{"updated":"2026-10-14","license":"CC BY 4.0","source":"https://onpremworld.com","docs":"https://onpremworld.com/openapi.json","count":37,"items":[{"id":"rtx-3060-12gb","vendor":"NVIDIA","name":"GeForce RTX 3060 (12 GB)","type":"gpu","memoryOptionsGB":[12],"memoryType":"GDDR6","memoryBandwidthGBs":360,"tdpW":170,"launchDate":"2021-02","launchMsrpUsd":329,"architecture":"Ampere","softwareSupport":"CUDA (compute capability 8.6, per NVIDIA spec page); llama.cpp, Ollama, vLLM, ExLlamaV2","sources":{"memoryGB":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3060-3060ti/","memoryType":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3060-3060ti/","memoryBandwidthGBs":"https://www.nvidia.com/en-gb/geforce/news/game-on-you-asked-we-answered-qa/","tdpW":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3060-3060ti/","launchDate":"https://www.nvidia.com/en-us/geforce/news/geforce-rtx-3060-out-now/","launchMsrpUsd":"https://nvidianews.nvidia.com/news/nvidia-introduces-geforce-rtx-3060-next-generation-of-the-worlds-most-popular-gpu","architecture":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3060-3060ti/","interface":"https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units"},"url":{"en":"https://onpremworld.com/en/hardware/rtx-3060-12gb","zh-tw":"https://onpremworld.com/zh-tw/hardware/rtx-3060-12gb"}},{"id":"rtx-3090","vendor":"NVIDIA","name":"GeForce RTX 3090","type":"gpu","memoryOptionsGB":[24],"memoryType":"GDDR6X","memoryBandwidthGBs":936,"tdpW":350,"launchDate":"2020-09","launchMsrpUsd":1499,"architecture":"Ampere","softwareSupport":"CUDA (compute capability 8.6, per NVIDIA spec page); llama.cpp, Ollama, vLLM, ExLlamaV2","sources":{"memoryGB":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3090-3090ti/","memoryType":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3090-3090ti/","memoryBandwidthGBs":"https://www.nvidia.com/content/PDF/nvidia-ampere-ga-102-gpu-architecture-whitepaper-v2.pdf","tdpW":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3090-3090ti/","launchDate":"https://nvidianews.nvidia.com/news/nvidia-delivers-greatest-ever-generational-leap-in-performance-with-geforce-rtx-30-series-gpus","launchMsrpUsd":"https://nvidianews.nvidia.com/news/nvidia-delivers-greatest-ever-generational-leap-in-performance-with-geforce-rtx-30-series-gpus","architecture":"https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3090-3090ti/","interface":"https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units"},"url":{"en":"https://onpremworld.com/en/hardware/rtx-3090","zh-tw":"https://onpremworld.com/zh-tw/hardware/rtx-3090"}},{"id":"rtx-4060-ti-16gb","vendor":"NVIDIA","name":"GeForce RTX 4060 Ti 16GB","type":"gpu","memoryOptionsGB":[16],"memoryType":"GDDR6","memoryBandwidthGBs":288,"tdpW":165,"launchDate":"2023-07","launchMsrpUsd":499,"architecture":"Ada Lovelace","softwareSupport":"CUDA (compute capability 8.9, per NVIDIA spec page; Ada 4th-gen Tensor Cores add FP8); llama.cpp, Ollama, vLLM, ExLlamaV2","sources":{"memoryGB":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4060-4060ti/","memoryType":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4060-4060ti/","memoryBandwidthGBs":"https://www.profesionalreview.com/2023/07/21/nvidia-rtx-4060-ti-16gb-review/","tdpW":"https://www.profesionalreview.com/2023/07/21/nvidia-rtx-4060-ti-16gb-review/","launchDate":"https://nvidianews.nvidia.com/news/geforce-rtx-4060-family-is-here-nvidias-revolutionary-ada-lovelace-architecture-comes-to-core-gamers-everywhere-starting-at-299","launchMsrpUsd":"https://nvidianews.nvidia.com/news/geforce-rtx-4060-family-is-here-nvidias-revolutionary-ada-lovelace-architecture-comes-to-core-gamers-everywhere-starting-at-299","architecture":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4060-4060ti/","interface":"https://www.profesionalreview.com/2023/07/21/nvidia-rtx-4060-ti-16gb-review/"},"url":{"en":"https://onpremworld.com/en/hardware/rtx-4060-ti-16gb","zh-tw":"https://onpremworld.com/zh-tw/hardware/rtx-4060-ti-16gb"}},{"id":"rtx-4080-super","vendor":"NVIDIA","name":"GeForce RTX 4080 SUPER","type":"gpu","memoryOptionsGB":[16],"memoryType":"GDDR6X","memoryBandwidthGBs":736,"tdpW":320,"launchDate":"2024-01","launchMsrpUsd":999,"architecture":"Ada Lovelace","softwareSupport":"CUDA (compute capability 8.9, per NVIDIA spec page; Ada 4th-gen Tensor Cores add FP8); llama.cpp, Ollama, vLLM, ExLlamaV2","sources":{"memoryGB":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4080-family/","memoryType":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4080-family/","memoryBandwidthGBs":"https://www.nvidia.com/en-us/geforce/news/geforce-rtx-4080-4070-ti-4070-super-gpu/","tdpW":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4080-family/","launchDate":"https://nvidianews.nvidia.com/news/geforce-rtx-40-super-series","launchMsrpUsd":"https://nvidianews.nvidia.com/news/geforce-rtx-40-super-series","architecture":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4080-family/","interface":"https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units"},"url":{"en":"https://onpremworld.com/en/hardware/rtx-4080-super","zh-tw":"https://onpremworld.com/zh-tw/hardware/rtx-4080-super"}},{"id":"rtx-4090","vendor":"NVIDIA","name":"GeForce RTX 4090","type":"gpu","memoryOptionsGB":[24],"memoryType":"GDDR6X","memoryBandwidthGBs":1008,"tdpW":450,"launchDate":"2022-10","launchMsrpUsd":1599,"architecture":"Ada Lovelace","softwareSupport":"CUDA (compute capability 8.9, per NVIDIA spec page; Ada 4th-gen Tensor Cores add FP8); llama.cpp, Ollama, vLLM, ExLlamaV2","sources":{"memoryGB":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4090/","memoryType":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4090/","memoryBandwidthGBs":"https://images.nvidia.com/aem-dam/Solutions/geforce/ada/nvidia-ada-gpu-architecture.pdf","tdpW":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4090/","launchDate":"https://nvidianews.nvidia.com/news/nvidia-delivers-quantum-leap-in-performance-introduces-new-era-of-neural-rendering-with-geforce-rtx-40-series","launchMsrpUsd":"https://nvidianews.nvidia.com/news/nvidia-delivers-quantum-leap-in-performance-introduces-new-era-of-neural-rendering-with-geforce-rtx-40-series","architecture":"https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4090/","interface":"https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units"},"url":{"en":"https://onpremworld.com/en/hardware/rtx-4090","zh-tw":"https://onpremworld.com/zh-tw/hardware/rtx-4090"}},{"id":"rtx-5060-ti-16gb","vendor":"NVIDIA","name":"GeForce