Compute4Cash is offering to pay real money for your unwanted GPU cycles. It's an intriguing idea - but what does it intend to compute with your hardware?
Renting out unused GPU cycles to repay the cost of the card is a new idea put into practice by Compute4Cash. All you have to do is make sure you have the right GPU-based graphics card, download its client and start earning money. Payment is automatic via PayPal but it is unclear how much your GPU has to compute to clock up a Work Unit or WU. Whatever a WU actually is, in February you will be paid $0.20 falling to $0.10 per WU when the special offer ends.
So what is this all about?
The company's domain name was registered at the end of 2010 and a website appeared soon after, complete with various publicity efforts. A new page has been created on Instructables and there are Google ads aimed at getting people signed up. There are comments on the Instructables page speculating on whether or not it would actually be worth going out and buying a pile of GPU cards just to make some money. Others factor in the electricity consumed by these power-hungry cards and seem to conclude that it isn't - of course, the monetary benefit is all based on the size of a WU.
The Compute4Cash site shows how many people are connected and how many WUs per hour are being logged. Around 20 WU/hr seems to be figure and this means that the company is paying out around $4 per hour which is hardly big money. The total number of WU generated is given as around 2500 which means that the company's total budget is around $500 - again hardly big money.
You can speculate that the economics are set so that Compute4Cash is actually buying your GPU cycles at below cost and hence running a model for less than the cost of the power used. You could also speculate that the whole thing is a scam and the client you have to install is using more than GPU cycles. When you invite a paying guest in there is no way of determining how trustworthy it is.
The whole thing may disappear overnight but ... now the genie is out of the bottle it will probably be reinvented any number of times. The real question is not "is this a scam?" but "is this a viable economic model for distributed computation?".
Building a GPU-based super computer by using spare GPU cycles might be a good way to do research tasks like simulating the weather or any complex system. It is also a good way to crack passwords and decrypt documents and GPUs have already been proven to be good at this. Suppose you rent out your GPU and it turns out that the computation is the batch cracking of millions of passwords - are you an accomplice?
Geeks3D has some information from Compute4Cash about what it claims to be computing. It is allegedly a hash with specific properties for the "financial sector" . The task is encrypted compression and you won't get into trouble for helping compute it. Of course there is no way to verify this and in general secrecy is going to be a feature of such GPU computation projects. After all if it is worth paying to have computed then you would want to keep it secret.