Getting Started

Dragonfly can run on a workstation with the following minimum specifications. However, you should note that performance and productivity will likely be limited and that such a system may not be able to handle larger sized research datasets.

• Windows 7/8/8.1/10/11 (64-bit)* or Linux (64-bit), starting from Ubuntu 18.04
• AVX compliant CPU
• GPU with 1+ GB of VRAM (NVIDIA or AMD) or Intel integrated graphics processor (IGP) with UHD Graphics**
• Support for Open GL 4.3 or higher
•  8 GB of RAM
• 10+ GB of local hard disk space for download and installation
  
  * Windows 10/11 is required for deep learning.
  ** Performance will likely not be comparable to running Dragonfly 3D World on a system with a dedicated GPU.
   

Dragonfly needs a high-performance workstation in order to handle the image processing and visualization demands of large data. If you want maximal performance, then you should exceed these specifications and purchase whatever your budget permits.

• Dedicated GPU graphics card with at least 6 GB of VRAM.
• 32+ GB of RAM (plan on having 4x as much RAM as the largest scans you need to process).
• 2.4 GHz or faster Intel 4-core Xeon or i7 CPU (or AMD equivalent).
• 2 TB of local hard disk storage.

Your hardware choices should be prioritized as follows:

1. GPU (graphics processing unit)
2. RAM (random access memory)
3. CPU (central processing unit)
4. Hard disk storage

One of the most important considerations for optimizing Dragonfly performance is the graphics card. For reference, the minimum and recommended GPU requirements are summarized below. You can also refer to the following table for additional information about 3D rendering and Deep Learning support for specific NVIDIA and AMD GPUs.

NOTE As of Dragonfly 3D World version 2024.1, the application can be installed on a system/laptop with only an Intel integrated graphics processor (IGP) with UHD Graphics. In this case, performance will likely not be comparable to running Dragonfly 3D World on a system with a dedicated GPU. In addition, some advanced features are not supported on systems/laptops without a dedicated GPU.

Recommended Specifications
A dedicated NVIDIA or AMD graphics card with at least 6 GB VRAM, support for OpenGL 4.3 or higher, and the latest WHQL driver for Windows operating systems.

You should note that some functions, such as Deep Learning, PuMA, and CT Reconstruction may require increased graphics card performance. In general, performance will improve with more video RAM (VRAM) and more GPU processing capability (GPU cores). You should also note that the amount of GPU memory needed also depends on the size of your data and that high-end graphics cards have 16 to 32 GB of memory.

When comparing graphics boards, there are many different criteria and performance numbers to consider. Some are more important than others, and some are more important for certain kinds of tasks. It is important that you consider your specific requirements. Wikipedia articles and the NIVIDIA And AMD websites provide specific performance metrics for NVIDIA GeForce/Quadro and AMD Radeon cards.

NVIDIA and AMD GPUs for 3D Rendering and Deep Learning

The following tables list NVIDIA and AMD GPU support for 3D rendering and Deep Learning in Dragonfly starting from version 2021.1 to the current release. Tables last updated 2023-07-27.
 

NVIDIA GPUs	3D Rendering	Deep Learning
GeForce RTX 40 Series	x	x*
GeForce RTX 30 Series	x	x*
GeForce GTX 20 Series	x	x
GeForce GTX 10 Series	x	x
GeForce GTX 900 Series	x	x
GeForce GTX 800 Series	x	x
Quadro RTX A6000	x	x*
Quadro RTX A5000	x	x*
Quadro RTX A4000	x	x*
Quadro RTX 8000	x	x
Quadro RTX 6000	x	x
Quadro RTX 5000	x	x
Quadro RTX 4000	x	x
Quadro GP100	x	x
Quadro GV100	x	x
Quadro P6000	x	x
Quadro P5000	x	x
Quadro P4000	x	x

* NVIDIA GeForce RTX 30 Series and NVIDIA GeForce RTX 40 Series cards, as well as Quadro RTX A6000, Quadro RTX A5000, and Quadro RTX A4000 cards, are only supported for Deep Learning starting with Dragonfly version 2022.1. Earlier versions of Dragonfly only support those cards for 3D rendering, NOT for Deep Learning.

