M5 Ultra Mac Studio 智能体测试胜单台 GB10，但落后双 Spark 集群
Tom's Hardware 将配备 36 核 CPU、80 核 GPU 和最高 512GB 统一内存的 M5 Ultra Mac Studio，与单台 Dell Pro Precision GB10 进行对比，并称后者与 Nvidia DGX Spark 几乎相同。使用 llama.cpp 运行 Qwen 3.8 27B，并启用其原生多 token 预测后，Mac 在 Artificial Analysis AgentPerf-Local 测试中的用时不到 GB10 的一半。报告还列出了内存带宽对比：Mac 为 1.2 TB/s，GB10 为 273 GB/s。但在 DeepSeek V4.1 Flash 1017 测试中，Mac 的 llama.cpp 配置处理最长测试上下文需 10–25 分钟，而采用 DSpark 推测解码和 TensorFold 的双 Spark 集群最多用时不到 5 分钟。报告称，Mac Studio 的交付周期已延长至 2027 年 2 月。
测试所用 Mac 配置售价约 1.23 万美元，而配备 4TB 硬盘的两台 Spark 系统据称总价约 1.5 万至 2 万美元或以上。结果显示，Mac 的原始硬件规格更强，并在智能体测试中击败单台 Spark 类系统；但在大型模型推理软件、模型兼容性和支持方面，Nvidia 集群具有优势。
Apple’s updated Mac Studio with the M5 Ultra SoC has already proven itself as one of the most powerful systems around, thanks to its 36-core CPU and 80-core GPU with Neural Accelerators for the matrix math operations that drive AI compute. Beyond that raw power, options for up to 512GB of unified memory have many local AI enthusiasts lining up for the 13 to 15 weeks it’ll take for Apple to ship one right now.
But is our tested Mac Studio config worth a whopping $12,300 or more for those folks? We already tested the maxed-out M5 Ultra chip with the popular Qwen 3.8 27B in our initial review, but that compact dense model doesn’t come anywhere close to exercising 256GB of memory. We wanted to fill that pool to the brim with cutting-edge large models like DeepSeek V4.1 Flash 1017 and GLM 5.3 Flash.
We also have some new benchmarks at our disposal to see how the Mac Studio stacks up in agentic tasks on the GPU side in the wake of our review of the Microsoft Surface Laptop Ultra with Nvidia’s RTX Spark chip. So there's much to dig into.
#Apple#Nvidia#Mac Studio#M5 Ultra#Dell Pro Precision GB10#DGX Spark#DeepSeek V4.1 Flash 1017#GLM 5.3 Flash#Qwen 3.8 27B#llama.cpp#MLX#TensorFold