RTX 5060 Ti 16GB","type":"gpu","memoryOptionsGB":[16],"memoryType":"GDDR7","memoryBandwidthGBs":448,"tdpW":180,"launchDate":"2025-04","launchMsrpUsd":429,"architecture":"Blackwell","softwareSupport":"CUDA (compute capability 12.0, per NVIDIA spec page; sm_120 needs CUDA 12.8+ toolkit); 5th-gen Tensor Cores with FP4; llama.cpp, Ollama, vLLM, ExLlamaV2","sources":{"memoryGB":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5060-family/","memoryType":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5060-family/","memoryBandwidthGBs":"https://www.nvidia.com/en-us/geforce/graphics-cards/compare/","tdpW":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5060-family/","launchDate":"https://nvidianews.nvidia.com/news/nvidia-blackwell-geforce-rtx-arrives-for-every-gamer-starting-at-299","launchMsrpUsd":"https://nvidianews.nvidia.com/news/nvidia-blackwell-geforce-rtx-arrives-for-every-gamer-starting-at-299","architecture":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5060-family/","interface":"https://www.profesionalreview.com/2025/04/17/rtx-5060-ti-escalado-pcie-5-0-4-0-y-3-0/"},"url":{"en":"https://onpremworld.com/en/hardware/rtx-5060-ti-16gb","zh-tw":"https://onpremworld.com/zh-tw/hardware/rtx-5060-ti-16gb"}},{"id":"rtx-5070","vendor":"NVIDIA","name":"GeForce RTX 5070","type":"gpu","memoryOptionsGB":[12],"memoryType":"GDDR7","memoryBandwidthGBs":672,"tdpW":250,"launchDate":"2025-03","launchMsrpUsd":549,"architecture":"Blackwell","softwareSupport":"CUDA (compute capability 12.0, per NVIDIA spec page; sm_120 needs CUDA 12.8+ toolkit); 5th-gen Tensor Cores with FP4; llama.cpp, Ollama, vLLM, ExLlamaV2","sources":{"memoryGB":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5070-family/","memoryType":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5070-family/","memoryBandwidthGBs":"https://www.nvidia.com/en-us/geforce/graphics-cards/compare/","tdpW":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5070-family/","launchDate":"https://www.nvidia.com/en-us/geforce/news/rtx-5070-out-now/","launchMsrpUsd":"https://nvidianews.nvidia.com/news/nvidia-blackwell-geforce-rtx-50-series-opens-new-world-of-ai-computer-graphics","architecture":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5070-family/","interface":"https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units"},"url":{"en":"https://onpremworld.com/en/hardware/rtx-5070","zh-tw":"https://onpremworld.com/zh-tw/hardware/rtx-5070"}},{"id":"rtx-5070-ti","vendor":"NVIDIA","name":"GeForce RTX 5070 Ti","type":"gpu","memoryOptionsGB":[16],"memoryType":"GDDR7","memoryBandwidthGBs":896,"tdpW":300,"launchDate":"2025-02","launchMsrpUsd":749,"architecture":"Blackwell","softwareSupport":"CUDA (compute capability 12.0, per NVIDIA spec page; sm_120 needs CUDA 12.8+ toolkit); 5th-gen Tensor Cores with FP4; llama.cpp, Ollama, vLLM, ExLlamaV2","sources":{"memoryGB":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5070-family/","memoryType":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5070-family/","memoryBandwidthGBs":"https://www.nvidia.com/en-us/geforce/graphics-cards/compare/","tdpW":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5070-family/","launchDate":"https://www.nvidia.com/en-us/geforce/news/rtx-5070-ti-out-now/","launchMsrpUsd":"https://nvidianews.nvidia.com/news/nvidia-blackwell-geforce-rtx-50-series-opens-new-world-of-ai-computer-graphics","architecture":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5070-family/","interface":"https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units"},"url":{"en":"https://onpremworld.com/en/hardware/rtx-5070-ti","zh-tw":"https://onpremworld.com/zh-tw/hardware/rtx-5070-ti"}},{"id":"rtx-5080","vendor":"NVIDIA","name":"GeForce RTX 5080","type":"gpu","memoryOptionsGB":[16],"memoryType":"GDDR7","memoryBandwidthGBs":960,"tdpW":360,"launchDate":"2025-01","launchMsrpUsd":999,"architecture":"Blackwell","softwareSupport":"CUDA (compute capability 12.0, per NVIDIA spec page; sm_120 needs CUDA 12.8+ toolkit); 5th-gen Tensor Cores with FP4; llama.cpp, Ollama, vLLM, ExLlamaV2","sources":{"memoryGB":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5080/","memoryType":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5080/","memoryBandwidthGBs":"https://www.nvidia.com/en-us/geforce/graphics-cards/compare/","tdpW":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5080/","launchDate":"https://nvidianews.nvidia.com/news/nvidia-blackwell-geforce-rtx-50-series-opens-new-world-of-ai-computer-graphics","launchMsrpUsd":"https://nvidianews.nvidia.com/news/nvidia-blackwell-geforce-rtx-50-series-opens-new-world-of-ai-computer-graphics","architecture":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5080/","interface":"https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units"},"url":{"en":"https://onpremworld.com/en/hardware/rtx-5080","zh-tw":"https://onpremworld.com/zh-tw/hardware/rtx-5080"}},{"id":"rtx-5090","vendor":"NVIDIA","name":"GeForce RTX 5090","type":"gpu","memoryOptionsGB":[32],"memoryType":"GDDR7","memoryBandwidthGBs":1792,"tdpW":575,"launchDate":"2025-01","launchMsrpUsd":1999,"architecture":"Blackwell","softwareSupport":"CUDA (compute capability 12.0, per NVIDIA spec page; sm_120 needs CUDA 12.8+ toolkit); 5th-gen Tensor Cores with FP4; llama.cpp, Ollama, vLLM, ExLlamaV2","sources":{"memoryGB":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5090/","memoryType":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5090/","memoryBandwidthGBs":"https://www.nvidia.com/en-us/geforce/graphics-cards/compare/","tdpW":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5090/","launchDate":"https://nvidianews.nvidia.com/news/nvidia-blackwell-geforce-rtx-50-series-opens-new-world-of-ai-computer-graphics","launchMsrpUsd":"https://nvidianews.nvidia.com/news/nvidia-blackwell-geforce-rtx-50-series-opens-new-world-of-ai-computer-graphics","architecture":"https://www.nvidia.com/en-us/geforce/graphics-cards/50-series/rtx-5090/","interface":"https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units"},"url":{"en":"https://onpremworld.com/en/hardware/rtx-5090","zh-tw":"https://onpremworld.com/zh-tw/hardware/rtx-5090"}},{"id":"amd-rx-7900-xt","vendor":"AMD","name":"Radeon