": "", "story": "Nvidia's Risky Business"}, {"id": 784394544, "item_id": 49024659, "url": "https://news.ycombinator.com/item?id=49024659", "by": "cyber_kinetist", "date": "2026-07-23", "stance": "recommend", "quote": "you should just buy an NVIDIA card, CUDA is just that good", "excerpt": "The one and only Holy Bible of Graphics Programming. If you're starting to learn computer graphics, just study through the entire site and do the examples one by one. It doesn't matter one bit that it uses a slightly outdated API called OpenGL - you're supposed to learn how to render things first, not about some weird obscure hardware / driver details! After you've learned it, you can start learning CUDA if you want ", "context": "", "use_case": "GPU compute", "switched_from": "", "story": "Learn OpenGL, extensive tutorial resource for learning Modern OpenGL"}, {"id": 802743408, "item_id": 50171463, "url": "https://news.ycombinator.com/item?id=50171463", "by": "jacquesm", "date": "2026-10-20", "stance": "compare", "quote": "CUDA still has an edge but every release like this chips away at that.", "excerpt": "Those are the two main interfaces: the inference hardware and the endpoints. Endpoints are pretty much standardized to the point that you can switch them out in seconds. CUDA still has an edge but every release like this chips away at that. Even so, TPUs are unobtanium so while I appreciate Google's effort here I also recognize that it is not something that will be useful to me. In that sense AMD is still better posi", "context": "", "use_case": "inference hardware stack", "switched_from": "", "story": "Google open-sources TorchTPU, a native PyTorch back end for TPUs"}]}, {"label": "Compute power", "summary": "It unlocks far more power than CPUs for ML and other suited workloads.", "quotes": [{"id": 788242464, "item_id": 49265154, "url": "https://news.ycombinator.com/item?id=49265154", "by": "kevstev", "date": "2026-08-11", "stance": "recommend", "quote": "there is literally no alternative to CUDA.", "excerpt": "I think OP is overstating it a bit tbh. So Nvidia has the market for the hardware, which helps, but there is literally no alternative to CUDA. Nvidia keeps it a scalpel for skilled users, its not super easy to use, but unlocks orders of more magnitudes of power for the use cases it excels at vs CPUs. I don't have access to anything like it in the Apple ecosystem. AMD has had years to try and counter it, but just has ", "context": "", "use_case": "GPU computing", "switched_from": "", "story": "Nvidia's Risky Business"}, {"id": 794997360, "item_id": 49687335, "url": "https://news.ycombinator.com/item?id=49687335", "by": "Nurysso", "date": "2026-09-13", "stance": "recommend", "quote": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun", "excerpt": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun and on amd its a war especially on rdna 2 cards which i have. hopefully one day we will be able to properly translate cuda for its amd counter parts", "context": "RDNA 2 AMD GPU owner", "use_case": "machine learning", "switched_from": "", "story": "CUDA for AMD on Windows"}]}, {"label": "Usability", "summary": "One rendering developer switched from Vulkan because writing a fast rasterizer in CUDA was easier.", "quotes": [{"id": 782471296, "item_id": 48904456, "url": "https://news.ycombinator.com/item?id=48904456", "by": "mschuetz", "date": "2026-07-14", "stance": "recommend", "quote": "Easier to write a fast software rasterizer in cuda, than to draw a single triangle in Vulkan", "excerpt": "> unimportant implementation detail that should be assessed on performance alone rather than usability I disagree. Usability is important, or you end up with garbage like Vulkan. I'd rather have 99% of the performance for 100% usability, than 100% of the performance for 5% of the usability. Unfortunately when it comes to graphics APIs, the choices are between bad and worse, which is why I switched to doing real time ", "context": "real-time rendering developer", "use_case": "real-time rendering", "switched_from": "Vulkan", "story": "Writing a bindless GPU abstraction layer"}, {"id": 794997360, "item_id": 49687335, "url": "https://news.ycombinator.com/item?id=49687335", "by": "Nurysso", "date": "2026-09-13", "stance": "recommend", "quote": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun", "excerpt": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun and on amd its a war especially on rdna 2 cards which