[[ build.model.nick ]]

Choose the price, not the parts. Each model is built with the GPUs, CPU, RAM, and storage that maximizes Deep Learning performance per dollar.

[[ build.model.image_alt ]]

[[ build.model.nick ]]

Basic

OS Ubuntu 18.04 + Lambda Stack
GPUs 4x NVIDIA V100 (32 GB)
Fabric 4-way NVLink
CPU 2x Intel Xeon Silver 4116
Memory 256 GB memory
STORAGE 2 TB NVMe (OS Install)
NETWORK 10 Gbps ethernet

Premium

OS Ubuntu 18.04 + Lambda Stack
GPUs 8x NVIDIA V100 (32 GB)
Fabric 8-way NVLink hybrid cube mesh topology
CPU 2x Intel Xeon Gold 6148
Memory 512 GB memory
STORAGE (OS) 1.92 TB NVMe (OS Install)
STORAGE (Data) 3.84 TB NVMe (Data)
NETWORK 100 Gbps InfiniBand

Max

OS Ubuntu 18.04 + Lambda Stack
GPUs 8x NVIDIA V100 (32 GB)
Fabric 8-way NVLink hybrid cube mesh topology
CPU 2x Intel Xeon Platinum 8168
Memory 512 GB memory
STORAGE (OS) 3.84 TB NVMe (OS Install)
STORAGE (Data) 7.68 TB NVMe (Data)
NETWORK 100 Gbps InfiniBand

Customize

Not seeing what you want?

Add a Protection Plan

3-year protection for [[ getWarrantyPrice(true) ]]

[[ getSubtotal(true)]]
Talk to an engineer
(650) 479-5530

About the [[ build.model.title ]] Basic

GPUs

GPUs are the most critical piece of hardware for Deep Learning. The Data Center Basic has 4x NVIDIA GTX V100 GPUs (Volta Architecture) connected with NVLink fabric. For Deep Learning, the V100 offers the best raw power performance of any GPU on the market. Each V100 has 15.7 TFLOPs of FP32 performance (the standard precision for Deep Learning training). For most tasks, the V100 is 1.72x faster than the 1080 Ti and 1.21x faster than the Titan V.

Processor

During training, the CPUs preprocess data and feed it to the GPUs. Slow processors will cause the GPUs to waste cycles waiting for this data. Core count and PCIe lane count are important CPU performance factors. More cores means faster data preprocessing; more PCIe lanes means faster transmission of that data to the GPUs. The Basic has two Skylake Intel Xeon Silver 4116 (24 cores total @ 2.1 Ghz) processors. The Basic's CPUs, combined with its NVLink-enabled motherboard, provides high bandwidth GPU-to-GPU and CPU-to-GPU communication.

Motherboard

A motherboard's GPU interconnect fabric significantly impacts Deep Learning performance. PCIe lanes are data pipes that enable communication between the GPUs and CPU. NVLink is a fabric used for GPU-to-GPU communication. More lanes is better: for example, a device with 16 PCIe lanes can send data faster than a device with 4. When training a neural net, the GPUs and CPU send huge amounts of data to each other. To ensure speedy communication, the Basic's motherboard provides ample PCIe lanes as well as NVLink.

Memory

A Deep Learning computer should have at least as much RAM as GPU memory. For example, a machine with 2x NVIDIA V100 GPUs should have at least 64 GB of memory (V100s have 32 GB of memory each). The Basic has 4x V100 GPUs and 256 GB of memory, so it follows this rule of thumb. If you work with large data sets (e.g. many large images), consider upgrading to Premium, which has 512 GB of memory.

Storage

Most datasets do not fit in RAM. In such cases, during model training, subsets must be repeatedly swapped in and out of RAM from nonvolatile storage. Such a pipeline requires fast, solid state storage; without it, the GPUs would waste cycles waiting for their next batch of data. The Basic was designed with this constraint in mind; it has a high bandwidth 2 TB PCIe NVMe SSD.

Network

The Basic has 10 Gbps ethernet. For downloads, your ISP will almost certainly be the bottleneck. The main benefit of 10 Gbps ethernet is fast data transfer between systems on your own network. For multi-node distributed training, you'll need to upgrade to either the Premium or Max which have 100 Gbps InfiniBand NICs.

Who bought a Basic?

About the [[ build.model.title ]] Premium

GPUs

GPUs are the most critical piece of hardware for Deep Learning. The Data Center Premium has 8x NVIDIA GTX V100 GPUs (Volta Architecture) connected with NVLink fabric. For Deep Learning, the V100 offers the best raw power performance of any GPU on the market. Each V100 has 15.7 TFLOPs of FP32 performance (the standard precision for Deep Learning training). For most tasks, the V100 is 1.72x faster than the 1080 Ti and 1.21x faster than the Titan V.

