Configuring your Hardware
We make your AI algorithms run faster.
Use CPU or GPU instances to run your models on NimbleBox.ai. Choose from a range of hardware configurations
NimbleBox.ai provides a range of GPUs that can be provisioned based on your requirement.
Being a modular platform, NimbleBox.ai provides customization based on your workload.
CPU Configurations Available
GPU Configurations Available
2 Core, 7.5GB RAM
NVIDIA K80 (12GB VRAM)
4 Core, 15GB RAM
NVIDIA T4 (16GB VRAM)
8 Core, 30 GB RAM
NVIDIA P100 (16GB VRAM)
16 Core, 60GB RAM
NVIDIA V100 (16GB VRAM)
A detailed breakdown of the GPUs available of NimbleBox
​
T4
K80
P100
V100
CUDA Cores
2560
4992
3584
5120
Memory Size (GB)
16
12
16
16
Type
GDDR6
GDDR5
CoWoS GPU memory
HBM2
Memory Bandwidth (GB/sec)
320
2Γ— 240
732
900
Processing Power
8100
1864–2912
4760–5304
7450
Commissioning additional GPUs speeds up parallel processing.

Selecting your customized hardware on NimbleBox.ai

Customize your hardware
    1.
    Click on 'Start Project' from the NimbleBox Dashboard.
    2.
    Click on the 'Hardware Configuration' icon (on the top-right) and then customize the hardware by clicking on the dropdown.
The 16 Core, 60GB RAM requires a minimum of 2 GPUs

Additional Configurations

    Pre-emptible instances: Pre-emptible instances cut down running costs significantly. This is recommended for developing or making quick edits to code and non critical work, not for training. WARNING: Pre-emptive instances may shut down automatically if resources on NimbleBox are limited, irrespective of the auto-shutdown timer. The maximum runtime for a pre-emptive instance is 24 hours.
    Dedicated instances: Have complete control of work without any interruption. Choosing an non-preemptible/dedicated instance will stop your instance only when the auto-shutdown timer has reached its limit, or your credits run out.
Pre-emptible instances may shut down automatically if resources on NimbleBox.ai are limited, irrespective of the auto-shutdown timer.
The maximum runtime for a pre-emptive instance is 24 hours.
Custom configuration:
You may request a specific configuration depending on your workload, and our team shall provision it within 5 business days.
Last modified 1mo ago