RX 7900 XT","type":"gpu","memoryOptionsGB":[20],"memoryType":"GDDR6","memoryBandwidthGBs":800,"tdpW":315,"launchDate":"2022-12","launchMsrpUsd":899,"architecture":"RDNA 3","softwareSupport":"ROCm/HIP: listed as supported (RDNA3, gfx1100) in the ROCm 10.1.0 compatibility matrix (Linux & Windows) and HIP SDK for Windows 7.2.0; AMD extended its ROCm PyTorch support to the RX 7900 XT in Nov 2023. Vulkan.","sources":{"memoryGB":"https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xt.html","memoryType":"https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xt.html","memoryBandwidthGBs":"https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xt.html","tdpW":"https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xt.html","launchDate":"https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xt.html","launchMsrpUsd":"https://www.amd.com/en/newsroom/press-releases/2022-11-3-amd-unveils-world-s-most-advanced-gaming-graphics-.html","architecture":"https://www.amd.com/en/newsroom/press-releases/2022-11-3-amd-unveils-world-s-most-advanced-gaming-graphics-.html","interface":"https://www.custompc.com/amd-radeon-rx-7900-xt-review","softwareSupport":"https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html"},"url":{"en":"https://onpremworld.com/en/hardware/amd-rx-7900-xt","zh-tw":"https://onpremworld.com/zh-tw/hardware/amd-rx-7900-xt"}},{"id":"amd-rx-7900-xtx","vendor":"AMD","name":"Radeon RX 7900 XTX","type":"gpu","memoryOptionsGB":[24],"memoryType":"GDDR6","memoryBandwidthGBs":960,"tdpW":355,"launchDate":"2022-12","launchMsrpUsd":999,"architecture":"RDNA 3","softwareSupport":"ROCm/HIP: listed as supported (RDNA3, gfx1100) in the ROCm 10.1.0 compatibility matrix (Linux & Windows) and HIP SDK for Windows 7.2.0; first officially supported for PyTorch with ROCm 5.7 (Oct 2023). Vulkan (e.g. llama.cpp Vulkan backend).","sources":{"memoryGB":"https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xtx.html","memoryType":"https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xtx.html","memoryBandwidthGBs":"https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xtx.html","tdpW":"https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xtx.html","launchDate":"https://www.amd.com/en/products/graphics/desktops/radeon/7000-series/amd-radeon-rx-7900xtx.html","launchMsrpUsd":"https://www.amd.com/en/newsroom/press-releases/2022-11-3-amd-unveils-world-s-most-advanced-gaming-graphics-.html","architecture":"https://www.amd.com/en/newsroom/press-releases/2022-11-3-amd-unveils-world-s-most-advanced-gaming-graphics-.html","interface":"https://www.custompc.com/amd-radeon-rx-7900-xtx-review","softwareSupport":"https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html"},"url":{"en":"https://onpremworld.com/en/hardware/amd-rx-7900-xtx","zh-tw":"https://onpremworld.com/zh-tw/hardware/amd-rx-7900-xtx"}},{"id":"amd-rx-9060-xt-16gb","vendor":"AMD","name":"Radeon RX 9060 XT 16GB","type":"gpu","memoryOptionsGB":[16],"memoryType":"GDDR6","memoryBandwidthGBs":320,"tdpW":160,"launchDate":"2025-06","launchMsrpUsd":349,"architecture":"RDNA 4","softwareSupport":"ROCm/HIP: listed as supported (RDNA4, gfx1200) in the ROCm 10.1.0 compatibility matrix (Linux & Windows) and HIP SDK for Windows 7.2.0; support first added in ROCm 6.4.1 (compute workloads). Vulkan.","sources":{"memoryGB":"https://www.amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9060xt.html","memoryType":"https://www.amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9060xt.html","memoryBandwidthGBs":"https://www.amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9060xt.html","tdpW":"https://www.amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9060xt.html","launchDate":"https://www.tomshardware.com/pc-components/gpus/amd-radeon-rx-9060-xt-launches-on-june-5-starting-at-usd299","launchMsrpUsd":"https://www.amd.com/en/newsroom/press-releases/2025-5-20-amd-introduces-new-radeon-graphics-cards-and-ryzen.html","architecture":"https://www.amd.com/en/newsroom/press-releases/2025-5-20-amd-introduces-new-radeon-graphics-cards-and-ryzen.html","interface":"https://www.amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9060xt.html","softwareSupport":"https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html"},"url":{"en":"https://onpremworld.com/en/hardware/amd-rx-9060-xt-16gb","zh-tw":"https://onpremworld.com/zh-tw/hardware/amd-rx-9060-xt-16gb"}},{"id":"amd-rx-9070-xt","vendor":"AMD","name":"Radeon RX 9070 XT","type":"gpu","memoryOptionsGB":[16],"memoryType":"GDDR6","memoryBandwidthGBs":640,"tdpW":304,"launchDate":"2025-03","launchMsrpUsd":599,"architecture":"RDNA 4","softwareSupport":"ROCm/HIP: listed as supported (RDNA4, gfx1201) in the ROCm 10.1.0 compatibility matrix (Linux & Windows) and HIP SDK for Windows 7.2.0; support first added in ROCm 6.4.1 (compute workloads). Vulkan.","sources":{"memoryGB":"https://www.amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9070xt.html","memoryType":"https://www.amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9070xt.html","memoryBandwidthGBs":"https://www.amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9070xt.html","tdpW":"https://www.amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9070xt.html","launchDate":"https://www.amd.com/en/newsroom/press-releases/2025-2-28-amd-unveils-next-generation-amd-rdna-4-architectu.html","launchMsrpUsd":"https://www.amd.com/en/newsroom/press-releases/2025-2-28-amd-unveils-next-generation-amd-rdna-4-architectu.html","architecture":"https://www.amd.com/en/newsroom/press-releases/2025-2-28-amd-unveils-next-generation-amd-rdna-4-architectu.html","interface":"https://www.amd.com/en/products/graphics/desktops/radeon/9000-series/amd-radeon-rx-9070xt.html","softwareSupport":"https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html"},"url":{"en":"https://onpremworld.com/en/hardware/amd-rx-9070-xt","zh-tw":"https://onpremworld.com/zh-tw/hardware/amd-rx-9070-xt"}},{"id":"intel-arc-b580","vendor":"Intel","name":"Arc