Primecoin GPU miner: 9.1 CPD on a 280x / pool / 2% dev.fee

[crazy987]

Well, for me after about 3-4 hours mining nothing found yet

[dga]

If crackers can make light work of patching licensed software from the likes of Microsoft and Adobe, what makes you think that your licensing is any better?

We use 3rd party copy-protection software. Not sure if it's better or worse than Microsoft's and Adobe's.

Holders of the license would get early access to updates. There's still a room for additional 20-30% optimization.

As soon as the binary is available to someone who wants to disassemble it, you cannot prevent it from being reverse engineered, simple as that.

So you see no difference in reverse engineering and patching?

As a reply to some of the posters in this thread from someone who's done this publicly before:

(a)  Self-mining usually doesn't work for the devs.  Most of us don't have big GPU farms - we like to code, not manage clusters, or we'd be managing clusters.   Second, it's very capital-intensive, and therefore risky.  Letting people who already have GPUs mine is a far more effective way to go - more people running the software, no need to buy more GPUs into the mining ecosystem.

(b)  The copy protection only has to be "good enough."  If they charge, e.g., 0.07 BTC per GPU -- something in the range of a week's worth of mining output -- is someone with 50 GPUs really going to find it worth putting in 10+ hours of work to break the copy protection?  Or pay 3.5BTC and call it a day?  Of course, there's always going to be someone hanging around bored who wants to crack your software, but it'll take a while, and the self-interested parties won't release it publicly.

(c) If we're being honest, the big concern is not spiking the diff for the sake of difficulty:  It's making sure that the people who pay for the software make a profit on it before the diff rises to the point where it's no more profitable than mining litecoin.  That enables the devs to profit, and provides a reason for purchasers to take an up-front risk of paying for the software in the very uncertain world of crypto-currency mining.  But that also means that the people who buy the software also have an incentive not to let it out into the wild:  Because they'll never recoup their investment if they do.

If they don't hand out the miner binary, and have license-based copy protection, it's not crazy at all to think that they would get a few months of activity before someone released a cracked version of it or some other dev took the thermonuclear option and released an open source GPU miner.  (No, that won't be me, I'm sticking to RIC. :-).

It's not dissimilar to pre-orders for hardware miners, if you think about it that way:  Some people pay a lot of money up front to have access first.  They're funding the time and money it took to develop the hardware.  After that, availability increases, price decreases, but the mining difficulty increases and later buyers get cheaper hardware but lower income.  It's just that hardware has built-in copy protection (grin).

[primeGPU]

Problem with releasing with dev fee, which is usually the method i prefer to see is the difficulty will climb very fast to the point where its no longer profitable to use the miner, meanwhile the price will tank due to dump fest. +1 to the dev's for trying to take a very responsible course on releasing this I hope the software protection holds out, its in no ones interest to see the difficulty go any higher while the price continues to drop. As to the license fee i hope its not so high that smaller miners cannot afford entry to the market leaving only whales to benefit. Will the price be on a per card basis? Using my rig as an example 4x 280x + E3 1230v3 would 50xpm per gpu be a suitable price, would be nice to have a ballpark figure that you are considering.

Yes, the price will be on a per card basis.
We don't want to set a high price for this for the exactly same reasons that you mentioned:
A whale is more probable to dump the coins than an enthusiast, so we prefer to cater to the latest.
We think that 100xpms is definitely too high, and 20xpms is too low, that's all we are sure about right now
That numbers are for the current network difficulty. Next month we will lower the price according to its rise.

[gpools]

Problem with releasing with dev fee, which is usually the method i prefer to see is the difficulty will climb very fast to the point where its no longer profitable to use the miner, meanwhile the price will tank due to dump fest. +1 to the dev's for trying to take a very responsible course on releasing this I hope the software protection holds out, its in no ones interest to see the difficulty go any higher while the price continues to drop. As to the license fee i hope its not so high that smaller miners cannot afford entry to the market leaving only whales to benefit. Will the price be on a per card basis? Using my rig as an example 4x 280x + E3 1230v3 would 50xpm per gpu be a suitable price, would be nice to have a ballpark figure that you are considering.

