阿里巴巴人工智能实验室 AiILabs & TVM PART 1 : ARM32 CPU CONTENT PART 2 : HIFI4 DSP PART 3 : _ PowervVR GPU [和| Alibaba AL.Labs 阿里巴巴人工智能实验室 ARM 32 CPU Resolution Quantization Orize Kernel ALIOS TVM Alibaba HIFI4 DSP HIFI4 DSP [和| Alibaba AL.Labs 阿里巴巴人工智能实验室 PowerVR GPU Alibaba Al.Labs 阿里巴巴人工智能实验室 PowerVR support by TVM NNVM Compiler -Execution graph -Model layers functions

MiniMax 等。  其次是由 TogetherAI、Groq、Fireworks、Replicate、硅基流动等组成的 GPU 推理集群 服务提供商，它们处理扩展与缩减等技术难题，并在基本计算费用基础上收取额外费用，从 而让应用公司无需承担构建和管理 GPU 推理集群的高昂成本，而是可以直接利用抽象化的 AI 基础设施服务。  第三类是传统的云计算平台，例如亚马逊的 Amazon Vertex AI 等，允许应用开发者轻松部署和使用标准化或定制化的 AI 模型， 并通过 API 接口调用这些模型。  最后一类是本地推理，SGLang、vLLM、TensorRT-LLM 在生产级 GPU 服务负载中表现 出色，受到许多有本地托管模型需求的应用开发者的欢迎，此外，Ollama 和 LM Studio 也 是在个人计算机上运行模型的优选方案。 62 / 111 除模型层面外，应 软件，例如：微控制处理器（MCU）会运行实时操作系统或者直接运行某个特定程序；中央处 理器（CPU）往往会运行 Windows、Linux 等复杂操作系统作为底座支撑整个软件栈；图形 处理器（GPU）一般不加载操作系统而是直接运行图形图像处理程序，神经网络处理器 （NPU）则直接运行深度学习相关程序。 处理器芯片设计是一项很复杂的任务，整个过程犹如一座冰山。冰山水面上是用户或者大 众看到

the GE, and over 400 bug fixes. 2.72 -- October 2014: Cycles gets volume and SSS support on the GPU, pie menus are added and tooltips greatly improved, the Intersection modeling tool is added, new Sun Automatic Automatically use GLSL which runs on the GPU for performance but falls back to the CPU for large images which might be slow when loaded with the GPU. 2D Texture Uses CPU for display transform and images. Cycles can use either the CPU or certain GPUs to render images, for more information see the GPU Rendering page. None When set to None or when the only option is None: the CPU will be used as the

to the GE, and over 400 bug fixes. 2.72 – October 2014: Cycles gets volume and SSS support on the GPU, pie menus are added and tooltips greatly improved, the Intersection modeling tool is added, new Sun Automatic Automatically use GLSL which runs on the GPU for performance but falls back to the CPU for large images which might be slow when loaded with the GPU. 2D Texture Uses CPU for display transform and images. Cycles can use either the CPU or certain GPUs to render images, for more information see the GPU Rendering page. None When set to None or when the only option is None: the CPU will be used as the

the GE, and over 400 bug fixes. 2.72 -- October 2014: Cycles gets volume and SSS support on the GPU, pie menus are added and tooltips greatly improved, the Intersection modeling tool is added, new Sun supported: Automatically use GLSL which runs on the GPU for performance but falls back to the CPU for large images which might be slow when loaded with the GPU. Uses CPU for display transform and render images images. Cycles can use either the CPU or certain GPUs to render images, for more information see the GPU Rendering page. 無 When set to None or when the only option is None: the CPU will be used as the computing

to the GE, and over 400 bug fixes. 2.72 – October 2014: Cycles gets volume and SSS support on the GPU, pie menus are added and tooltips greatly improved, the Intersection modeling tool is added, new Sun supported: Automatically use GLSL which runs on the GPU for performance but falls back to the CPU for large images which might be slow when loaded with the GPU. Uses CPU for display transform and render images images. Cycles can use either the CPU or certain GPUs to render images, for more information see the GPU Rendering page. None When set to None or when the only option is None: the CPU will be used as the