B580","type":"gpu","memoryOptionsGB":[12],"memoryType":"GDDR6","memoryBandwidthGBs":456,"tdpW":190,"launchDate":"2024-12","launchMsrpUsd":249,"architecture":"Battlemage (Xe2)","softwareSupport":"oneAPI/SYCL (llama.cpp SYCL backend lists Arc B580 as supported); upstream PyTorch XPU (PyTorch 2.14 docs list Arc B-Series/Battlemage); Vulkan. IPEX-LLM had a B580 quickstart but Intel has archived the repo (no longer maintained).","sources":{"memoryGB":"https://www.intel.com/content/www/us/en/products/sku/241598/intel-arc-b580-graphics/specifications.html","memoryType":"https://www.intel.com/content/www/us/en/products/sku/241598/intel-arc-b580-graphics/specifications.html","memoryBandwidthGBs":"https://www.intel.com/content/www/us/en/products/sku/241598/intel-arc-b580-graphics/specifications.html","tdpW":"https://www.intel.com/content/www/us/en/products/sku/241598/intel-arc-b580-graphics/specifications.html","launchDate":"https://www.intel.com/content/www/us/en/newsroom/news/intel-launches-arc-b-series-graphics-cards.html","launchMsrpUsd":"https://www.intel.com/content/www/us/en/newsroom/news/intel-launches-arc-b-series-graphics-cards.html","architecture":"https://www.intel.com/content/www/us/en/products/sku/241598/intel-arc-b580-graphics/specifications.html","interface":"https://www.intel.com/content/www/us/en/products/sku/241598/intel-arc-b580-graphics/specifications.html","softwareSupport":"https://github.com/ggml-org/llama.cpp/blob/master/docs/backend/SYCL.md"},"url":{"en":"https://onpremworld.com/en/hardware/intel-arc-b580","zh-tw":"https://onpremworld.com/zh-tw/hardware/intel-arc-b580"}},{"id":"nvidia-rtx-a6000","vendor":"NVIDIA","name":"RTX A6000","type":"workstation-gpu","memoryOptionsGB":[48],"memoryType":"GDDR6 (ECC)","memoryBandwidthGBs":768,"tdpW":300,"launchDate":"2020-12","launchMsrpUsd":4650,"architecture":"Ampere (GA102)","softwareSupport":"CUDA (Ampere); broadly supported by llama.cpp, vLLM, TensorRT-LLM, Ollama, LM Studio; NVLink bridge pairs two cards (96 GB combined).","sources":{"memoryGB":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/quadro-product-literature/proviz-print-nvidia-rtx-a6000-datasheet-us-nvidia-1454980-r9-web%20(1).pdf","memoryType":"https://www.nvidia.com/en-us/design-visualization/rtx-a6000/","memoryBandwidthGBs":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/quadro-product-literature/proviz-print-nvidia-rtx-a6000-datasheet-us-nvidia-1454980-r9-web%20(1).pdf","tdpW":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/quadro-product-literature/proviz-print-nvidia-rtx-a6000-datasheet-us-nvidia-1454980-r9-web%20(1).pdf","launchDate":"https://lowyat.net/2020/223037/nvidia-announces-rtx-a6000-and-a40-gpus-for-professionals-and-data-centre-workloads","launchMsrpUsd":"https://www.computerbase.de/2023-01/nvidia-a6000-ada-generation-6800-us-dollar/","architecture":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/quadro-product-literature/NVIDIA-RTX-Blackwell-PRO-GPU-Architecture-v1.0.pdf","interface":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/quadro-product-literature/proviz-print-nvidia-rtx-a6000-datasheet-us-nvidia-1454980-r9-web%20(1).pdf"},"url":{"en":"https://onpremworld.com/en/hardware/nvidia-rtx-a6000","zh-tw":"https://onpremworld.com/zh-tw/hardware/nvidia-rtx-a6000"}},{"id":"nvidia-rtx-6000-ada","vendor":"NVIDIA","name":"RTX 6000 Ada Generation","type":"workstation-gpu","memoryOptionsGB":[48],"memoryType":"GDDR6 (ECC)","memoryBandwidthGBs":960,"tdpW":300,"launchDate":"2022-12","launchMsrpUsd":6800,"architecture":"Ada Lovelace (AD102)","softwareSupport":"CUDA (Ada Lovelace, FP8 Tensor Cores); supported by llama.cpp, vLLM, TensorRT-LLM, Ollama, LM Studio.","sources":{"memoryGB":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/rtx-6000/proviz-print-rtx6000-datasheet-web-2504660.pdf","memoryType":"https://www.nvidia.com/en-us/design-visualization/rtx-6000/","memoryBandwidthGBs":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/rtx-6000/proviz-print-rtx6000-datasheet-web-2504660.pdf","tdpW":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/rtx-6000/proviz-print-rtx6000-datasheet-web-2504660.pdf","launchDate":"https://nvidianews.nvidia.com/news/nvidias-new-ada-lovelace-rtx-gpu-arrives-for-designers-and-creators","launchMsrpUsd":"https://www.computerbase.de/2023-01/nvidia-a6000-ada-generation-6800-us-dollar/","architecture":"https://www.nvidia.com/en-us/design-visualization/rtx-6000/","interface":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/rtx-6000/proviz-print-rtx6000-datasheet-web-2504660.pdf"},"url":{"en":"https://onpremworld.com/en/hardware/nvidia-rtx-6000-ada","zh-tw":"https://onpremworld.com/zh-tw/hardware/nvidia-rtx-6000-ada"}},{"id":"nvidia-rtx-pro-5000-blackwell-72gb","vendor":"NVIDIA","name":"RTX PRO 5000 72GB Blackwell","type":"workstation-gpu","memoryOptionsGB":[72],"memoryType":"GDDR7 (ECC)","memoryBandwidthGBs":1344,"tdpW":300,"launchDate":"2025-12","launchMsrpUsd":null,"architecture":"Blackwell","softwareSupport":"CUDA (NVIDIA datasheet lists CUDA 12.8); FP4/NVFP4 Tensor Cores; supported by llama.cpp, vLLM, TensorRT-LLM, Ollama.","sources":{"memoryGB":"https://www.nvidia.com/content/dam/en-zz/Solutions/products/workstations/professional-desktop-gpus/rtx-pro-5000-blackwell/workstation-datasheet-blackwell-rtx-pro-5000-5488550-nvidia.pdf","memoryType":"https://www.nvidia.com/en-us/products/workstations/professional-desktop-gpus/rtx-pro-5000/","memoryBandwidthGBs":"https://www.nvidia.com/content/dam/en-zz/Solutions/products/workstations/professional-desktop-gpus/rtx-pro-5000-blackwell/workstation-datasheet-blackwell-rtx-pro-5000-5488550-nvidia.pdf","tdpW":"https://www.nvidia.com/content/dam/en-zz/Solutions/products/workstations/professional-desktop-gpus/rtx-pro-5000-blackwell/workstation-datasheet-blackwell-rtx-pro-5000-5488550-nvidia.pdf","launchDate":"https://blogs.nvidia.com/blog/rtx-pro-5000-72gb-blackwell-gpu/","architecture":"https://blogs.nvidia.com/blog/rtx-pro-5000-72gb-blackwell-gpu/","interface":"https://www.nvidia.com/content/dam/en-zz/Solutions/products/workstations/professional-desktop-gpus/rtx-pro-5000-blackwell/workstation-datasheet-blackwell-rtx-pro-5000-5488550-nvidia.pdf"},"url":{"en":"https://onpremworld.com/en/hardware/nvidia-rtx-pro-5000-blackwell-72gb","zh-tw":"https://onpremworld.com/zh-tw/hardware/nvidia-rtx-pro-5000-blackwell-72gb"}},{"id":"nvidia-rtx-pro-6000-blackwell","vendor":"NVIDIA","name":"RTX