Topic: Cost Defecting Mining (Read 6696 times)

Check out our service. You can use it to mine for a very healthy profit. Feel free to post an analysis.

http://www.bitcoinrigs.com/products/Rig-Rental-Service.html

Do not mine at a current loss. Period. If you are mining at a current loss (even if you expect BTC to go up in the future), then it would be better to utilize the money that you would be spending mining to purchase BTC through mtgox.

If you can mine money at a current profit, then do that instead of buying.

I love you I love you I love you! When my $20 clears I'll show my appreciation

Oh, if you want to get some of your money back from this litlte experiment, go here. Any Amazon AWS account is eligible, I got $20 from it even though I only use a micro instance.

The plot thickens... to somewhere between treacle & caramel.

You can actually rent Amazon images cheaper than the "On Demand" rate. How much cheaper? Well, for the GPU instance, about a third of the price.

How so? Well, you can either rent the server a year or three at a time (that's $5630/year for the GPU instance, yielding ~13.40 blocks - 670 BTC. How's the ol' cognitive biases going?) - OR - you can bid for instances. This is called "spot" instancing. Think of it as the Mt Gox of instance rental. They even have pretty graphs of the history of bids.

So, spot pricing... well, the GPU instance goes for about $0.76 per instance hour. That's a good 1/3 of the usual price. So lets shoot back to my original post and push these numbers around...

• $0.76 hour^-1 x 8 instances = $145.92 day^-1
• 1.64 Ghs^-1 now= 1 block every ~3.4 days
• 1 block = 50 BTC @ USD$3.47 = $173.50 / 3.4 days = $51 day^-1

You're still only making 34% of your money back (if you sell now). You're better off playing pokies. Seriously.

So still not good value. But it's much better. So unless you have spare cash now and expect the value of a bitcoin to more than double before you want to cash out, buy an ATI/AMD 8XXX; and even then... But maybe that's what you want to do? Who am I to judge! I'm someone who just got their Amazon AWS Bill (~$100 if you're interested, for a good two days work!). I'm not begging; 'cos I think the information is easily worth that much - but if you somewhat agree, donation link in the dooblydoo.

Oh, and I also charted up the costs/benefit of spot pricing across the Amazon range. Just to be sure GPU instances were superior. BTW, sorry it's an image; couldn't find a nicer way to export Google Spreadsheets.

https://i.imgur.com/tYXZ6.png

Yep; GPU FTW.

From the Amazon Pricing Page, I found this little nugget:


So that's $0.42 in addition to the $2.10, as well as any S3 and SDB usage fees. You're paying more for the convenience of enhanced interoperability.

lets hope amazon (or any other cluster) doesnt realize they can destroy the whole mining economy of bitcoin in an instant if they use their spare computing time. Would be funny to see some big player let loose their hashing power. in the end the most economical miner will win though, and thats the student in his dorm, where power supply is for free essentially ^^

Thanks!  When I get a few more BTC in my wallet I'll try to remember to send a couple over, because I've been wondering about this.  Have you looked at this at all with regards to the MapReduce pricing @ $0.42/h?  I'm not entirely sure what the difference is from the two, but I'm curious if it would make a difference.

Thanks!ß

The fastest CPU rig has less than (or equal to) the CPU power of the GPU cluster, but at 0.70USD so you can get three of them for the price of the GPU rig. Which max'd out give 25Mhs^-1 each, giving 75Mhs^-1 for the same price of the GPU cluster, which yields almost 3x as many hashes.

If they swapped out the Nvidia's for AMD/ATIs it'd not cost them any more, yet would make it very worthwhile, and would likely change the state of bitcoin slightly (would Amazon target bitcoin miners? upon request? who knows!)

So no, I don't think the CPU rigs could be faster (none of the smaller rigs have GPUs, maybe if they offered a 'micro' with the Tesla, but I very much doubt they would). I ran a CPU rig for a few hours and just quit it. No point.

What about the tiny instances? Hundreds of tiny instances running their little CPUs? No idea. Anyone (with more cash spare) up for the challenge? I'll do it for BTC 1CWc9rkmprfwrF6fkjPRnrohxugvUwSRSg

Funny thing, today i pondered over the same question. You showed that it is not profitable at all. Do you think renting only the obviously faster CUDAs could yield profit?

- or - How to Lose Money Whilst Really Really Trying

Howdy folks. Against the better judgement and superior mathematics of my Bitcoin brethren, I decided to venture into the land of rental GPU power. Burnt myself the better part of 24 hours, as well as a hole in my pocket.

Enter Amazon EC2 Cluster GPU Instances

The stats they give:

• 2 x Intel Xeon X5570, quad-core “Nehalem” architecture
• 2 x NVIDIA Tesla “Fermi” M2050 GPUs
• 64-bit platform
• USD$2.10 per hour
• Up to 8 instances

The stats I got:

• 8 CPU threads x 3200 khs^-1 = 25.6 Mhs^-1
• 2 GPU CUDA threads x 90 Mhs^-1 = 180 Mhs^-1
• 64-bit binaries
• $2.10 hour^-1 x 8 instances = $403.20 day^-1

The stats that matter:

• 25.6 Mhs^-1 + 180 Mhs^-1 = 205.6 Mhs^-1 x 8 instances
• 1.64 Ghs^-1 @ $403.20 day^-1
• 1.64 Ghs^-1 now= 1 block every ~3.4 days
• 1 block = 50 BTC @ USD$3.47 = $173.50 / 3.4 days = $51 day^-1

Permit me a LOL

The experiment did yield spoils, however: in the form of 64bit Linux binaries. Any and all donations are appreciated.

NOTE: These are built against the Amazon 64 Bit Linux AMI's, so YMMV. Debian/Ubuntu users should get TuxSoul's Debian packages instead.

• CUDA RPC Miner & CUDA 2.0 CUBinary  - 1DqJmB2SU4kQPWj3SYS2o9Ap59zJAQQhSG
• CPU Miner Demon with 64bit SSE2 support  - 1LauUnmyHnmBrxcAiYgkfpWQpE9h7k8pmQ

I also built binaries for running your own pool; but they are horribly inefficient.

• CPU Remote Miner Client - 1G4WwqZYkX6XToTQGrqP1Q7P2WJqeYoF2U
• CUDA Remote Miner Client - 1GBJCAzumixa7DyvNiJfZxnzUD1L74cqH1

I can't say I didn't expect financial loss with this experiment; but If I've saved you any time or effort with this post, donations are most welcome. If I never have to see a makefile again it will be too soon.

 - Metonymous -
162Dw2KSJpAJrpVQ5Ert5427GMpnsSLPmb.