i have. hopefully one day we will be able to properly translate cuda for its amd counter parts", "context": "RDNA 2 AMD GPU owner", "use_case": "machine learning", "switched_from": "", "story": "CUDA for AMD on Windows"}]}, {"label": "Broad card support", "summary": "CUDA works across every card Nvidia makes, not just a select few.", "quotes": [{"id": 734999856, "item_id": 45937491, "url": "https://news.ycombinator.com/item?id=45937491", "by": "boxerab", "date": "2025-11-15", "stance": "recommend", "quote": "A stack that can be used for EVERY card they produce, like CUDA, not just a select few.", "excerpt": "This is a great project, but the bigger question is: why isn't AMD doing this themselves? It continues to boggle my mind how much they don't seem to get the importance of a mature software stack when it is so obviously the key to the success of team red. A stack that can be used for EVERY card they produce, like CUDA, not just a select few. I used to believe that AMD the underdog would catch up some day, but I've mor", "context": "", "use_case": "GPU compute stack", "switched_from": "", "story": "AMD GPUs Go Brrr"}]}], "caveats": [{"label": "Nvidia lock-in", "summary": "CUDA requires an Nvidia GPU, and AMD owners describe the experience as a struggle.", "quotes": [{"id": 794997360, "item_id": 49687335, "url": "https://news.ycombinator.com/item?id=49687335", "by": "Nurysso", "date": "2026-09-13", "stance": "recommend", "quote": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun", "excerpt": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun and on amd its a war especially on rdna 2 cards which i have. hopefully one day we will be able to properly translate cuda for its amd counter parts", "context": "RDNA 2 AMD GPU owner", "use_case": "machine learning", "switched_from": "", "story": "CUDA for AMD on Windows"}, {"id": 784394544, "item_id": 49024659, "url": "https://news.ycombinator.com/item?id=49024659", "by": "cyber_kinetist", "date": "2026-07-23", "stance": "recommend", "quote": "you should just buy an NVIDIA card, CUDA is just that good", "excerpt": "The one and only Holy Bible of Graphics Programming. If you're starting to learn computer graphics, just study through the entire site and do the examples one by one. It doesn't matter one bit that it uses a slightly outdated API called OpenGL - you're supposed to learn how to render things first, not about some weird obscure hardware / driver details! After you've learned it, you can start learning CUDA if you want ", "context": "", "use_case": "GPU compute", "switched_from": "", "story": "Learn OpenGL, extensive tutorial resource for learning Modern OpenGL"}, {"id": 734999856, "item_id": 45937491, "url": "https://news.ycombinator.com/item?id=45937491", "by": "boxerab", "date": "2025-11-15", "stance": "recommend", "quote": "A stack that can be used for EVERY card they produce, like CUDA, not just a select few.", "excerpt": "This is a great project, but the bigger question is: why isn't AMD doing this themselves? It continues to boggle my mind how much they don't seem to get the importance of a mature software stack when it is so obviously the key to the success of team red. A stack that can be used for EVERY card they produce, like CUDA, not just a select few. I used to believe that AMD the underdog would catch up some day, but I've mor", "context": "", "use_case": "GPU compute stack", "switched_from": "", "story": "AMD GPUs Go Brrr"}]}, {"label": "Hardware documentation", "summary": "Nvidia poorly documents feature support per GPU, so users rely on third-party tables.", "quotes": [{"id": 782479504, "item_id": 48904969, "url": "https://news.ycombinator.com/item?id=48904969", "by": "DiabloD3", "date": "2026-07-14", "stance": "warn", "quote": "Nvidia is bad at documenting which hardware does what things, and CUDA users often have to use third party tables", "excerpt": "I love how people say things like \"extension spaghetti\", as if all other non-standard APIs have the same problem: hardware gets new features that people want to use from that API, API gains extension to use that hardware feature. CUDA is no different, in fact, often worse. Nvidia is bad at documenting which hardware does what things, and CUDA users often have to use third party tables to figure out what hardware can'", "context": "", "use_case": "GPU programming", "switched_from": "", "story": "Alternative(s) to run CUDA on non-Nvidia hardware"}]}, {"label": "Learning