PRO 6000 Blackwell Workstation Edition","type":"workstation-gpu","memoryOptionsGB":[96],"memoryType":"GDDR7 (ECC)","memoryBandwidthGBs":1792,"tdpW":600,"launchDate":"2025-04","launchMsrpUsd":8565,"architecture":"Blackwell (GB202)","softwareSupport":"CUDA (Blackwell, 5th-gen Tensor Cores with FP4); MIG up to 4 instances; supported by llama.cpp, vLLM, TensorRT-LLM, Ollama.","sources":{"memoryGB":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/quadro-product-literature/workstation-datasheet-blackwell-rtx-pro6000-x-nvidia-us-3519208-web.pdf","memoryType":"https://www.nvidia.com/en-us/products/workstations/professional-desktop-gpus/rtx-pro-6000/","memoryBandwidthGBs":"https://www.nvidia.com/en-us/products/workstations/professional-desktop-gpus/rtx-pro-6000/","tdpW":"https://www.nvidia.com/en-us/products/workstations/professional-desktop-gpus/rtx-pro-6000/","launchDate":"https://nvidianews.nvidia.com/news/nvidia-blackwell-rtx-pro-workstations-servers-agentic-ai","launchMsrpUsd":"https://www.tomshardware.com/pc-components/gpus/nvidia-raises-rtx-pro-6000-blackwell-gpu-pricing-to-usd13-250-55-percent-increase-over-msrp-in-a-years-time","architecture":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/quadro-product-literature/workstation-datasheet-blackwell-rtx-pro6000-x-nvidia-us-3519208-web.pdf","interface":"https://www.nvidia.com/content/dam/en-zz/Solutions/design-visualization/quadro-product-literature/workstation-datasheet-blackwell-rtx-pro6000-x-nvidia-us-3519208-web.pdf"},"url":{"en":"https://onpremworld.com/en/hardware/nvidia-rtx-pro-6000-blackwell","zh-tw":"https://onpremworld.com/zh-tw/hardware/nvidia-rtx-pro-6000-blackwell"}},{"id":"amd-radeon-ai-pro-r9700","vendor":"AMD","name":"Radeon AI PRO R9700","type":"workstation-gpu","memoryOptionsGB":[32],"memoryType":"GDDR6 (ECC on Linux only)","memoryBandwidthGBs":640,"tdpW":300,"launchDate":"2025-07","launchMsrpUsd":1299,"architecture":"RDNA 4","softwareSupport":"ROCm (AMD cites ROCm 6.4.1 for R9700) with PyTorch/ONNX Runtime; Vulkan llama.cpp (AMD's own LM Studio benchmarks use Vulkan llama.cpp); Windows and Linux.","sources":{"memoryGB":"https://www.amd.com/en/products/graphics/workstations/radeon-ai-pro/ai-9000-series/amd-radeon-ai-pro-r9700.html","memoryType":"https://www.amd.com/en/products/graphics/workstations/radeon-ai-pro/ai-9000-series/amd-radeon-ai-pro-r9700.html","memoryBandwidthGBs":"https://www.amd.com/en/products/graphics/workstations/radeon-ai-pro/ai-9000-series/amd-radeon-ai-pro-r9700.html","tdpW":"https://www.amd.com/en/products/graphics/workstations/radeon-ai-pro/ai-9000-series/amd-radeon-ai-pro-r9700.html","launchDate":"https://www.amd.com/en/blogs/2025/amd-radeon-ai-pro-r9700-to-be-available-in-workstation.html","launchMsrpUsd":"https://www.amd.com/en/products/graphics/workstations/radeon-ai-pro.html","architecture":"https://www.amd.com/en/products/graphics/workstations/radeon-ai-pro/ai-9000-series/amd-radeon-ai-pro-r9700.html","interface":"https://www.amd.com/en/products/graphics/workstations/radeon-ai-pro/ai-9000-series/amd-radeon-ai-pro-r9700.html"},"url":{"en":"https://onpremworld.com/en/hardware/amd-radeon-ai-pro-r9700","zh-tw":"https://onpremworld.com/zh-tw/hardware/amd-radeon-ai-pro-r9700"}},{"id":"intel-arc-pro-b60","vendor":"Intel","name":"Arc Pro B60","type":"workstation-gpu","memoryOptionsGB":[24],"memoryType":"GDDR6","memoryBandwidthGBs":456,"tdpW":200,"launchDate":"2025-06","launchMsrpUsd":null,"architecture":"Xe2 (Battlemage)","softwareSupport":"oneAPI/SYCL and OpenVINO (Intel spec page); llama.cpp SYCL backend lists Arc B-series and Arc Pro B60; Vulkan 1.3; Intel LLM Scaler vLLM containers target Arc Pro B60/B70 (github.com/intel/llm-scaler); note IPEX-LLM repo was archived in Jan 2026.","sources":{"memoryGB":"https://www.intel.com/content/www/us/en/products/sku/243916/intel-arc-pro-b60-graphics/specifications.html","memoryType":"https://www.intel.com/content/www/us/en/products/sku/243916/intel-arc-pro-b60-graphics/specifications.html","memoryBandwidthGBs":"https://www.intel.com/content/www/us/en/products/sku/243916/intel-arc-pro-b60-graphics/specifications.html","tdpW":"https://www.intel.com/content/www/us/en/products/sku/243916/intel-arc-pro-b60-graphics/specifications.html","launchDate":"https://www.silicon.co.uk/press-release/computex-2025-intel-unveils-new-gpus-for-ai-and-workstations","architecture":"https://www.intel.com/content/www/us/en/products/sku/243916/intel-arc-pro-b60-graphics/specifications.html","interface":"https://download.intel.com/newsroom/2025/client-computing/Intel-Arc-Pro-B60-Data-Sheet.pdf"},"url":{"en":"https://onpremworld.com/en/hardware/intel-arc-pro-b60","zh-tw":"https://onpremworld.com/zh-tw/hardware/intel-arc-pro-b60"}},{"id":"intel-arc-pro-b70","vendor":"Intel","name":"Arc Pro B70","type":"workstation-gpu","memoryOptionsGB":[32],"memoryType":"GDDR6 (ECC)","memoryBandwidthGBs":608,"tdpW":230,"launchDate":"2026-03","launchMsrpUsd":949,"architecture":"Xe2 (Battlemage)","softwareSupport":"oneAPI/SYCL and OpenVINO (Intel spec page); llama.cpp SYCL backend lists Arc B-series and Arc Pro B60; Vulkan 1.3; Intel LLM Scaler vLLM containers target Arc Pro B60/B70 (github.com/intel/llm-scaler); note IPEX-LLM repo was archived in Jan 