Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 980. (Read 2347664 times)

6x 750ti cards on kopiemtu USB stick install (Lubuntu 14.04)

@Grout : Great to know!

@Hashbrown : What are you mining with?

ok, will try to compile the latest and see if i can solo mine on lyra2v2. previous versions would crash after 2 minutes, but djm's lyra miner works great. will post back with results.

Thanks, compiling HEAD version did the trick, I was going insane... Cards are reporting 6000kH instead of 6100, but this time, it's real.
Nice of ccminer to display cards temperatures and fan speeds, by the way :-)

RELEASE PACKAGE dot 61 FIXED--

SP_ fixed release dot 61, re-packaged it, and posted the fixed release.  If you can't compile, the zip package works correctly now.  The original release dot 61 was causing me a loss of ~20% at the pool, it does work correctly now.

       --scryptr

The issue with a flat price (subscription or per-release) is that it creates a barrier to entry for people will little hardware. If the miner costs more than you can hope to make with your card, you don't buy it. Big farms, on the other hand, only spend a small fraction of their earnings to pay for the same miner.
That's why I prefer a percentage donation: the more you make, the more you support the developer. That's what I've been trying to do for SP, giving a day's worth of profits every two weeks. When I do this, it's just beers, but if a big farm did the same, it could become a living wage...

Try compiling directly from source not from 1.5.61 There were a fix about that on Quark after .61 released!
I had the same problem and now it everything looks alright

I have a really weird issue since upgrading to release 61. 5*750Ti rig (CentOS6, CUDA6.5), mining Quark: ccminer reports about 6100kH per card, and a total of 30500. But the pool reports about 24000, with 0% rejected shares.

First, I thought it came from the pool, so I tried with Nicehash and Hashpower. Same result on both: 24 MH.

Then, I thought one of the cards was dead and ccminer still reported it working for some reason. Removed one card, ran for 30 minutes on Nicehash, removed the next card, etc... I consistently got 18MH, never mind which card I removed...

Finally, I thought it came from ccminer and tried to run release 60 again. Same results!

I'm at a loss. Does anybody have any idea what could be causing this?


Edit: I tried lyra2v2 to see if it was quark-specific: r61 got about 50% reject with reason: high hash. r60 reports around 19.6MH, pool says 19.2 after 10 minutes. So it IS quark-specific. Any idea SP?

Maybe a subscription service, pay a monthly fee to get newest miners. Or a less desirable but possibly more effective route would be closed source miners that donate a percentage of hashes like the claymore AMD miner for cryptonote algo. I only say less desirable because I like open source. Or just set a price for each new version. I think people are willing to pay but when it's voluntary it's easy to forget or procrastinate or just be greedy.

I don't post much but have donated to the ccminer devs and will continue to, without you guys we couldn't mine 

Sent a couple beers to SP a58c7f816250c31b2f045e963a0e39aaeab8c98212bac77a6d1ee210665a8d99

I disagree, there is no ROI on developing software unless you consider your time. Ideal solution would be designing your own private kernel and then mine with it, that's assuming no one else can come along and top your rates, which effectively nullifies your time you spent developing your kernels. There is also investment in hardware. Putting aside ROI, you need to raise capital for it.

Big companies do everything in house, but we aren't big companies so things get split off because it's not lucrative to try and do everything (poorly) at once.

That aside, I believe the most money is in the exchanges. If you can design a good and well designed exchange and/or pool, it'll net you a decent amount of profit when people start to use you. I haven't actually seen a pool that's a exchange for instance. I think there is definitely still room for innovation here. Pool, exchange, and hashrate rental service all in one. People like when everything is all in one spot... it's very convenient.

As long as you keep things private though, it doesn't change that much for the average user. I don't think trying to get people in your BCT thread to donate is the equivalent of selling software either. As I explained months ago, displaying a product is probably a good place to start and a company could be still formed around this.