Processor

During training, the CPUs preprocess data and feed it to the GPUs. Slow processors will cause the GPUs to waste cycles waiting for this data. Core count and PCIe lane count are important CPU performance factors. More cores means faster data preprocessing; more PCIe lanes means faster transmission of that data to the GPUs. The Premium has two Skylake Intel Xeon Platinum 6148 (40 cores total @ 2.4Ghz) processors. The Premium's CPUs, combined with its NVLink-enabled motherboard, provides high bandwidth GPU-to-GPU and CPU-to-GPU communication.

Motherboard

A motherboard's GPU interconnect fabric significantly impacts Deep Learning performance. PCIe lanes are data pipes that enable communication between the GPUs and CPU. NVLink is a fabric used for GPU-to-GPU communication. More lanes is better: for example, a device with 16 PCIe lanes can send data faster than a device with 4. When training a neural net, the GPUs and CPU send huge amounts of data to each other. To ensure speedy communication, the Premium's motherboard provides ample PCIe lanes as well as NVLink.

Memory

A Deep Learning computer should have at least as much RAM as GPU memory. For example, a machine with 2x NVIDIA V100 GPUs should have at least 64 GB of memory (V100s have 32 GB of memory each). The Premium has 8x V100 GPUs and 512 GB of memory, so it follows this rule of thumb.

Storage

Most datasets do not fit in RAM. In such cases, during model training, subsets must be repeatedly swapped in and out of RAM from nonvolatile storage. Such a pipeline requires fast, solid state storage; without it, the GPUs would waste cycles waiting for their next batch of data. The Premium was designed with this constraint in mind; it has a high bandwidth 1.92 TB PCIe NVMe SSD for the OS and 3.84 TB of PCIe NVMe SSD for additional data storage.

Network

The Premium has both 10 Gbps ethernet and 100 Gbps InfiniBand. For downloads, your ISP will almost certainly be the bottleneck. The main benefit of the 100 Gbps InfiniBand is fast data transfer between systems on your own network. For multi-node distributed training, you'll need the Premium's 100 Gbps InfiniBand NIC.

Who bought a Premium?

About the [[ build.model.title ]] Max

GPUs

GPUs are the most critical piece of hardware for Deep Learning. The Data Center Max has 8x NVIDIA GTX V100 GPUs (Volta Architecture) connected with NVLink fabric. For Deep Learning, the V100 offers the best raw power performance of any GPU on the market. Each V100 has 15.7 TFLOPs of FP32 performance (the standard precision for Deep Learning training). For most tasks, the V100 is 1.72x faster than the 1080 Ti and 1.21x faster than the Titan V.

Processor

During training, the CPUs preprocess data and feed it to the GPUs. Slow processors will cause the GPUs to waste cycles waiting for this data. Core count and PCIe lane count are important CPU performance factors. More cores means faster data preprocessing; more PCIe lanes means faster transmission of that data to the GPUs. The Max has two Skylake Intel Xeon Platinum 8168 (48 cores @ 2.7Ghz) processors. The Max's CPUs, combined with its NVLink-enabled motherboard, provides high bandwidth GPU-to-GPU and CPU-to-GPU communication.

Motherboard

A motherboard's GPU interconnect fabric significantly impacts Deep Learning performance. PCIe lanes are data pipes that enable communication between the GPUs and CPU. NVLink is a fabric used for GPU-to-GPU communication. More lanes is better: for example, a device with 16 PCIe lanes can send data faster than a device with 4. When training a neural net, the GPUs and CPU send huge amounts of data to each other. To ensure speedy communication, the Max's motherboard provides ample PCIe lanes as well as NVLink.

Memory

A Deep Learning computer should have at least as much RAM as GPU memory. For example, a machine with 2x NVIDIA V100 GPUs should have at least 64 GB of memory (V100s have 32 GB of memory each). The Max has 8x V100 GPUs and 512 GB of memory, so it follows this rule of thumb.

Storage

Most datasets do not fit in RAM. In such cases, during model training, subsets must be repeatedly swapped in and out of RAM from nonvolatile storage. Such a pipeline requires fast, solid state storage; without it, the GPUs would waste cycles waiting for their next batch of data. The Max was designed with this constraint in mind; it has a high bandwidth 3.84 TB PCIe NVMe SSD for the OS and 7.68 TB of PCIe NVMe SSD for additional data storage.

Network

The Max has both 10 Gbps ethernet and 100 Gbps InfiniBand. For downloads, your ISP will almost certainly be the bottleneck. The main benefit of the 100 Gbps InfiniBand is fast data transfer between systems on your own network. For multi-node distributed training, you'll need the Max's 100 Gbps InfiniBand NIC.

Who bought a Max?

[[ component.name ]]

[[ option.description ]] [[ build.getPriceDiff(component, option)]]