 

AMD/ATI GPUs	3D Rendering	Deep Learning
Radeon RX 7000 XT/XTX Series	x	x
Radeon RX 7000 Series	x	x
Radeon RX 6000 XT Series	x	x
Radeon RX 6000 Series	x	x
Radeon RX 5000 XT Series	x	x
Radeon RX 5000 Series	x	x
Radeon VII	x	x
Radeon RX Vega	x	x
Radeon R400 Series	x	x
Radeon R500 Series	x	x
Radeon R600 Series	x	x

When building a system yourself, you must confirm that your system is compatible with your preferred graphics card. The constraints are the interface of the motherboard (e.g. PCIe 3.0) and the power capacity of the power-supply unit (e.g. 800 W).

Some image processing tasks are very demanding of memory. We suggest you equip your workstation for 4x the size of the largest scan that you wish to work with. (e.g. If you will work with 25 GB scans, then you should arrange for 100+ GB of RAM).

Note that when building a system, you may wish to have a future upgrade path. Many budget systems will not permit you to upgrade beyond 32 GB of RAM. Always pay attention to the maximum supported memory for any system you purchase.

Dragonfly performance will benefit from faster clock speed CPUs and from more cores. If you are doing mostly visualization and minimal image processing, then CPU is less important. If you expect to do extensive image processing, then you should invest in faster clockspeed and more cores.

Dragonfly installations typically require 10+ GB of space. In addition, users will need space to store data they are working on, as well as intermediate and final results. Because users are typically working on large data, we encourage you to get at least 2 TB of local storage.

• Dedicated NVIDIA GPU graphics card with CUDA Compute Capability 3.5 or higher and at least 6 GB of VRAM.
  IMPORTANT NVIDIA GeForce RTX 30xx cards are supported for Deep Learning starting with Dragonfly version 2021.3. Earlier versions only support those cards for rendering, not for Deep Learning.
  NOTE All of our recommended GPUs have the required CUDA capability. You can also go to https://developer.nvidia.com/cuda-gpus to check if your NVIDIA GPU is CUDA-capable.
• 32+ GB of RAM (plan on having at least 4x as much RAM as the largest scans you need to process).
• 2.4 GHz or faster Intel 4-core Xeon or i7 CPU (or AMD equivalent) with the AVX extension. Below is a list of both Intel and AMD CPUs that support AVX.
  • Intel
    • Sandy Bridge processor, 2011
    • Sandy Bridge E processor, 2011
    • Ivy Bridge processor, 2012
    • Ivy Bridge E processor, 2013
    • Haswell processor, 2013
    • Haswell E processor, 2014
    • Broadwell processor, 2014
    • Broadwell E processor, 2016
    • Skylake processor, 2015
    • Kaby Lake processor, 2016 (ULV mobile)/2017 (desktop/mobile)
    • Skylake-X processor, 2017
    • Coffee Lake processor, 2017
    • Cannon Lake (microarchitecture) processor, 2018
    • Cascade Lake processor, 2018
    • Ice Lake processor, 2018

    NOTE Not all CPUs from the listed families support AVX. Generally, CPUs with the commercial denomination "Core i3/i5/i7" support them, whereas "Pentium" and "Celeron" CPUs don't.

  • AMD
    • Jaguar-based processors and newer
    • Jaguar-based processors and newer
    • "Heavy Equipment" processors
      • Bulldozer-based processors, 2011
      • Piledriver-based processors, 2012
      • Steamroller-based processors, 2014
      • Excavator-based processors and newer, 2015
    • Zen-based processors, 2017
    • Zen+-based processors, 2018

    Issues regarding compatibility between future Intel and AMD processors are discussed under XOP instruction set.

• 2+ TB of local hard disk storage.