Axiom came along and the GPU miner was short lived, there is plenty of room for improvement in XMR. Ethereum looks like it's going to be around awhile and the miner definitely could be improved. Neoscrypt could also be improved and X11 you guys talked about it not having been touched in quite a long time. There is plenty of room for improvement. That's just in the algos, there is also interface and features...


Also, don't sell your AMD cards. Mine Ethereum with them. They're much better at mining then Nvidia cards. I'd highly suggest looking at them before booting them to the curb.

I have been thinking why is that 64-bit binary hashing slower than 32-bit binary.
Tried to tweak around the CUDA C/C++ configuration properties but still having hard time to catch up with 32-bit binary - trailing 200kH/s.

i have to admit it too - that growing a farm ( along with the growing costs ) seems to be the better way to go when it comes to doing anything on a coin level ... and im not just talking about 'profitability' in fiat ...

you have the issue that kernel design and optimization doesnt bring you enough fiat currency as opposed to building a larger farm to make more coin to trade across ...

i have been in the same boat for a long while now - which is why i dont look at it as anything but a way of making 'coin' - rather than 'money' ... and its a delayed satisfaction in terms of trading at a good rate to btc then to fiat ...

choosing the right coin is another issue also - not just the farm - which is why my investment is concentrated on the farm which mines the spread of coins ...

the flipside is - that the short term is a costly sum in fiat ... the long term is anyones guess ... either way its a gamble ... and one i never get over my head in ... ive always got into this with the one 'system' if you will ... NEVER invest more than you are willing to lose ... NEVER ...

the most precious of all this is the miner itself ... the software that drives all this ...

i want to be able to help with 'incentive' to continue development and financial gain to grow - as would most of us here that use the fruits of your ( and other devs ) labor ...

help us help you by giving us more ideas to bring this about ... i need ideas not just from my self - but also from the community and from the devs themselves that can benefit greatly from successful ideas to incentivize the development of opensource to the level that it can not be outperformed very easily ...

what would your ideal way be to incentivize the devs to continue? ... what would be the practical way? ... what can the community and i do to help the devs? ...

just donate? ... buy you equipment? ... or is there more that we as a community can do to help? ...

im back in the office on thursday - im just too damn sick to move around too much at the moment ... so home and bed it is for me ... but this down time has given me thought time - and there must be something more than just a few donations to help you devs produce better optimizations - that in turn help us miners ...

call me sympathetic - call me imaginative - call me sick with delirium ... this 'free' time on my hands has got me thinking a great deal about how little or how much we can all work together ... for the benefit of one another and not a select few ...

ill probably read this back tomorrow and shirk at what ive written but - im all ears with your ideas sp and djm34 and pallas and every other dev that has put a great deal of work into this community ...

im all ears - and also off to bed ...

#crysx

You mentioned adding support for cryptonight mining some time in the future, is this still on the cards? (pardon the pun :p)

Thanks for all your hard work

When you test, you should let the gpu's rest for a while. 5 mins before you start the test, and then run for 5 minutes. But small optimalizations are hard to test, so I usually just look at the instruction count. Less instructions are normally bether.

966 show +250kh on x11 for me. now testing 965
PS. Now on 966 quark +100 kh (49500). It`s very strange...

Could you test x11. I got bether results with 965 on x11.

(Reducing the instruction count but adding const memory access)

Actually, I should split all the x11 kernals from the quark kernals and tune them seperatly, because the performance is not always bether for both of them.

Sorry, yes - my mistake. 965!

Are you mixing 955 with 965?

I think I will revert the 966 commit.

I wonder how gtx 950 performs? Pls share the results when u have a time.Maybe i have possiblity to buy some new cards and i'm thinking to buy some 950's.. )

Perfomance lost in 955 100kh (compared with commit 933), but pool side is not bad. 966 300kh lost, pool side go down 20% stable.
PS. 40 minutes on nicehash speed stable 40-50MH (my 49200 now in miner), 200 kh lower then 955.