curve", "summary": "One commenter says it is a tool for skilled users and not super easy to use.", "quotes": [{"id": 788242464, "item_id": 49265154, "url": "https://news.ycombinator.com/item?id=49265154", "by": "kevstev", "date": "2026-08-11", "stance": "recommend", "quote": "there is literally no alternative to CUDA.", "excerpt": "I think OP is overstating it a bit tbh. So Nvidia has the market for the hardware, which helps, but there is literally no alternative to CUDA. Nvidia keeps it a scalpel for skilled users, its not super easy to use, but unlocks orders of more magnitudes of power for the use cases it excels at vs CPUs. I don't have access to anything like it in the Apple ecosystem. AMD has had years to try and counter it, but just has ", "context": "", "use_case": "GPU computing", "switched_from": "", "story": "Nvidia's Risky Business"}]}], "who": [{"by": "Nurysso", "context": "RDNA 2 AMD GPU owner", "url": "https://news.ycombinator.com/item?id=49687335", "date": "2026-09-13"}, {"by": "mschuetz", "context": "real-time rendering developer", "url": "https://news.ycombinator.com/item?id=48904456", "date": "2026-07-14"}], "tags": ["for ML work", "for GPU compute", "for real-time rendering", "for Nvidia GPU owners"], "use_cases": ["gpu compute", "inference hardware stack", "machine learning", "gpu computing", "gpu programming", "real-time rendering"], "compared": [{"slug": "vulkan", "name": "Vulkan", "co": 5, "switched_from": 1, "switched_to": 0}, {"slug": "rocm", "name": "ROCm", "co": 2, "switched_from": 0, "switched_to": 0}, {"slug": "opengl", "name": "OpenGL", "co": 2, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=50169053", "title": "Google open-sources TorchTPU, a native PyTorch back end for TPUs", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49684356", "title": "CUDA for AMD on Windows", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49255710", "title": "Nvidia's Risky Business", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49078762", "title": "Can AMD Break the CUDA Moat? AMD Advancing AI 2026", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=49022634", "title": "Learn OpenGL, extensive tutorial resource for learning Modern OpenGL", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=48903715", "title": "Alternative(s) to run CUDA on non-Nvidia hardware", "mentions": 1}], "sentiment_by_month": {"2025-11": {"recommend": 1}, "2025-12": {}, "2026-01": {}, "2026-02": {}, "2026-03": {}, "2026-04": {}, "2026-05": {}, "2026-06": {}, "2026-07": {"recommend": 2, "warn": 1, "compare": 1}, "2026-08": {"recommend": 1}, "2026-09": {"recommend": 1}, "2026-10": {"compare": 1}}, "quotes": [{"id": 802743408, "item_id": 50171463, "url": "https://news.ycombinator.com/item?id=50171463", "by": "jacquesm", "date": "2026-10-20", "stance": "compare", "quote": "CUDA still has an edge but every release like this chips away at that.", "excerpt": "Those are the two main interfaces: the inference hardware and the endpoints. Endpoints are pretty much standardized to the point that you can switch them out in seconds. CUDA still has an edge but every release like this chips away at that. Even so, TPUs are unobtanium so while I appreciate Google's effort here I also recognize that it is not something that will be useful to me. In that sense AMD is still better posi", "context": "", "use_case": "inference hardware stack", "switched_from": "", "story": "Google open-sources TorchTPU, a native PyTorch back end for TPUs"}, {"id": 794997360, "item_id": 49687335, "url": "https://news.ycombinator.com/item?id=49687335", "by": "Nurysso", "date": "2026-09-13", "stance": "recommend", "quote": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun", "excerpt": "man i can't say how much i used to like cuda when i had a nvidia gpu it made ml so much fun and on amd its a war especially on rdna 2 cards which i have. hopefully one day we will be able to properly translate cuda for its amd counter parts", "context": "RDNA 2 AMD GPU owner", "use_case": "machine learning", "switched_from": "", "story": "CUDA for AMD on Windows"}, {"id": 788242464, "item_id": 49265154, "url": "https://news.ycombinator.com/item?id=49265154", "by": "kevstev", "date": "2026-08-11", "stance": "recommend", "quote": "there is literally no alternative to CUDA.", "excerpt": "I think OP is overstating it a bit tbh. So Nvidia has the market for the hardware, which helps, but there is literally no alternative to CUDA. Nvidia keeps