2026.","sources":{"memoryGB":"https://www.intel.com/content/www/us/en/products/sku/245797/intel-arc-pro-b70-graphics/specifications.html","memoryType":"https://www.intel.com/content/www/us/en/products/sku/245797/intel-arc-pro-b70-graphics/specifications.html","memoryBandwidthGBs":"https://www.intel.com/content/www/us/en/products/sku/245797/intel-arc-pro-b70-graphics/specifications.html","tdpW":"https://www.intel.com/content/www/us/en/products/sku/245797/intel-arc-pro-b70-graphics/specifications.html","launchDate":"https://hothardware.com/news/intel-arc-pro-b70-b65-cards-debut","launchMsrpUsd":"https://hothardware.com/news/intel-arc-pro-b70-b65-cards-debut","architecture":"https://www.intel.com/content/www/us/en/products/sku/245797/intel-arc-pro-b70-graphics/specifications.html","interface":"https://www.intel.com/content/www/us/en/products/sku/245797/intel-arc-pro-b70-graphics/specifications.html"},"url":{"en":"https://onpremworld.com/en/hardware/intel-arc-pro-b70","zh-tw":"https://onpremworld.com/zh-tw/hardware/intel-arc-pro-b70"}},{"id":"apple-m4-pro","vendor":"Apple","name":"M4 Pro","type":"apple-silicon","memoryOptionsGB":[24,48,64],"memoryType":"LPDDR5X","memoryBandwidthGBs":273,"tdpW":null,"launchDate":"2024-11","launchMsrpUsd":1399,"architecture":"Apple M4 family (second-generation 3 nm; TSMC N3E)","softwareSupport":"Metal (MPS) / MLX / llama.cpp Metal backend; Core ML","sources":{"memoryGB":"https://support.apple.com/en-us/121555","memoryType":"https://en.wikipedia.org/wiki/Apple_M4","memoryBandwidthGBs":"https://support.apple.com/en-us/121555","launchDate":"https://www.apple.com/newsroom/2024/10/apples-new-mac-mini-is-more-mighty-more-mini-and-built-for-apple-intelligence/","launchMsrpUsd":"https://www.apple.com/newsroom/2024/10/apples-new-mac-mini-is-more-mighty-more-mini-and-built-for-apple-intelligence/","architecture":"https://www.apple.com/newsroom/2024/10/apple-introduces-m4-pro-and-m4-max/","interface":"https://support.apple.com/en-us/121555"},"url":{"en":"https://onpremworld.com/en/hardware/apple-m4-pro","zh-tw":"https://onpremworld.com/zh-tw/hardware/apple-m4-pro"}},{"id":"apple-m4-max-32c-gpu","vendor":"Apple","name":"M4 Max (14-core CPU / 32-core GPU)","type":"apple-silicon","memoryOptionsGB":[36],"memoryType":"LPDDR5X","memoryBandwidthGBs":410,"tdpW":null,"launchDate":"2024-11","launchMsrpUsd":3199,"architecture":"Apple M4 family (second-generation 3 nm; TSMC N3E)","softwareSupport":"Metal (MPS) / MLX / llama.cpp Metal backend; Core ML","sources":{"memoryGB":"https://support.apple.com/en-us/121553","memoryType":"https://en.wikipedia.org/wiki/Apple_M4","memoryBandwidthGBs":"https://support.apple.com/en-us/121553","launchDate":"https://www.apple.com/newsroom/2024/10/new-macbook-pro-features-m4-family-of-chips-and-apple-intelligence/","launchMsrpUsd":"https://web.archive.org/web/20241121061833/https://www.apple.com/shop/buy-mac/macbook-pro/14-inch-m4-max","architecture":"https://www.apple.com/newsroom/2024/10/apple-introduces-m4-pro-and-m4-max/","interface":"https://support.apple.com/en-us/121553"},"url":{"en":"https://onpremworld.com/en/hardware/apple-m4-max-32c-gpu","zh-tw":"https://onpremworld.com/zh-tw/hardware/apple-m4-max-32c-gpu"}},{"id":"apple-m4-max-40c-gpu","vendor":"Apple","name":"M4 Max (16-core CPU / 40-core GPU)","type":"apple-silicon","memoryOptionsGB":[48,64,128],"memoryType":"LPDDR5X","memoryBandwidthGBs":546,"tdpW":null,"launchDate":"2024-11","launchMsrpUsd":3699,"architecture":"Apple M4 family (second-generation 3 nm; TSMC N3E)","softwareSupport":"Metal (MPS) / MLX / llama.cpp Metal backend; Core ML","sources":{"memoryGB":"https://support.apple.com/en-us/121553","memoryType":"https://en.wikipedia.org/wiki/Apple_M4","memoryBandwidthGBs":"https://support.apple.com/en-us/121553","launchDate":"https://www.apple.com/newsroom/2024/10/new-macbook-pro-features-m4-family-of-chips-and-apple-intelligence/","launchMsrpUsd":"https://web.archive.org/web/20250107193915/https://www.apple.com/shop/buy-mac/macbook-pro/14-inch-space-black-standard-display-apple-m4-max-with-14-core-cpu-and-32-core-gpu-36gb-memory-1tb","architecture":"https://www.apple.com/newsroom/2024/10/apple-introduces-m4-pro-and-m4-max/","interface":"https://support.apple.com/en-us/121553"},"url":{"en":"https://onpremworld.com/en/hardware/apple-m4-max-40c-gpu","zh-tw":"https://onpremworld.com/zh-tw/hardware/apple-m4-max-40c-gpu"}},{"id":"apple-m3-ultra","vendor":"Apple","name":"M3 Ultra","type":"apple-silicon","memoryOptionsGB":[96,256,512],"memoryType":"LPDDR5","memoryBandwidthGBs":819,"tdpW":null,"launchDate":"2025-03","launchMsrpUsd":3999,"architecture":"Apple M3 family (3 nm; TSMC N3B), two M3 Max dies joined by UltraFusion","softwareSupport":"Metal (MPS) / MLX / llama.cpp Metal backend; Core ML","sources":{"memoryGB":"https://www.apple.com/newsroom/2025/03/apple-reveals-m3-ultra-taking-apple-silicon-to-a-new-extreme/","memoryType":"https://en.wikipedia.org/wiki/Apple_M3","memoryBandwidthGBs":"https://support.apple.com/en-us/122211","launchDate":"https://www.apple.com/newsroom/2025/03/apple-unveils-new-mac-studio-the-most-powerful-mac-ever/","launchMsrpUsd":"https://web.archive.org/web/20250312162945/https://www.apple.com/shop/buy-mac/mac-studio","architecture":"https://www.apple.com/newsroom/2025/03/apple-reveals-m3-ultra-taking-apple-silicon-to-a-new-extreme/","interface":"https://support.apple.com/en-us/122211"},"url":{"en":"https://onpremworld.com/en/hardware/apple-m3-ultra","zh-tw":"https://onpremworld.com/zh-tw/hardware/apple-m3-ultra"}},{"id":"apple-m5","vendor":"Apple","name":"M5","type":"apple-silicon","memoryOptionsGB":[16,24,32],"memoryType":"LPDDR5X","memoryBandwidthGBs":153,"tdpW":null,"launchDate":"2025-10","launchMsrpUsd":1599,"architecture":"Apple M5 family (third-generation 3 nm; TSMC N3P)","softwareSupport":"Metal (MPS) / MLX / llama.cpp Metal backend; Core ML; M5 GPU has a Neural Accelerator in each core","sources":{"memoryGB":"https://www.apple.com/macbook-pro/specs/","memoryType":"https://en.wikipedia.org/wiki/Apple_M5","memoryBandwidthGBs":"https://www.apple.com/newsroom/2025/10/apple-unleashes-m5-the-next-big-leap-in-ai-performance-for-apple-silicon/","launchDate":"https://www.apple.com/newsroom/2025/10/new-ipad-pro-14-inch-macbook-pro-and-apple-vision-pro-now-available/","launchMsrpUsd":"https://www.apple.com/newsroom/2025/10/apple-unveils-new-14-inch-macbook-pro-powered-by-the-m5-chip/","architecture":"https://www.apple.com/newsroom/2025/10/apple-unleashes-m5-the-next-big-leap-in-ai-performance-for-apple-silicon/","interface":"https://www.apple.com/macbook-pro/specs/"},"url":{"en":"https://onpremworld.com/en/hardware/apple-m5","zh-tw":"https://onpremworld.com/zh-tw/hardware/apple-m5"}},{"id":"apple-m5-pro","vendor":"Apple","name":"M5 