Hello !
FLOPS is an abbreviation of FLoating-point Operations Per Second, and is a common measure of a computer’s speed. According to bitcoinwatch.com, the hashrate estimate for the bitcoin network has passed 1 exaFLOPS (or 1000 petaFLOPS)
The bitcoin mining, the process which maintains the bitcoin network, uses almost no floating point operations, relying entirely on integer calculations instead.
The FLOPS estimate is based on the cost for a system to do bitcoin hash operations on its graphics card, rather than perform other tasks. A conversion rate of 1 hash = 12.7K FLOPS is used to determine the general speed of the network contribution.
The ASIC device is very specialized and doesn’t use floating point operations at all.
An exaFLOPS is an impressive number, regardless of whether it is technically correct.
If all supercomputers were combined, they’d have less than 10% of what bitcoin is using.
My question is:
Can I use an ASIC device (let say 100 GH/s) for science and not for mining ?
Is this the right tool ?
It is wrong to express the computing power of the bitcoin network in terms of FLOPS. Like you said, FLOPS stands for FLoating point OPerations per Second. Bitcoin mining consists purely of integer operation. Not single floating point is operated on.
The number people quote for the FLOPS-count for the Bitcoin network is just a mostly arbitrary guess. Back when the majority of the network consisted of CPU and GPU miners, it was probably an okay estimate, since those devices can be used for all kinds of operations. But ASICs can do one thing only and Bitcoin-ASICs only do mining operations.
You could design and produce ASICs for scientific computations and it would be very efficient. However, in science your computer model is often improved or completely replaced, which would make developing ASICs for science a very time-consuming, expensive and ultimately not very worthwhile endeavour. Bitcoin on the other hand has the advantage that the mining algorithm is fixed and while in theory it could be changed, in practice this isn't going to happen unless SHA-256 is found to have a vulnerability.