Workstation hardware blueprints
The Deep Origin OS allows you to configure powerful workstations for your computational biology R&D. This page is designed to help you understand the hardware choices that are available.
Hardware blueprint definition
The hardware blueprint defines the hardware that is available to your workstation. The following selections are available:
- Compute cluster
- The compute cluster defines where your workstation is provisioned. You can choose from Deep Origin's AWS regions, or use your own AWS account. More information is available here.
- Central processing power (vCPU)
- The processing power defines your workstation's capacity to process data and run code. One vCPU is equal to one thread of a CPU core.
- Accelerated processing power (NVIDIA GPU)
- The accelerated processing power defines your ability to train and use machine learning models, such as AlphaFold and DeepVariant, and execute parallel programs, such as molecular dynamics simulations with GROMACS. Choose a larger GPU to train and deploy larger models. See the section on GPUs below for advice on choosing the right size.
- Memory
- Defines how much information your workstation can store for immediate processing with your vCPUs.
- Three options are available, with different amounts of RAM per vCPU.
- Small memory: Provides 2 GB RAM per vCPU.
- Medium memory: Provides 4 GB RAM per vCPU.
- Large memory: Provides 8 GB RAM per vCPU.
- High-performance persistent local storage
- The persistent local storage defines how much information your workstation can persist across sessions of your workstation. The storage will be mounted to the home directory (
/home/bench-user/
) of the workstation. In addition, users can mount shared storage drives to the workstations and access managed data from workstations.
- The persistent local storage defines how much information your workstation can persist across sessions of your workstation. The storage will be mounted to the home directory (
Hardware blueprint options for CPU-only workstations
The possible hardware selections for CPU-only workstations are flexible - you can choose any combination of vCPU, memory and persistent local storage. The maximum hardware capacity of CPU-only workstations is outlined below.
Trial users are limited to workstations with a basic resources. To upgrade your account, please contact customer support.
Feature | Maximum |
---|---|
Central processing power | 192 vCPU |
Memory | 1536 GB |
Persistent local storage | 16 TB |
Hardware blueprint options for NVIDIA GPU-enabled workstations
The hardware options for workstations with NVIDIA GPUs are outlined below. The available options are constrained by the configurations that our underyling cloud providers offer.
GPU size | GPU model | GPU memory | Tensor cores | CUDA cores | vCPUs | Memory |
---|---|---|---|---|---|---|
Small | T4 | 16 GB | 320 | 2560 | 4 | 16 GB |
Small | T4 | 16 GB | 320 | 2560 | 8 | 32 GB |
Small | T4 | 16 GB | 320 | 2560 | 16 | 64 GB |
Small | T4 | 16 GB | 320 | 2560 | 32 | 128 GB |
Small | T4 | 16 GB | 320 | 2560 | 64 | 256 GB |
Medium | A10G | 24 GB | 288 | 9216 | 4 | 16 GB |
Medium | A10G | 24 GB | 288 | 9216 | 8 | 32 GB |
Medium | A10G | 24 GB | 288 | 9216 | 16 | 64 GB |
Medium | A10G | 24 GB | 288 | 9216 | 32 | 128 GB |
Medium | A10G | 24 GB | 288 | 9216 | 64 | 256 GB |
Large | V100 | 16 GB | 640 | 5120 | 8 | 61 GB |
Software blueprint options for GPU-enabled workstations
Each Deep Origin software blueprint provides the drivers for needed for NVIDIA GPUs.
Testing the functionality of a GPU
To test for the presence of a GPU and functional drivers, run nvidia-smi
from a terminal (such as with JupyterLab or SSH). You should receive a response indicating the GPU model, driver versions, and more information.
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 510.47.03 Driver Version: 510.47.03 CUDA Version: 11.6 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 Tesla T4 On | 00000000:00:1E.0 Off | 0 |
| N/A 41C P8 15W / 70W | 0MiB / 15360MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+
Editing the hardware blueprint of a workstation
You can edit the hardware blueprints of existing workstations to change their capacity. The following options can be changed: name, summary, vCPU, GPU, memory, persistent local storage, and auto-stop option. To edit the hardware blueprint of a running or stopped workstation, follow these steps:
The software blueprint and compute cluster of a workstation cannot be changed.
Before editing the hardware of a workstation, we recommend saving all active work. Running workstations must be restarted to edit their Hardware. This will stop all active processes.
- Navigate to your workstation.
- Click the actions menu (three dots) at the right of the workstation.
- In the actions menu, click the "Edit" option.
- In the wizard that opens, edit the configuration of your workstation.
- Advance to the final review step of the wizard. This step presents a summary of the new configuration and its costs.
- Click the "Save changes" button.
- If your workstation was running, it will be restarted. If your workstation was stopped, it will remain stopped.