http://www.wired.com/wiredenterprise/20 ... -language/
Noir (or anyone else familiar with how GPU's work), about halfway through the article, it mentions that Amazon offers GPU processing as a cloud service. How exactly does that work and under what circumstances would it be beneficial? And what are the performance comparisons?
Just curious.
GPU Article
Moderator: Moderators
- Ylyrra
- Member
- Posts: 5704
- https://www.behance.net/kuchnie-warszawa
- Joined: Fri Jan 16, 2004 10:34 pm
- Location: Georgia, USA
GPUs are really good at specific types of processing. They are used highly in cryptography work.
NVIDIA has a generic high-level article on it here: http://www.nvidia.com/object/what-is-gpu-computing.html
Ars did a great series of articles on password cracking, here's probably the best one: http://arstechnica.com/security/2012/08 ... r-assault/
In there they say that a single Radeon HD7970 can do something like 8.2 billion password calculations a second.
I'm sure there's other applications GPU computing is good for as well.
NVIDIA has a generic high-level article on it here: http://www.nvidia.com/object/what-is-gpu-computing.html
Ars did a great series of articles on password cracking, here's probably the best one: http://arstechnica.com/security/2012/08 ... r-assault/
In there they say that a single Radeon HD7970 can do something like 8.2 billion password calculations a second.
I'm sure there's other applications GPU computing is good for as well.
Basically it depends on the type of calculation you need to perform. A GPU is optimized to get data from memory/cpu and push it through its pipeline and out as an image to the display. The bottle neck is getting the data back out of the GPU.
It just so happens that if you need to perform the same set of calculations over and over on similar data with very little branching then the GPU can split the data between its 100+ cores and do them in parallel. However if you need to calculate something that has a bunch of condition logic then its likely to run faster on the CPU.
It just so happens that if you need to perform the same set of calculations over and over on similar data with very little branching then the GPU can split the data between its 100+ cores and do them in parallel. However if you need to calculate something that has a bunch of condition logic then its likely to run faster on the CPU.
Password hackers use multiple GPU's to run passwords against libraries to make it faster.
<a href="http://eq.magelo.com/profile/1397234" target="_blank"><img src="http://eq.sig.magelo.com/1397234.png" border="0"></a>