Topic: How many OPs for £1 on BTC and other clouds? (Read 764 times)
Well, you can look at the price of GPU instances on Amazon for estimates, if you're looking for float-heavy computations in CUDA. Doing the stuff privately would be probably cheaper (no datacenter etc.) but still it gives an estimate.
How many operations per second, or perhaps more precisely FLOPS can I buy for £1, or if you prefer $1?
Some crude back-of-the-envelope guesstimation shows that £1 should buy you more than 5 GFLOPS (general purpose) these days, meaning to buy the hardware. There are also operations costs and overhead to keep running; you can get a few GFLOPS per watt or even much more.How much are you really going to pay? More guesstimation indicates your £1 will buy you a TFLOPS for an hour. But I have no confidence in that unless people chime in with how much they are really paying.
Quote
Is there a chart showing the price history for thease idearly together?
FLOPS on Wikipedia, also "Performance per watt"
You could count how many operations 1 "hash" needs and then extrapolate from MH/s numbers...
One problem is, that with specialized hardware you'd get insanely high efficiency (meaning a lot more operations on Bitcoin relevant work) but you wouldn't be able to use this for anything else. Also monitoring something down to the instruction level is not trivial with modern CPUs which do all kinds of pre-computation, caching of results, instruction sets that do complex operations in one step etc.
One of the closest measurements could be that for example my CPU uses in full load 100W, electricity costs me 10€-cents per kWh, so I could let it run for max. 10 hours (assuming I have the space, cooling and hardware for free) if you pay me 1 €. You could then benchmark whatever you want to run (e.g. Bitcoin mining) on that CPU and work your way from there.
From difficulty you can derive the average hash rate of the past ~2 weeks. With this number you can then (as you know the amount of coins generated + fees paid, as well as the exchange rate to GBP or USD) calculate how much ALL miners earn each day and how much 1 MH/s earns each day. Then you can see how much MH/s different hardware can calculate and see how high/low electricity rates need to be to break even.
One problem is, that with specialized hardware you'd get insanely high efficiency (meaning a lot more operations on Bitcoin relevant work) but you wouldn't be able to use this for anything else. Also monitoring something down to the instruction level is not trivial with modern CPUs which do all kinds of pre-computation, caching of results, instruction sets that do complex operations in one step etc.
One of the closest measurements could be that for example my CPU uses in full load 100W, electricity costs me 10€-cents per kWh, so I could let it run for max. 10 hours (assuming I have the space, cooling and hardware for free) if you pay me 1 €. You could then benchmark whatever you want to run (e.g. Bitcoin mining) on that CPU and work your way from there.
From difficulty you can derive the average hash rate of the past ~2 weeks. With this number you can then (as you know the amount of coins generated + fees paid, as well as the exchange rate to GBP or USD) calculate how much ALL miners earn each day and how much 1 MH/s earns each day. Then you can see how much MH/s different hardware can calculate and see how high/low electricity rates need to be to break even.
Then, how many (BOPS?) bit (bitwise?) operations per second, or Integer operations?
Equivalent for all operations in one of those forms, I'm trying to get a feel for roughly how much it would be for a given cost.
Equivalent for all operations in one of those forms, I'm trying to get a feel for roughly how much it would be for a given cost.
How many operations per second, or perhaps more precisely FLOPS can I buy for £1, or if you prefer $1?
In comparison can the price and the difficulty be used to work this out for the BTC nework? I suspect not but I also suspect estimates to have been made. I think the number would be need to be reduced by some proportion to compensate for the way that the BTC cloud contains hardware designed only for mining BTC.
Is there a chart showing the price history for thease idearly together?
In comparison can the price and the difficulty be used to work this out for the BTC nework? I suspect not but I also suspect estimates to have been made. I think the number would be need to be reduced by some proportion to compensate for the way that the BTC cloud contains hardware designed only for mining BTC.
Is there a chart showing the price history for thease idearly together?
FLOPS performance is kinda meaningless in the Bitcoin network, since it's just doing bitwise and logical operations on integers. Whatever estimates are there assume too much, simply because every architecture is different in how FLOPS performance translates to hashing performance. (See NVidia vs. ATI)
How many operations per second, or perhaps more precisely FLOPS can I buy for £1, or if you prefer $1?
In comparison can the price and the difficulty be used to work this out for the BTC nework? I suspect not but I also suspect estimates to have been made. I think the number would be need to be reduced by some proportion to compensate for the way that the BTC cloud contains hardware designed only for mining BTC.
Is there a chart showing the price history for thease idearly together?
In comparison can the price and the difficulty be used to work this out for the BTC nework? I suspect not but I also suspect estimates to have been made. I think the number would be need to be reduced by some proportion to compensate for the way that the BTC cloud contains hardware designed only for mining BTC.
Is there a chart showing the price history for thease idearly together?
Jump to: