The ATI’s Radeon HD 5970 is a Supercomputer in your desktop

March 27, 2010
131 Views

The Radeon HD 5970 is the fastest single card graphics solution – but it is also a powerful supercomputer!

Bright Side of News recently wrote an article reporting on ElcomSoft’s GPU computing attempts.  They used various GPUs to hack WPA-PSK passwords.  Check out the results here.   ElcomSoft leverages the immense computing power of latest generation graphics cards.  Cards such as ATI’s Radeon HD 5970, and nVidia’s GTX 295.  The HD 5970 is ATI’s latest offering, it has 2 GPUs, with 3200 stream processing units, and 2 GB of DDR5 RAM.  Stream processors are a form of parallel processors that are optimized for I/O intense operations.  This GPU is basically an incredibly powerful computing device inside the computer traditionally optimized for video processing.  However, nVidia uses  a process called “CUDA”, and ATI one called “Stream” to utilize the GPUs as parallel processing chips.   These processes enable incredibly powerful and capable parallel processing from GPUs.

CUDA and ATI Stream are valued  by crytopgraphers, developers, and programmers as supercomputers.  They have found these to be (relatively)

The Radeon HD 5970 is the fastest single card graphics solution – but it is also a powerful supercomputer!

Bright Side of News recently wrote an article reporting on ElcomSoft’s GPU computing attempts.  They used various GPUs to hack WPA-PSK passwords.  Check out the results here.   ElcomSoft leverages the immense computing power of latest generation graphics cards.  Cards such as ATI’s Radeon HD 5970, and nVidia’s GTX 295.  The HD 5970 is ATI’s latest offering, it has 2 GPUs, with 3200 stream processing units, and 2 GB of DDR5 RAM.  Stream processors are a form of parallel processors that are optimized for I/O intense operations.  This GPU is basically an incredibly powerful computing device inside the computer traditionally optimized for video processing.  However, nVidia uses  a process called “CUDA”, and ATI one called “Stream” to utilize the GPUs as parallel processing chips.   These processes enable incredibly powerful and capable parallel processing from GPUs.

CUDA and ATI Stream are valued  by crytopgraphers, developers, and programmers as supercomputers.  They have found these to be (relatively) inexpensive ways to boost CPU cycles.  An ATI Radeon HD 5970 runs $650 on the streets, and yet is 25x as powerful as the $250 Intel core i7 920 CPU (at the task of password cracking).  The Core i7 920 is mainly a consumer chip, but has found use in powerful workstations.  Intel has recently released a six-core version on the Core i7 platform, the 980X, which retails around $1,000, yet barely breaks 6,000 guessed passwords per second (the HD 5970 hits 100k, while the 920 is about 4k).  Thus, utilizing GPU processing can be substantially less expensive than upgrading CPUs (with greater pay off).

So what does this mean?  I think we might see malicious actors start to use GPU enabled laptops or remote desktop access to hack into wireless networks.  An actor with a netbook or a properly equipped smartphone can dial into their GPU enabled desktop at home enabling access to this supercomputer.  This capability to crack passwords faster than ever will provide them with powerful capabilities to get into networks, and from there, cause harm.  Additionally,  CUDA/ATI Stream allow users to plug into large grids to enable massive supercomputing.  SETI and Folding@Home are two such efforts that use distributed computing resources to complete monumental tasks.  Smart hackers/adversaries could turn botnets into distributed, powerful supercomputer grids, and leverage them to deploy even more powerful attacks/exploits.

ASUS G Series G73JH-X1 17.3" Windows 7 Home Premium 64-bit NoteBook - Retail

This laptop could hack your WIFI network in just a few seconds!

GPU computing is here to stay, and there will be a wide variety of formats that benefit consumers and corporate users.  I think we will see GPU computing used in more and more fashions.    There are a few barriers to entry (high dollar GPUs, expensive motherboards, large space requirements, and large power requirements).  These GPUs cannot be plugged into your standard desktop PC.  They need motherboards with PCI Express Ports, and a dedicated 300W+  each.  But the ability to readily add incredible computing power to your PC (or laptop) is one that cannot be ignored.

To our readers: do you have any experience with GPU computing? If so, please leave your thoughts, conclusions, and experiences below.

Link to original post