| Forum Home | ||||
| PC World Chat | ||||
| Thread ID: 85212 | 2007-12-03 10:30:00 | PS3 can be used to quickly crack passwords | motorbyclist (188) | PC World Chat |
| Post ID | Timestamp | Content | User | ||
| 617594 | 2007-12-03 10:30:00 | so i've heard of a scientist using several to outperform NASA's supercomputers and how it cost less than hiring them for a week with a waiting period, but this is even more impressive (www.tv3.co.nz) actually, don't GPUs crack passwords very quickly too? |
motorbyclist (188) | ||
| 617595 | 2007-12-03 22:49:00 | Yeah, it's all to do with the CPU architecture, and how it processes data. GPU's are good for 3D graphics, which is basically maths, and digital cryptography is using maths to obscure something to the point that it becomes more expensive to crack the code than the code is worth. The people at Folding@Home have cottoned on to this, releasing a GPU-based client for FAH (dunno if it's out of beta though). | ubergeek85 (131) | ||
| 617596 | 2007-12-03 23:15:00 | The reason GPUs are so good at cryptography tasks is because they are massively parallel and have programmable hardware. Instead of building every process on top of a standard instruction set (as general-purpose CPUs require) the GPU hardware can be reconfigured to execute the process directly at a hardware level. Because there is no abstraction layer, performance is vastly superior. |
Erayd (23) | ||
| 617597 | 2007-12-04 02:50:00 | Man, it would be awesome if any of that made any sense to me ^ | --Wolf-- (128) | ||
| 617598 | 2007-12-04 03:12:00 | Man, it would be awesome if any of that made any sense to me ^ :lol: |
motorbyclist (188) | ||
| 617599 | 2007-12-04 04:08:00 | The reason GPUs are so good at cryptography tasks is because they are massively parallel and have programmable hardware. Instead of building every process on top of a standard instruction set (as general-purpose CPUs require) the GPU hardware can be reconfigured to execute the process directly at a hardware level. Because there is no abstraction layer, performance is vastly superior. Man, it would be awesome if any of that made any sense to me ^Apologies for being overly cryptic - explanation coming right up: Massively parallel: Able to process many tasks at once, rather than slicing them up and processing one piece at a time. Programmable hardware: Hardware that allows the design of the chip itself to be reconfigured on-the-fly via software. Standard Instruction Set: A set of basic instructions common to a specific class of hardware - the x86 instruction set (common to almost every desktop CPU produced today) is an example. All software that runs on computers using these CPUs is built up from the basic 'building blocks' of the x86 instruction set. The CPU translates these instructions into its own internal logic on-the-fly as software is executed. ...execute the process directly at a hardware level: The hardware executes the software logic directly; there is no 'instruction set' to translate. ...no abstraction layer...: There is no need to translate an 'instruction set' in order to execute software.Does that clarify things? |
Erayd (23) | ||
| 617600 | 2007-12-04 05:21:00 | after seeing the guy talk at kiwicon..... the guy has coded MD5 for the cell chip and it run at speeds that make you think he made a mistake. this could be changed easily to do windows password hash too. running on a linux distro for the PS3, the code idea is much the same as what can be use on GPU. EG vector math this kind of cracking speed has been seen before but not without custom chips or PGA's, to get this speed with an off the shelf system is the cool trick. |
robsonde (120) | ||
| 1 | |||||