it a scalpel for skilled users, its not super easy to use, but unlocks orders of more magnitudes of power for the use cases it excels at vs CPUs. I don't have access to anything like it in the Apple ecosystem. AMD has had years to try and counter it, but just has ", "context": "", "use_case": "GPU computing", "switched_from": "", "story": "Nvidia's Risky Business"}, {"id": 785274225, "item_id": 49079639, "url": "https://news.ycombinator.com/item?id=49079639", "by": "jauntywundrkind", "date": "2026-07-27", "stance": "compare", "quote": "Nvidia says CUDA Tiles says CUDA Tiles MLIR is open source.", "excerpt": "I mean, Nvidia says CUDA Tiles says CUDA Tiles MLIR is open source. https://news.ycombinator.com/item?id=46330732 Afaik, AMD, Google, and everyone else is still using some combo of mostly XLA, with StableHLO, IREE, etc? https://openxla.org/ I was going to lightly blow off more steam about AMD's TheRock all-in-one monorepo-ish thing for their graphics drivers being basically unbuildable on anything but a single distro", "context": "", "use_case": "GPU compute", "switched_from": "", "story": "Can AMD Break the CUDA Moat? AMD Advancing AI 2026"}, {"id": 784394544, "item_id": 49024659, "url": "https://news.ycombinator.com/item?id=49024659", "by": "cyber_kinetist", "date": "2026-07-23", "stance": "recommend", "quote": "you should just buy an NVIDIA card, CUDA is just that good", "excerpt": "The one and only Holy Bible of Graphics Programming. If you're starting to learn computer graphics, just study through the entire site and do the examples one by one. It doesn't matter one bit that it uses a slightly outdated API called OpenGL - you're supposed to learn how to render things first, not about some weird obscure hardware / driver details! After you've learned it, you can start learning CUDA if you want ", "context": "", "use_case": "GPU compute", "switched_from": "", "story": "Learn OpenGL, extensive tutorial resource for learning Modern OpenGL"}, {"id": 782479504, "item_id": 48904969, "url": "https://news.ycombinator.com/item?id=48904969", "by": "DiabloD3", "date": "2026-07-14", "stance": "warn", "quote": "Nvidia is bad at documenting which hardware does what things, and CUDA users often have to use third party tables", "excerpt": "I love how people say things like \"extension spaghetti\", as if all other non-standard APIs have the same problem: hardware gets new features that people want to use from that API, API gains extension to use that hardware feature. CUDA is no different, in fact, often worse. Nvidia is bad at documenting which hardware does what things, and CUDA users often have to use third party tables to figure out what hardware can'", "context": "", "use_case": "GPU programming", "switched_from": "", "story": "Alternative(s) to run CUDA on non-Nvidia hardware"}, {"id": 782471296, "item_id": 48904456, "url": "https://news.ycombinator.com/item?id=48904456", "by": "mschuetz", "date": "2026-07-14", "stance": "recommend", "quote": "Easier to write a fast software rasterizer in cuda, than to draw a single triangle in Vulkan", "excerpt": "> unimportant implementation detail that should be assessed on performance alone rather than usability I disagree. Usability is important, or you end up with garbage like Vulkan. I'd rather have 99% of the performance for 100% usability, than 100% of the performance for 5% of the usability. Unfortunately when it comes to graphics APIs, the choices are between bad and worse, which is why I switched to doing real time ", "context": "real-time rendering developer", "use_case": "real-time rendering", "switched_from": "Vulkan", "story": "Writing a bindless GPU abstraction layer"}, {"id": 734999856, "item_id": 45937491, "url": "https://news.ycombinator.com/item?id=45937491", "by": "boxerab", "date": "2025-11-15", "stance": "recommend", "quote": "A stack that can be used for EVERY card they produce, like CUDA, not just a select few.", "excerpt": "This is a great project, but the bigger question is: why isn't AMD doing this themselves? It continues to boggle my mind how much they don't seem to get the importance of a mature software stack when it is so obviously the key to the success of team red. A stack that can be used for EVERY card they produce, like CUDA, not just a select few. I used to believe that AMD the underdog would catch up some day, but I've mor", "context": "", "use_case": "GPU compute stack", "switched_from": "", "story": "AMD GPUs Go Brrr"}]}