机器学习工作站_2020年最佳机器学习工作站

机器学习工作站

Security of data is crucial for many applications and often you would need to run experiments locally at your company whenever you deal with sensitive data. That’s why a perfect solution will be buying a workstation. There are many specialised machine learning workstations available on the market currently and in this text I will review some of them.

数据的安全性对于许多应用程序至关重要，通常，每当您处理敏感数据时，您都需要在公司本地进行实验。 这就是为什么完美的解决方案是购买工作站的原因。 当前市场上有许多专用的机器学习工作站，在本文中，我将对其中的一些进行回顾。

顶级机器学习工作站 (Top Machine Learning Workstations)

My list starts with NVIDIA top-class AI workstation and generally goes down to more affordable machines. Choosing a workstation will depend on your applications and budget.

我的榜单从NVIDIA顶级AI工作站开始，然后通常是价格更便宜的机器。 选择工作站取决于您的应用程序和预算。

NVIDIA DGX STATION (NVIDIA DGX STATION)

NVIDIA® DGX Station™ is the world’s first purpose-built AI workstation, powered by four NVIDIA Tesla® V100 GPUs. It delivers 500 teraFLOPS (TFLOPS) of deep learning performance — the equivalent of hundreds of traditional servers — conveniently packaged in a workstation form factor built on NVIDIA NVLink™ technology.

NVIDIA®DGX Station™是世界上第一个专用AI工作站，由四个NVIDIATesla®V100 GPU驱动。 它提供了500 teraFLOPS(TFLOPS)的深度学习性能-相当于数百台传统服务器-方便地包装在基于NVIDIA NVLink™技术的工作站中。

Truly an enterprise-grade solution and a welcome gift for any machine learning team.

真正为任何机器学习团队提供的企业级解决方案和可喜的礼物。

HyperStation DLX-4R — Quad GPU开发工作站 (HyperStation DLX-4R — Quad GPU development workstation)

Featuring a 14-core, high-performance Intel Xeon W processor, up to four GPUs, and 512GB DDR4–2666MHz, the HyperStation DLX-4R allows developers and researchers to experiment on models of a more complicated nature, using much larger data sets.

HyperStation DLX-4R具有14核高性能Intel Xeon W处理器，多达四个GPU和512GB DDR4–2666MHz，可让开发人员和研究人员使用大得多的数据集来试验性质更复杂的模型。

Features:

特征：

• Intel Xeon W-2175 2.5Hz (4.3GHz Turbo) 14 Core Processor
  英特尔至强W-2175 2.5Hz(4.3GHz Turbo)14核心处理器
• 4x NVIDIA RTX 2080 Ti Turbo 11GB Graphics (4352 CUDA Cores per GPU)
  4个NVIDIA RTX 2080 Ti Turbo 11GB显卡(每个GPU 4352 CUDA内核)
• 128GB ECC DDR4 2666MHz quad-channel Memory
  128GB ECC DDR4 2666MHz四通道内存
• 512GB NVMe M.2 SSD + 2x 4TB Enterprise-class HDD
  512GB NVMe M.2 SSD + 2个4TB企业级硬盘
• Support for up to 4 double-width GPUs @ PCIe x16 (PLX switched)
  支持多达4个全宽GPU @ PCIe x16(切换PLX)
• 2x 10 Gigabit Ethernet
  2个10千兆以太网

BIZON X5000 G2 — AMD锐龙Threadripper 4 GPU深度学习工作站PC (BIZON X5000 G2 — AMD Ryzen Threadripper 4 GPU Deep Learning Workstation PC)

Deep learning workstation by Bizon is powered by the latest AMD Threadipper CPUs, NVIDIA GPUs and whisper-quiet liquid cooling. Starting at $7,990.

Bizon的深度学习工作站配备了最新的AMD Threadipper CPU，NVIDIA GPU和悄悄的液体冷却技术。 起价为7,990美元。

Features:

特征：

• AMD RYZEN Threadripper 3nd Gen 39xx (Up to 64 Cores)
  AMD RYZEN Threadripper 3nd Gen 39xx(最多64核)
• 32 GB DDR4 3200 MHz Memory (up to 256 GB)
  32 GB DDR4 3200 MHz内存(最大256 GB)
• 4 x NVIDIA RTX 2080 SUPER 8GB (optional 4 x RTX 2080 Ti, Titan RTX)
  4个NVIDIA RTX 2080 SUPER 8GB(可选4个RTX 2080 Ti，Titan RTX)
• Cooling
  冷却
• CPU liquid cooling system
  CPU液体冷却系统
• SSD 250 GB PCI-e SSD (Up to 7.6 TB SSD)
  SSD 250 GB PCI-e SSD(最高7.6 TB SSD)
• Up to 4 x 12 TB HDD
  多达4个12 TB硬盘
• SATA-3, RAID, USB 3.0, USB 3.1, USB Type-C, M.2 PCI-E, WiFi, Bluetooth.
  SATA-3，RAID，USB 3.0，USB 3.1，USB Type-C，M.2 PCI-E，WiFi，蓝牙。

Lambda的2个RTX 2080 Ti (2x RTX 2080 Ti from Lambda)

2x RTX 2080 Ti is an awesome looking, great performing deep learning workstation made by Lambda, a company specialized in hardware for machine learning experiments. With its affordable pricing, it’s a great way to build your company’s computing power.

2x RTX 2080 Ti是由Lambda(一家专门从事机器学习实验的硬件的公司)制造的外观超赞，性能卓越的深度学习工作站。 凭借价格合理的价格，这是增强公司计算能力的好方法。

Features:

特征：

• 2-Way NVLink
  2路NVLink
• Intel i9–10900X CPU (10 Cores)
  英特尔i9–10900X CPU(10核)
• 64 GB Memory
  64 GB内存
• 2 TB NVMe SSD
  2 TB NVMe固态硬盘
• Price: $ 6,193
  价格：$ 6,193

自己构建一个深度学习工作站 (Build a deep learning workstation yourself)

The last option you might want to consider when you need a workstation but you are on a budget is building one yourself. You need to be highly technical to do that, but it might be worth it. Have a look at this tutorial here.

当您需要工作站但预算有限时，您可能要考虑的最后一个选择是自行构建。 您需要具有很高的技术水平才能做到这一点，但这可能是值得的。 在这里看看这个教程 。

购买哪个工作站进行机器学习？ (Which workstation to buy for machine learning?)

For the final decision which workstation to choose for your machine learning experiments, you need to evaluate properly how big models you will run. Based on that you can look at the parameters of these workstations to choose the one with enough computing power.

为了最终决定为您的机器学习实验选择哪个工作站，您需要正确评估将运行的大型模型。 基于此，您可以查看这些工作站的参数以选择具有足够计算能力的工作站。

翻译自: https://medium.com/data-science-rush/best-machine-learning-workstations-in-2020-de9900028f19

机器学习工作站