Pro","type":"apple-silicon","memoryOptionsGB":[24,48,64],"memoryType":"LPDDR5X","memoryBandwidthGBs":307,"tdpW":null,"launchDate":"2026-03","launchMsrpUsd":2199,"architecture":"Apple M5 Pro (Fusion Architecture: two third-generation 3 nm dies)","softwareSupport":"Metal (MPS) / MLX / llama.cpp Metal backend; Core ML; GPU Neural Accelerator in each core","sources":{"memoryGB":"https://www.apple.com/macbook-pro/specs/","memoryType":"https://en.wikipedia.org/wiki/Apple_M5","memoryBandwidthGBs":"https://www.apple.com/newsroom/2026/03/apple-debuts-m5-pro-and-m5-max-to-supercharge-the-most-demanding-pro-workflows/","launchDate":"https://www.apple.com/newsroom/2026/03/macbook-neo-iphone-17e-ipad-air-with-m4-and-more-are-now-available/","launchMsrpUsd":"https://www.apple.com/newsroom/2026/03/apple-introduces-macbook-pro-with-all-new-m5-pro-and-m5-max/","architecture":"https://www.apple.com/newsroom/2026/03/apple-debuts-m5-pro-and-m5-max-to-supercharge-the-most-demanding-pro-workflows/","interface":"https://www.apple.com/macbook-pro/specs/"},"url":{"en":"https://onpremworld.com/en/hardware/apple-m5-pro","zh-tw":"https://onpremworld.com/zh-tw/hardware/apple-m5-pro"}},{"id":"apple-m5-max-32c-gpu","vendor":"Apple","name":"M5 Max (18-core CPU / 32-core GPU)","type":"apple-silicon","memoryOptionsGB":[36],"memoryType":"LPDDR5X","memoryBandwidthGBs":460,"tdpW":null,"launchDate":"2026-03","launchMsrpUsd":3599,"architecture":"Apple M5 Max (Fusion Architecture: two third-generation 3 nm dies)","softwareSupport":"Metal (MPS) / MLX / llama.cpp Metal backend; Core ML; GPU Neural Accelerator in each core","sources":{"memoryGB":"https://www.apple.com/macbook-pro/specs/","memoryType":"https://www.tomshardware.com/pc-components/cpus/apples-18-core-m5-max-destroys-96-core-ryzen-threadripper-pro-9995wx-in-geekbench-gpu-performance-is-much-less-impressive","memoryBandwidthGBs":"https://www.apple.com/macbook-pro/specs/","launchDate":"https://www.apple.com/newsroom/2026/03/macbook-neo-iphone-17e-ipad-air-with-m4-and-more-are-now-available/","launchMsrpUsd":"https://www.apple.com/newsroom/2026/03/apple-introduces-macbook-pro-with-all-new-m5-pro-and-m5-max/","architecture":"https://www.apple.com/newsroom/2026/03/apple-debuts-m5-pro-and-m5-max-to-supercharge-the-most-demanding-pro-workflows/","interface":"https://www.apple.com/macbook-pro/specs/"},"url":{"en":"https://onpremworld.com/en/hardware/apple-m5-max-32c-gpu","zh-tw":"https://onpremworld.com/zh-tw/hardware/apple-m5-max-32c-gpu"}},{"id":"apple-m5-max-40c-gpu","vendor":"Apple","name":"M5 Max (18-core CPU / 40-core GPU)","type":"apple-silicon","memoryOptionsGB":[48,64,128],"memoryType":"LPDDR5X","memoryBandwidthGBs":614,"tdpW":null,"launchDate":"2026-03","launchMsrpUsd":4099,"architecture":"Apple M5 Max (Fusion Architecture: two third-generation 3 nm dies)","softwareSupport":"Metal (MPS) / MLX / llama.cpp Metal backend; Core ML; GPU Neural Accelerator in each core","sources":{"memoryGB":"https://www.apple.com/macbook-pro/specs/","memoryType":"https://www.tomshardware.com/pc-components/cpus/apples-18-core-m5-max-destroys-96-core-ryzen-threadripper-pro-9995wx-in-geekbench-gpu-performance-is-much-less-impressive","memoryBandwidthGBs":"https://www.apple.com/macbook-pro/specs/","launchDate":"https://www.apple.com/newsroom/2026/03/macbook-neo-iphone-17e-ipad-air-with-m4-and-more-are-now-available/","launchMsrpUsd":"https://web.archive.org/web/20260315010528/https://www.apple.com/shop/buy-mac/macbook-pro","architecture":"https://www.apple.com/newsroom/2026/03/apple-debuts-m5-pro-and-m5-max-to-supercharge-the-most-demanding-pro-workflows/","interface":"https://www.apple.com/macbook-pro/specs/"},"url":{"en":"https://onpremworld.com/en/hardware/apple-m5-max-40c-gpu","zh-tw":"https://onpremworld.com/zh-tw/hardware/apple-m5-max-40c-gpu"}},{"id":"apple-m5-ultra","vendor":"Apple","name":"M5 Ultra","type":"apple-silicon","memoryOptionsGB":[96,256,512],"memoryType":"LPDDR5X","memoryBandwidthGBs":1200,"tdpW":null,"launchDate":"2026-09","launchMsrpUsd":5499,"architecture":"Apple M5 Ultra (quad-die: two dual-die M5 Max joined by UltraFusion; 3 nm)","softwareSupport":"Metal (MPS) / MLX / llama.cpp Metal backend; Core ML; Core AI framework; GPU Neural Accelerators; Thunderbolt 5 RDMA clustering","sources":{"memoryGB":"https://www.apple.com/mac-studio/specs/","memoryType":"https://en.wikipedia.org/wiki/Apple_M5","memoryBandwidthGBs":"https://www.apple.com/mac-studio/specs/","launchDate":"https://www.apple.com/newsroom/2026/09/the-new-mac-mini-and-mac-studio-are-available-today/","launchMsrpUsd":"https://www.apple.com/newsroom/2026/08/apple-introduces-new-mac-studio-with-m5-max-and-m5-ultra/","architecture":"https://www.apple.com/newsroom/2026/08/apple-introduces-m6-and-m5-ultra-for-a-big-leap-in-performance-and-ai-compute/","interface":"https://www.apple.com/mac-studio/specs/"},"url":{"en":"https://onpremworld.com/en/hardware/apple-m5-ultra","zh-tw":"https://onpremworld.com/zh-tw/hardware/apple-m5-ultra"}},{"id":"apple-m6-16gb","vendor":"Apple","name":"M6 (16GB configuration, 153 GB/s)","type":"apple-silicon","memoryOptionsGB":[16],"memoryType":null,"memoryBandwidthGBs":153,"tdpW":null,"launchDate":"2026-09","launchMsrpUsd":899,"architecture":"Apple M6 (2 nm)","softwareSupport":"Metal (MPS) / MLX / llama.cpp Metal backend; Core ML; GPU Neural Accelerators; Dual 16-core Neural Engine","sources":{"memoryGB":"https://www.apple.com/mac-mini/specs/","memoryBandwidthGBs":"https://www.apple.com/mac-mini/specs/","launchDate":"https://www.apple.com/newsroom/2026/09/the-new-mac-mini-and-mac-studio-are-available-today/","launchMsrpUsd":"https://www.apple.com/newsroom/2026/08/apple-unveils-a-more-powerful-mac-mini-featuring-the-all-new-m6-and-m5-pro/","architecture":"https://www.apple.com/newsroom/2026/08/apple-introduces-m6-and-m5-ultra-for-a-big-leap-in-performance-and-ai-compute/","interface":"https://www.apple.com/mac-mini/specs/"},"url":{"en":"https://onpremworld.com/en/hardware/apple-m6-16gb","zh-tw":"https://onpremworld.com/zh-tw/hardware/apple-m6-16gb"}},{"id":"apple-m6-24-32gb","vendor":"Apple","name":"M6 (24GB/32GB configurations, 170 