This question arises quite often.

Selling your hash power makes sense.

Why?

Because you give up your risk to your customers and monetize immediately your investments instead of holding and holding and holding. If you have even a very small amount of loans then having a constant income lowering the risk can be extremely beneficial.

Yes indeed, he would have probably pulled my hair a bit... but I still believe GPU multipools are going to stay - they're really a good thing to start while you wait for the FPGA cards to be designed.

I honestly wouldn't even consider it with less than an hundred thousand... I've had a quick discussion with a dude some time ago who was going to invest in an established farm by buying a share. I think he was going to invest some 7-digits number.

I'd say that's what it takes to get into BTC mining with reasonable risk by now. I told him to do a GPU multipool instead!

I find somewhat ironic that MYR didn't make it to the list in time by last edit 2014-01-05 yet it's still alive and the best has yet to come!

Yes, I understand what you're trying to say.
I only wanted to make clear there's a need for improvement.
At the end of the day HW alone does nothing and SW alone does very little. NV right now has a certainly better ecosystem around it.

It is related to the "shape" the kernel gets.
Imagine putting a square peg in a round hole. To do this, GPUs split the beg in pieces, pass each smaller piece then put them back on the other side.
AMD believes this is not efficient usage of transistors and this is a reason they won the console war again. NV by contrast believes they should optimize the worst case and this is what they do.

Chained hashing is even worse: you have taped several pegs together!

It took me almost two months to redo qubit + miner modifications (qubit is the tail of X11 FYI, it could be considered X5). I guess you get the idea.

It is possible - and perhaps even likely - many have the improved kernels already. If they have a room full of GPUs, they are consistent advantage. If you have one, not so much.

Not all algorithms can be improved. Some are so simple they're close to efficient... it's on a case-by-case basis. For example, Echo is nearly optimal. I honestly don't like X11/X13 much, the main problem I see is: it is inefficient for GPU users, it is inefficient for a small company doing FPGAs... it is a pipe dream for a big investor who wants to roll out its own ASIC already (they are shelling out 7-digits so paying four engineers instead of one isn't much of a problem). I simply don't like the idea but it's an hash like everything else out of here.

Just to make this clear: NVidia does not just have a 12+ months advantage in manufacturing capability... they also use different kernels, which were properly ported.

The state of things for AMD miners is sad. The code hasn't been ported to GPU, it has basically been mashed to run with a direct conversion of the CPU code taken from SPHlib. Unfortunately, AMD hinted a few times they're not interested in making slow code go fast but rather provide power to those who know where to look at.
Just porting (not even optimizing) made my qubit go up 80%.

So NV might be good but you can bet a large part of benefit is purely software.

Even reasonably intelligent people cannot understand cryptography. So, there's no proving it, we can have an idea.

Here's my understanding.

When your miner connects to the network it builds an initial merkle root by hashing block data + miner data. I know at GetWork level this hash is built to be unique albeit I have only focused on stratum.

When scanning for nonces, we look for partial collisions on the hash obtained by a function of this merkle root.
If we all start from the same initial state, then the fastest miner always wins at solving the block deterministically, minus the non-determinism involved in network propagation.
So what you have is a form of time-shifting somehow where the initial hash somehow shifts your results. Your (nonce, nonce2) still goes on as usual (stratum parlance) but due to the different bits in the initial hash this is effectively transformed in a non-linear space.

Now, some coins/algos produce this initial hash, but let's assume this is not an issue and we just use SHA256D.
What we obtain is a sequence of garbage numbers. We append a nonce to those numbers and go hashing.

Now, for all purposes, an hash function is a highly nonlinear deterministic function which we can consider opaque.
The output is an hash. We consider the nonce candidate to sending if it counts at least D leading/trailing bits.

Now, given a 80-byte hash Hi you can be about 99.999999...% sure of the following

    Algo1(Hi) != Algo2(Hi) != ... != AlgoN(Hi)

It is also very there will be at most one golden nonce in the above results and usually much less.
Different algorithms map our input space nonlinearly in a different way but this isn't consistently different from just using a different initial merkle. Even more so because the merkle is nonlinear in itself.

So honestly I don't know from where your concern is stemming.
It seems to me everything is just an extension of the PoW concept.

When it comes to putting multiple algos in the same chain, the obvious point is ensuring their diff is adjusted accordingly, which is what the MYR team is doing. Truth to be told the numbers so far seem to support the system is working with minor deviations from expected results. Which is unsurprising to me.

Those GPU hashes are ridiculous. I'm going in circles in trying to understand what CryptoNight really is (all I see is statements) but there's very little info, I think I might just hit the code in the future.

If you guys think 2nd gen difference to be technical you're wrong big way.
Not only the tool, but how is it used. A simple mindset shift can make a difference.

Thank you CrashOD, that is possibly better than what I hoped for!
It's very unfortunate the GPU section lacks wattages. I bookmarked this as I'll have to read it more carefully.

Hello!
In the last few days I'm taking a look at CryptoNight as I'm considering to diversificate my mining operations to some CryptoNote currency.
However, I'm having a lot of difficulty understanding what's going on with the whole mining thing. Hashrates are extremely low, but I assume this is the case... I'm having difficulty finding a place when I can look at this data.