Yes, the price will be on a per card basis.
We don't want to set a high price for this for the exactly same reasons that you mentioned:
A whale is more probable to dump the coins than an enthusiast, so we prefer to cater to the latest.
We think that 100xpms is definitely too high, and 20xpms is too low, that's all we are sure about right now
That numbers are for the current network difficulty. Next month we will lower the price according to its rise.

report my rig 3 hour find 1 block

[jack80]

Work on Nvidia card ? .

[aamarket]

it does not - and although I asked why ( e.g. here : https://bitcointalk.org/index.php?topic=548230.msg6031104#msg6031104 ),
devs don't care to answer ... feels like they are hiding something ...

[jimmyGee]

If crackers can make light work of patching licensed software from the likes of Microsoft and Adobe, what makes you think that your licensing is any better?

We use 3rd party copy-protection software. Not sure if it's better or worse than Microsoft's and Adobe's.

Holders of the license would get early access to updates. There's still a room for additional 20-30% optimization.

As soon as the binary is available to someone who wants to disassemble it, you cannot prevent it from being reverse engineered, simple as that.

So you see no difference in reverse engineering and patching?

As a reply to some of the posters in this thread from someone who's done this publicly before:

(a)  Self-mining usually doesn't work for the devs.  Most of us don't have big GPU farms - we like to code, not manage clusters, or we'd be managing clusters.   Second, it's very capital-intensive, and therefore risky.  Letting people who already have GPUs mine is a far more effective way to go - more people running the software, no need to buy more GPUs into the mining ecosystem.

(b)  The copy protection only has to be "good enough."  If they charge, e.g., 0.07 BTC per GPU -- something in the range of a week's worth of mining output -- is someone with 50 GPUs really going to find it worth putting in 10+ hours of work to break the copy protection?  Or pay 3.5BTC and call it a day?  Of course, there's always going to be someone hanging around bored who wants to crack your software, but it'll take a while, and the self-interested parties won't release it publicly.

(c) If we're being honest, the big concern is not spiking the diff for the sake of difficulty:  It's making sure that the people who pay for the software make a profit on it before the diff rises to the point where it's no more profitable than mining litecoin.  That enables the devs to profit, and provides a reason for purchasers to take an up-front risk of paying for the software in the very uncertain world of crypto-currency mining.  But that also means that the people who buy the software also have an incentive not to let it out into the wild:  Because they'll never recoup their investment if they do.

If they don't hand out the miner binary, and have license-based copy protection, it's not crazy at all to think that they would get a few months of activity before someone released a cracked version of it or some other dev took the thermonuclear option and released an open source GPU miner.  (No, that won't be me, I'm sticking to RIC. :-).

It's not dissimilar to pre-orders for hardware miners, if you think about it that way:  Some people pay a lot of money up front to have access first.  They're funding the time and money it took to develop the hardware.  After that, availability increases, price decreases, but the mining difficulty increases and later buyers get cheaper hardware but lower income.  It's just that hardware has built-in copy protection (grin).

The point I was trying to make is that they cannot control the number of cards mining. Impossible!

primeGPU stated:
"we are going to sell a limited number of licenses for the first 1-3 months.
The number should be 1440 GPUs - we are able to produce licenses for any number of GPUs (1-6).
This way the number of chains per minute on the network would rise by 1.2 and the difficulty would rise by 35-40%.
2nd month - for 2880 GPUs, maybe 2x this amount. We go open source in 6 months."

If someone buys one of the limited licenses @ 50xpm and it's restricted to 1 card... then of course it's worth their while to crack it if they have a gpu farm. Don't be so naive.
And you obviously know nothing about software development, it will only take a couple hours to crack...
And primeGPU statements show he's complete lack of knowledge.

[primeGPU]

it does not - and although I asked why ( e.g. here : https://bitcointalk.org/index.php?topic=548230.msg6031104#msg6031104 ),
devs don't care to answer ... feels like they are hiding something ...