GB/s)","type":"apple-silicon","memoryOptionsGB":[24,32],"memoryType":null,"memoryBandwidthGBs":170,"tdpW":null,"launchDate":"2026-09","launchMsrpUsd":null,"architecture":"Apple M6 (2 nm)","softwareSupport":"Metal (MPS) / MLX / llama.cpp Metal backend; Core ML; GPU Neural Accelerators; Dual 16-core Neural Engine","sources":{"memoryGB":"https://www.apple.com/mac-mini/specs/","memoryBandwidthGBs":"https://www.apple.com/mac-mini/specs/","launchDate":"https://www.apple.com/newsroom/2026/09/the-new-mac-mini-and-mac-studio-are-available-today/","architecture":"https://www.apple.com/newsroom/2026/08/apple-introduces-m6-and-m5-ultra-for-a-big-leap-in-performance-and-ai-compute/","interface":"https://www.apple.com/mac-mini/specs/"},"url":{"en":"https://onpremworld.com/en/hardware/apple-m6-24-32gb","zh-tw":"https://onpremworld.com/zh-tw/hardware/apple-m6-24-32gb"}},{"id":"nvidia-dgx-spark","vendor":"NVIDIA","name":"DGX Spark","type":"unified-memory-pc","memoryOptionsGB":[64,128],"memoryType":"LPDDR5X","memoryBandwidthGBs":273,"tdpW":140,"launchDate":"2025-10","launchMsrpUsd":3999,"architecture":"Grace Blackwell (GB10: Blackwell GPU + 20-core Arm CPU)","softwareSupport":"CUDA on NVIDIA DGX OS (Arm64 Linux); NVIDIA DGX Spark playbooks (e.g. llama.cpp, NVFP4, speculative decoding across two Sparks); ConnectX-7 (200 Gb/s) for clustering up to four units.","sources":{"memoryGB":"https://www.nvidia.com/en-us/products/workstations/dgx-spark/","memoryType":"https://docs.nvidia.com/dgx/dgx-spark/hardware.html","memoryBandwidthGBs":"https://docs.nvidia.com/dgx/dgx-spark/hardware.html","tdpW":"https://docs.nvidia.com/dgx/dgx-spark/hardware.html","launchDate":"https://nvidianews.nvidia.com/news/nvidia-dgx-spark-arrives-for-worlds-ai-developers","launchMsrpUsd":"https://forums.developer.nvidia.com/t/2-23-2026-price-change-announcement/361713","architecture":"https://www.nvidia.com/en-us/products/workstations/dgx-spark/","interface":"https://www.nvidia.com/en-us/products/workstations/dgx-spark/"},"url":{"en":"https://onpremworld.com/en/hardware/nvidia-dgx-spark","zh-tw":"https://onpremworld.com/zh-tw/hardware/nvidia-dgx-spark"}},{"id":"nvidia-rtx-spark-n1x","vendor":"NVIDIA","name":"RTX Spark (N1X) laptops & compact desktops","type":"unified-memory-pc","memoryOptionsGB":[64,128],"memoryType":"LPDDR5X","memoryBandwidthGBs":300,"tdpW":80,"launchDate":"2026-10","launchMsrpUsd":null,"architecture":"Blackwell RTX GPU (up to 6,144 cores) + Grace Arm CPU (up to 20 cores)","softwareSupport":"CUDA runs natively on Windows 11 on Arm; NVIDIA says llama.cpp, vLLM, Ollama, Unsloth and LM Studio run at launch, plus PyTorch-CUDA for Windows on Arm and TensorRT/NVFP4.","sources":{"memoryGB":"https://www.nvidia.com/en-us/products/rtx-spark/","memoryType":"https://www.nvidia.com/en-us/products/rtx-spark/","memoryBandwidthGBs":"https://docs.nvidia.com/rtx-spark/rtx-spark-porting-guide/0.1.0/overview.html","tdpW":"https://www.nvidia.com/en-us/products/rtx-spark/","launchDate":"https://blogs.nvidia.com/blog/local-ai-rtx-spark-laptops-software-updates/","architecture":"https://www.nvidia.com/en-us/products/rtx-spark/","interface":"https://www.nvidia.com/en-us/products/rtx-spark/"},"url":{"en":"https://onpremworld.com/en/hardware/nvidia-rtx-spark-n1x","zh-tw":"https://onpremworld.com/zh-tw/hardware/nvidia-rtx-spark-n1x"}},{"id":"amd-ryzen-ai-max-plus-395","vendor":"AMD","name":"Ryzen AI Max+ 395 (Strix Halo) systems","type":"unified-memory-pc","memoryOptionsGB":[32,64,128],"memoryType":"LPDDR5X","memoryBandwidthGBs":256,"tdpW":120,"launchDate":"2025-03","launchMsrpUsd":1999,"architecture":"Zen 5 (16 cores) + RDNA 3.5 (Radeon 8060S, 40 CUs) + XDNA 2 NPU","softwareSupport":"ROCm (gfx1151 listed in AMD's Radeon/Ryzen ROCm 7.2.1 Linux and Windows support matrix; earliest supporting version not verified); Vulkan llama.cpp (used in AMD's own LM Studio testing); Windows and Linux.","sources":{"memoryGB":"https://www.amd.com/en/blogs/2025/amd-ryzen-ai-max-395-processor-breakthrough-ai-.html","memoryType":"https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-plus-395.html","memoryBandwidthGBs":"https://www.amd.com/en/blogs/2025/amd-ryzen-ai-max-395-processor-breakthrough-ai-.html","tdpW":"https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-plus-395.html","launchDate":"https://www.amd.com/en/blogs/2025/amd-ryzen-ai-max-395-processor-breakthrough-ai-.html","launchMsrpUsd":"https://frame.work/blog/introducing-the-framework-desktop","architecture":"https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-plus-395.html","interface":"https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-plus-395.html"},"url":{"en":"https://onpremworld.com/en/hardware/amd-ryzen-ai-max-plus-395","zh-tw":"https://onpremworld.com/zh-tw/hardware/amd-ryzen-ai-max-plus-395"}},{"id":"amd-ryzen-ai-max-plus-pro-495","vendor":"AMD","name":"Ryzen AI Max+ PRO 495 (Gorgon Halo) systems","type":"unified-memory-pc","memoryOptionsGB":[192],"memoryType":"LPDDR5X","memoryBandwidthGBs":273,"tdpW":120,"launchDate":"2026-10","launchMsrpUsd":null,"architecture":"Zen 5 (16 cores) + RDNA 3.5 (Radeon 8065S, 40 CUs) + XDNA 2 NPU","softwareSupport":"ROCm (AMD Ryzen AI Halo ships with a validated ROCm 'Best-Known Configuration'); AMD cites PyTorch, vLLM, llama.cpp, Ollama, ComfyUI and LM Studio support on Ryzen AI Halo; Windows and Linux.","sources":{"memoryGB":"https://www.amd.com/en/products/processors/desktops/ryzen/ryzen-ai-halo/ryzen-ai-max-plus-pro-495.html","memoryType":"https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-plus-pro-495.html","memoryBandwidthGBs":"https://www.amd.com/en/products/processors/desktops/ryzen/ryzen-ai-halo/ryzen-ai-max-plus-pro-495.html","tdpW":"https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-plus-pro-495.html","launchDate":"https://www.amd.com/en/blogs/2026/meet-the-next-amd-ryzen-ai-halo.html","architecture":"https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-plus-pro-495.html","interface":"https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-plus-pro-495.html"},"url":{"en":"https://onpremworld.com/en/hardware/amd-ryzen-ai-max-plus-pro-495","zh-tw":"https://onpremworld.com/zh-tw/hardware/amd-ryzen-ai-max-plus-pro-495"}}]}