For example, scrypt hashrates can be checked out at the LTC wiki more or less.
The other day I've found a very handy google spreadsheet about groestlcoin.
Myriad has a ... unfortunately not very populated ... database of hashrates across different hardware and algos.

I've been looking at the miner threads but the information here is very sparse.

Perhaps you could just dump there your hashrates?

In the past I would have considered 2nd gen coins have the following characteristics:

• Algorithm: not SHA256-D
• Block time: much faster, say 2 minutes or less

What they tried to do was to keep them accessible to miners (by changing algo) and to everyday users (10 minute block time is unacceptable for many brick-n-mortar).

In the past I would have considered LTC to be the first 2nd gen coin as they claimed to be interested in being minable by everyone. It has been shown this is not the case. LTC is in fact a "1.9" gen coin.

Most alts are in my opinion rough implementations of the 2nd gen concept. Only two surviving are perhaps VTC and MYR, and I'm not even sure VTC can make it as their way to keep it distributed, albeit effective, is not strictly a technical advancement.

MYR is technically ahead with its multi-pow approach. Perhaps it is the first true 2nd gen coin. After all, facts have shown that changing algos alone does not make a substantial difference. By contrast, multi-pow does.

I heard Eth is super smart. Lots and lots of things BTC cannot do apparently. I don't understand a single thing about them and those "contracts" but the devs went around, talked around, know their shit... today I found out ETH launched...


And I can buy it.

Now.

Are you kidding dude? I don't buy this for a second.

Thank you for the link! It sucks to read that!

Personally I always considered stratum a bit brittle. It's not like I require something to be fully extensible to be "future proof" but at least a version string on connection setup would have been nice.

I guess there goes my plan to take a look at cryptonight.

In the meanwhile, I connected my prototype miner to the server and it indeed gets confirmed logins as usual. 
I went back to the code of the legacy miner and found out I forgot to dump initiate_stratum. D'oh!
I now dump receive line. Easier and more accurate.

Thank you very much you got the ball rolling.

I don't like the amount of sarcasm implied there   

Any gem you feel to share?

Hello, I've just dumped some data from a p2pool server (MYR qubit in case this is relevant, albeit I don't think so).
It has come to my attention mining.notify got an ID. It is therefore no more a notify but a request.
It also seems to me the miner isn't confirming it (obviously, since it branches on the string name first).

So it seems to me the whole stratum protocol has gone awry big way and now everyone does whatever (s)he pleases while writing servers.

I'm in the process of writing an educational miner and I would like to understand this better.

I think everybody knows now the official stratum specs by Slush are way outdated but this is taking it to a new degree... of nonsense as far as I am concerned.

What would you suggest to do?

Update: it failed to lock the GPS, so I couldn't even get an idea of how this is supposed to work.
I'm fairly disappointed. It seems like users also had a discussion about this on reddit.

Hello, not very much activity here. I'm afraid bumping up this is necessary.

I have installed the app yesterday on a low-end android device. I am negatively impressed by the huge amount of memory required to install the app. Apparently, having 25MiB of free memory is a problem on low end and it required me to go through hoops to get it to install.

The instructions were not quite clear. The login screen might be ok for a native English user but something gets lost in translation I'm afraid: "sign in" and "sign up" doesn't make much difference to most here.
I guess this "energy" thing is required to avoid an user picking up everything in a single go. This sounds as a good idea in gameplay theory but it's just silly in practice. Besides I don't know how much time energy takes to regen, take notice people buying high-end will likely have more devices available (effectively multiplying N times). By contrast, in all my family (low-end users), we have 1 low-end device.

I am going to try it today going in a not-so-near city over here. I'm not expecting to turn a profit out of this but I seriously hope I'll be able to at least pay me an ice cream.

I am pretty worried the treasures seem to be measured in µBTC. If a treasure is 1000µBTC then at current price this is in the order of half an EUR. This cannot be possibly real. I can anticipate if the prize is that low, there will be bad, bad consequences in the perception of crypto here.
Mix this with the "energy" ruleset and the picture gets not very pretty.

I'll keep you posted.

Nobody said it's "valid". It's advertising. Thread is "X11 FPGA Farms now on offer". Just hit http://www.reddit.com/r/vertcoin/.

FPGA X11 farms reported to be available on VTC reddit.

I cannot yet say much on X11 in particular as I'm still looking at qubit (X5), so far:

• Luffa is a joke. I am surprised it was proposed as a SHA candidate, even more it made a few rounds. You need far less than a talented programmer to port it.
• Cubehash seems to have been designed since day one to be super-efficient in terms of die area and I expect this to be observable on FPGAs as well

I'm stuck with SHAVite right now. The amount of math involved is crazy since it employs AES but it's nothing a mathematician cannot explain in a couple of hours... and I never had a go at cryptography up to two weeks ago.

You don't need "facts" to understand someone might do FPGAs... just look at the code. It is incredibly simple. Given my experience, I hardly believe you need someone highly skilled to make this work in FPGAs.