No, NVidia cards are not suported. They are much slower than their AMD conuterparts in processing openCL code.
Sorry for not anwering earlier.

[primeGPU]

And primeGPU statements show he's complete lack of knowledge.

Sure, after 20 years in IT and after release of a GPU miner for XPM...
Edit: 5 of 20 years in a large software security company.

[ivanlabrie]

My latest findings:

Went from 3.6ghz to 4.25ghz cpu clock wise, and increased ram speed and bandwidth but also increased latency slightly.

End result: loss of performance! Seems like ddr3 latency plays a huge role here.

225 time to 282. 

[dextronomous]

Did you btw fall for the money, i guess  a lot would. You did say a few guys can get it free tough,

i did pm;ed you a few times, did not had an answer. Could you maybe reply or pm to tell if i can get one for free or buy a big
license with the money i dont have.

thanks for your programming skills, and not thinking about the big bucks but the smaller ones,
maybe after i receive i will believe you are true to tell i have to had to pay you for 1 gpu 50 xpm? hell what so much.

its free to reply. Wished i could get a working version so could gain some blocks, but now i gues we if wanted all had to tell that we have
had nothing and still dont and you, stab our eyes out and we still know youre the only one having the keygen for the program.


thanks in advance for sending the license

[primeGPU]

End result: loss of performance! Seems like ddr3 latency plays a huge role here.
225 time to 282. 

We had similar issues when trying to overclock the CPU.

[jimmyGee]

And primeGPU statements show he's complete lack of knowledge.

Sure, after 20 years in IT and after release of a GPU miner for XPM...
Edit: 5 of 20 years in a large software security company.

Sales and tech support doesn't count 

[primeGPU]

Guys, if you haven't received a reply to your PM - I'm trully sorry about that.
There's still some fascist limitations reagrding what number of PMs I can send.
Obv, I reply to the testers first. Also, keep in mind that we're now quite busy making a lot of small changes to the code that fix issues reported by the testers.
Sorry once again.

[primeGPU]

And primeGPU statements show he's complete lack of knowledge.

Sure, after 20 years in IT and after release of a GPU miner for XPM...
Edit: 5 of 20 years in a large software security company.

Sales and tech support doesn't count 

You know, I'm very bad at sales

[madMAx510510]

it does not - and although I asked why ( e.g. here : https://bitcointalk.org/index.php?topic=548230.msg6031104#msg6031104 ),
devs don't care to answer ... feels like they are hiding something ...

It really comes down to AMD having much better hardware for any GPGPU application with their GCN design. Nvidia is now doing the same with their new Maxwell GTX 750 Ti, but it will take them another year to release the high end cards.

Only the GTX 780 and Titan come close to a 7970 in GPGPU applications, but they cost more than double.

[SpeedDemon13]

If you need to test with a different card that's not GCN, ie 5870, 6950, 6970, etc., I have a couple ready to be tested with. If you plan to making it more universal than just the GCN cards, ie 7970, 280x, etc. I assume you have already the 4 testers already.

[primeGPU]

If you need to test with a different card that's not GCN, ie 5870, 6950, 6970, etc., I have a couple ready to be tested with. If you plan to making it more universal than just the GCN cards, ie 7970, 280x, etc.

We do plan to support other cards, but not right away.

I assume you have already the 4 testers already.

Actually, we do have a dozen already. It's just not all of them publish reports

[crazy987]

Well, almost 6 hours in and nothing found yet.
Going to play around with the GPU settings tonight, maybe it needs a diffrent setting.

[ivanlabrie]

Update:

Tweaked ram timings and it went back down to 225, so no change despite my cpu being overclocked (it'll be better for battlefield 4 multiplayer but that's about it lol )

So, yeah, stock clocks are good enough it seems.

As a sidenote, increased gpu clock to 1135 (was at 1100) and I now do 216 vs 225 at 1100mhz. gpu vram only makes it slightly worse if I touch it (again, latency).