Dr. Dobb's is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Channels ▼


CUDA, Supercomputing for the Masses: Part 1

Are you interested in getting orders-of-magnitude performance increases over standard multi-core processors, while programming with a high-level language such as C? And would you like that capability to scale across many devices as well?

Many people (myself included) have achieved this level of performance and scalability on non-trivial problems by using CUDA (short for "Compute Unified Device Architecture") from NVIDIA to program inexpensive multi-threaded GPUs. I purposefully stress "programming" because CUDA is an architecture designed to let you do your work, rather than forcing your work to fit within a limited set of performance libraries. With CUDA, you get to exploit your abilities to design software to achieve best performance on your multi-threaded hardware -- and have fun as well because figuring out the right mapping is captivating, plus the software development environment is both reasonable and straightforward.

This is the first of a series of articles to introduce you to the power of CUDA -- through working code -- and to the thought process to help you map applications onto multi-threaded hardware (such as GPUs) to get big performance increases. Of course, not all problems can be mapped efficiently onto multi-threaded hardware, so part of my thought process will be to distinguish what will and what won't work, plus provide a common-sense idea of what might work "well-enough".

"CUDA programming" and "GPGPU programming" are not the same (although CUDA runs on GPUs). Previously, writing software for a GPU meant programming in the language of the GPU. An acquaintance of mine once described this as a process similar to pulling data out of your elbow to get it to where you could look at it with your eyes. CUDA permits working with familiar programming concepts while developing software that can run on a GPU. It also avoids the performance overhead of graphics layer APIs by compiling your software directly to the hardware (GPU assembly language, for instance), thereby providing great performance.

The choice of CUDA device is up to you. Figures 1 and 2 show the CUDA N-body simulation program running on both a laptop and a discrete GPU based desktop PC.

Figure 1: nBody Astrophysics Simulation running on a Quadro FX 570M enabled laptop.

Figure 2: nBody Astrophysics Simulation running on a GeForce 8800 GTS 512MB enabled desktop

Can CUDA really increase application performance by one to two orders of magnitude -- or is all this hype rather than reality?

CUDA is a fairly new technology but there are already many examples in the literature and on the Internet highlighting significant performance boosts using current commodity GPU hardware. Tables 1 and 2 show summaries posted on the NVIDIA and Beckman Institute websites. At the heart of CUDA is the ability for programmers to keep thousands of threads busy. The current generation of NVIDIA GPUs can efficiently support a very large number of threads, and as a result they can deliver one to two orders of magnitude performance increase in application performance. These graphics processors are widely available to anyone at almost any price point. Newer boards will expand CUDA's capabilities by providing greater memory bandwidth, asynchronous data transfer, atomic operations, and double-precision floating point arithmetic among many hardware improvements. Look for the CUDA software environment to expand as the technology evolves and we eventually lose the distinction between GPUs and "many-core" processors. As developers, we have to anticipate that applications with many thousands of active threads will become common-place and look for CUDA to run on many platforms, including general-purpose processors.

Example Applications


Example Applications


Application Speedup

Seismic Database


66x to 100x

Mobile Phone Antenna Simulation



Molecular Dynamics


21x to 100x

Neuron Simulation



MRI processing


245x to 415x

Atmospheric Cloud Simulation



Table 1: NVIDIA summary from www.nvidia.com/object/IO_43499.html

GPU Performance Results, March 2008

GPU Performance Results, March 2008

GeForce8800GTX w/ CUDA 1.1, Driver 169.09

Calculation / Algorithm

Algorithm class

Speedup vs. Intel QX6700 CPU

Fluorescence microphotolysis

Iterative matrix / stencil


Pairlist calculation

Particle pair distance test

10x to 11x

Pairlist update

Particle pair distance test

5x to 15x

Molecular dynamics nonbonded force calculation

N-body cutoff force calculations

10x to 20x

Cutoff electron density sum

Particle-grid w/ cutoff

15x to 23x

Cutoff potential summation

Particle-grid w/ cutoff

12x to 21x

Direct Coulomb summation



Table 2: Beckman Institute table from www.ks.uiuc.edu/Research/vmd/publications/siam2008vmdcuda.pdf

Related Reading

More Insights

Currently we allow the following HTML tags in comments:

Single tags

These tags can be used alone and don't need an ending tag.

<br> Defines a single line break

<hr> Defines a horizontal line

Matching tags

These require an ending tag - e.g. <i>italic text</i>

<a> Defines an anchor

<b> Defines bold text

<big> Defines big text

<blockquote> Defines a long quotation

<caption> Defines a table caption

<cite> Defines a citation

<code> Defines computer code text

<em> Defines emphasized text

<fieldset> Defines a border around elements in a form

<h1> This is heading 1

<h2> This is heading 2

<h3> This is heading 3

<h4> This is heading 4

<h5> This is heading 5

<h6> This is heading 6

<i> Defines italic text

<p> Defines a paragraph

<pre> Defines preformatted text

<q> Defines a short quotation

<samp> Defines sample computer code text

<small> Defines small text

<span> Defines a section in a document

<s> Defines strikethrough text

<strike> Defines strikethrough text

<strong> Defines strong text

<sub> Defines subscripted text

<sup> Defines superscripted text

<u> Defines underlined text

Dr. Dobb's encourages readers to engage in spirited, healthy debate, including taking us to task. However, Dr. Dobb's moderates all comments posted to our site, and reserves the right to modify or remove any content that it determines to be derogatory, offensive, inflammatory, vulgar, irrelevant/off-topic, racist or obvious marketing or spam. Dr. Dobb's further reserves the right to disable the profile of any commenter participating in said activities.

Disqus Tips To upload an avatar photo, first complete your Disqus profile. | View the list of supported HTML tags you can use to style comments. | Please read our commenting policy.