数 Windows 配置要求 Mac 配置要求 适⽤场景 1.5B - RAM: 4GB - GPU: 集成显卡/现代CPU - 存储: 5GB - 内存: 8GB （M1/M2/M3） - 存储: 5GB 简单⽂本⽣成、基础代 码补全 7B - RAM: 8-10GB - GPU: GTX 1680（4-bit量 化） - 存储: 8GB - 内存: 16GB（M2 Pro/M3） Pro/M3） - 存储: 8GB 中等复杂度问答、代码 调试 14B - RAM: 24GB - GPU: RTX 3090（24GB VRAM） - 存储: 20GB - 内存: 32GB（M3 Max） - 存储: 20GB 复杂推理、技术⽂档⽣ 成 32B+ 企业级部署（需多卡并联） 暂不⽀持 科研计算、⼤规模数据 处理 2. 算⼒需求分析 模型 参数规 模 2*XE9680（16*H20 GPU） DeepSeek-R1-Distill- 70B 70B BF16 ≥180GB 4*L20 或 2*H20 GPU 三、国产芯⽚与硬件适配⽅案 1. 国内⽣态合作伙伴动态 企业 适配内容 性能对标（vs NVIDIA） 华为昇 腾 昇腾910B原⽣⽀持R1全系列，提供端到端推理优化 ⽅案 等效A100（FP16） 沐曦 GPU MXN系列⽀持70B模型BF16推理，显存利⽤率提升

动态绘画 ⼒场 碰撞 烘焙物理模拟 模拟节点 渲染 简介 EEVEE Cycles ⼯作台 摄像机 灯光 材质 着⾊节点 ⾊彩管理 Freestyle 层 & 通道 渲染输出 合成 简介 侧栏 GPU合成器 节点类型 运动追踪 & 遮罩 运动追踪 遮罩 视频编辑 简介 设置你的项⽬ 编辑您的项⽬ 资产、⽂件 & 数据系统 简介 Blender ⽂件 数据块 ⾃定义属性 关联库 资产库 变形运动模糊和Cycles⽀持⽕/烟，UI弹出菜单可拖动，雕刻模式 性能优化，新的动画插值类型，⼤量游戏引擎改进，和超过400处 bug修复。 2.72 -- 2014年10⽉: Cycles⽀持GPU渲染体积和SSS，添加饼菜单并⼤幅改进⼯具提 ⽰，添加相交建模⼯具，合成器新增Sun Beam节点，Freestyle可 在Cycles渲染器下⼯作，纹理绘制流程改进，和超过220处bug修 复。 for AMD on linux. 3.1 -- 2022年3⽉ Major point clouds improvements, Cycles Apple Metal GPU support, Subdivision GPU support, image editor handles larger images, Major performance gains for geometry nodes

Nodes Color Management Freestyle Layers & Passes Render Output Compositing Introduction Sidebar GPU Compositor Node Types Motion Tracking & Masking Introduction Motion Tracking Masking Video Editing to the GE, and over 400 bug fixes. 2.72 – October 2014: Cycles gets volume and SSS support on the GPU, pie menus are added and tooltips greatly improved, the Intersection modeling tool is added, new Sun for AMD on linux. 3.1 – March 2022 Major point clouds improvements, Cycles Apple Metal GPU support, Subdivision GPU support, image editor handles larger images, Major performance gains for geometry nodes

Nodes Color Management Freestyle Layers & Passes Render Output Compositing Introduction Sidebar GPU Compositor Node Types Motion Tracking & Masking Introduction Motion Tracking Masking Video Editing to the GE, and over 400 bug fixes. 2.72 – October 2014: Cycles gets volume and SSS support on the GPU, pie menus are added and tooltips greatly improved, the Intersection modeling tool is added, new Sun for AMD on linux. 3.1 – March 2022 Major point clouds improvements, Cycles Apple Metal GPU support, Subdivision GPU support, image editor handles larger images, Major performance gains for geometry nodes