Re: [ANN] sgminer v5 - new unified multi-algorithm on-the-fly kernel switching miner

http://zuik.org/429k.png

Humm. I'm not sure why I'm showing you this, maybe because I have something to brag and I like bragging.

I made some small optimizations on the Wolf's neoscrypt kernel. Now it does 430khs on 280x, 450khs on r9 290, and 260khs on r9 270.

I'm thinking what I should do with this. There's no point to release this to public as this would just make net hashrate rocket and everyone would still get the same profit.

Either release it for free, or bury it and never release it.

Optomizing the kernel and skyrocketing the hashrate by releasing it does nothing but allow cards to get what they should really be getting anyway.

Burying it changes nothing.

Selling it, simply makes people that can't afford to buy it because of the low ROI on GPU's angry with you, and the whales that can afford to spread the ROI across many GPU's get fatter.

Also makes devs poorer. Just saying.

Selling it responsibly can be good for everyone involved - don't be a dick and sell a 10x faster miner, release most of it slowly, and sell one that's around 2x faster. That way, the public isn't too far behind.

Program it into a miner and have the miner auto donate 1%.. Easy enough. Level playing field AND you get coin.

Don't you remem- oh, fuck it. I'm tired of teaching people crypto history.

Post a link, I'll figure it out, I'm good like that.

But finding that in the morass that is crypto, is damn near impossible.

It'd take several links, as it happened on BCT, but since you seem to care, I'll reiterate. There was a user named girino that made X11 50% faster, put in a 2% fee, and the community basically reverse engineered it, ripped out his code, and put it in sph-sgminer, which then got pulled into SGMiner. It is now called "darkcoin-mod" or "x11mod."

This community will do anything to not pay even a tiny fee - and anything can be reverse engineered, because if a computer can understand it, so can a human. So, the only thing you can do is make it very hard - but fuck up, and all your work will be for nothing. Having to risk that sucks.

It might be better if kernel was only released in binary form (*.bin), and not the source. The binaries could also be signed so if there is a leak, you could tell who has leaked..

Binary distribution would also eliminate the problem with different Catalyst versions. For example current neoscrypt kernel only compiles properly on Catalyst 14.6, anything newer or older generates HW errors or very slow speed. Then again all scrypt kernels compile best on 13.12, compiling on 14.x works but gives perhaps 30% lower hashrate..

The compiled binaries can be moved around with no problem. I have used the 14.6 neoscrypt binary with 13.12, and it works fine with full speed. Recompiling it under 13.12 halves the hashrate..

Still the biggest problem is, some users have 200 gpus and some have only a few. If you are selling a kernel, the price should be different! It should be "per GPU" price, but it is impossible to control. If the price is fixed, some big rig owner can easily pay a few BTC to get the kernel, and then ruin the profits for all "hobby" miners.

I like to optimize these just for my own fun. And obviously it's good if you get some profit from your work. But at the moment it seems the best way to profit, is just keep the kernel to yourself and enjoy the increased hashrate.. Any wider distribution will just lower my own profit.