CCminer(SP-MOD) Modded GPU kernels.

[bensam1231]

nothing about power consumption in that AMD release article.  which, for miners with > free electricity, means almost everything.

The 390x is rumored to have 300W TDP and they will be released on June 18th and the flagship device, Fury will be released on June 24th. I'll be curious what Wolf0 will crank out of those ones.

It's not going to matter. Wolf0 makes private kernels and sells them to the select few. Good reason if any to stay away from AMD. I was already there and saw that.

And yeah, efficiency...

I'm not interested in AMD mining at all, I'm all for nvidia but it is valuable to know how those cards will run because they could potentially cripple the green team especially because they will be cheap (R9 390X 8GB - $390; R9 390 8GB - $329).

And to be completely honest I wish more people would advertise their private miners because as much as I appreciate sp_, tpruvot, djm34 and other's work their optimizations are being used by everyone because they are free and in reality if everyone has fast miners then nobody really have an advantage.
It might be a good example (or just a coincidence) that since sp_ announced a week ago that he "pushed Quark on the AMD 280x from 1.7 MHASH to 4,5MHASH on standard clocks" the weekly average hashrate on yaamp for quark increased from 572 mh/s to 852 mh/s even though the profit decreased from 0.051 to 0.0398 mbtc/mh/d. And apparently there's a private russian miner which does 11mh/s (280x) and if that was free, quark mining would turn unprofitable very fast. So really, you either have very cheap electiricty, a fast private miner or tons of time to research coins nobody is looking at.

Miners mine mainly for profit so why shouldn't devs prioritize their profits as well? Assuming sp_ haven't changed his donation address he only received 4.6 BTC in the last 9 months meanwhile on yaamp alone currently there are 98 miners (with who knows how many cards per miner) using different SP-MOD ccminer forks. How many of those people ever donated or even knew this thread existed? Sure, private miners have plenty of cons on both sides but I'd certainly like to know the options.

Yup... outside of mining AMD has really good bang for your buck and these new ones sound like killer deals once again.

Also agree about people advertising their private miners and even suggested it to great lengths as small miners could get a idea of entry costs into mining which are currently obfuscated by not having a market you can buy these on. Strange isn't it? We live in a environment of extremely transparent and fluid money, yet what we use to mine it seems to be a extremely guarded secret...

I'm not against paying someone for their work and buying a 'product', but if people don't know it exists or refuse to sell it to everyone... it's kinda... That's what happened with AMD and why I abandon it.

Quoted for emphasis. Everyone should truly consider some sort of crowdfunding. I have a long running joke about people wanting more performance and sort but don't tell anyone (LOL)!

Yup and I mentioned this to both Wolf0 and DJM and they enjoyed a two page rant of making fun of me for suggesting such a thing. To paraphrase, a miner with a built in fee will always be re-engineered to be free, no one donates, and if they sold it so normal miners could afford it, it would get leaked sooner...

I'm not against a miner with a built in fee as well as long as it's reasonable (like 2%).

[Nubminer]

Speaking of one card per instance.  I am actually trying to control that.. I want to run one card per instance... Easy enough for the GPU (using -d ) however i cant seem to control wich core of the cpu processes for a specified GPU.  
Does the CPU affinity command control this and if so how do we utilize it.
Perhaps i can be pointed to the line of code and i can compile/build a custom version for my needs

When running one instance per card, you will need to use -b to specify a different port for each additional instance:
 

Code:

-b 127.0.0.1:5068 

And yes, --cpu-affinity allows you to set the specific core you want that instance to use:

Code:

--cpu-affinity 3

Thanks

[rednoW]

Where is djm34 with his optimized lyra ccminer? )))))))))))))))

[SDFG]

I'm not interested in AMD mining at all, I'm all for nvidia but it is valuable to know how those cards will run because they could potentially cripple the green team especially because they will be cheap (R9 390X 8GB - $390; R9 390 8GB - $329).

And to be completely honest I wish more people would advertise their private miners because as much as I appreciate sp_, tpruvot, djm34 and other's work their optimizations are being used by everyone because they are free and in reality if everyone has fast miners then nobody really have an advantage.
It might be a good example (or just a coincidence) that since sp_ announced a week ago that he "pushed Quark on the AMD 280x from 1.7 MHASH to 4,5MHASH on standard clocks" the weekly average hashrate on yaamp for quark increased from 572 mh/s to 852 mh/s even though the profit decreased from 0.051 to 0.0398 mbtc/mh/d. And apparently there's a private russian miner which does 11mh/s (280x)


11Mh was a long time now 13-15Mh 

[go6ooo1212]

Which are the speeds of r.52 on lyra2re ?

[bensam1231]

Which are the speeds of r.52 on lyra2re ?

970 gets me about 1030

Funny... you're complaining about people not advertising the existence and specifications of private miners, yet as far as I know, I was one of the first in the field to do this, and I continue to do so...

'Bragging' about your miner in various threads and telling people you don't like it's 'private' is not the same thing as advertising that you're selling it. You also don't have a set price for your product do you? That's not the same thing as SP saying he has a miner, telling you the specs, and that it's .1BTC for it. He'll tell you how much it hashes on what cards and how much power it uses. Not 'it runs very cool'.

You could even add these things to your signature... But you don't... why is that? The absolute easiest place to advertise in the world and get attention... Huh...

[rednoW]

970 gets me about 1030

gtx750 gets me 910 )))

[go6ooo1212]

970 gets me about 1030

gtx750 gets me 910 )))

Thanks , mates , I couldn't get more than 820kh/s with my 750TIs.
Which clock do you using with it..

[bensam1231]

970 gets me about 1030

gtx750 gets me 910 )))

That's probably why I don't mine lyra... The GPU usage graph goes up and down in a zig-zag pattern really fast too. I assume there is something wrong with the miner at least for 970s.

Using the TPR fork it doesn't zig-zag, but still only get around 1100 hash.

[rednoW]

Thanks , mates , I couldn't get more than 820kh/s with my 750TIs.
Which clock do you using with it..

1510/1610@1.125v fan 65% temp 51C (room temp 26C)

[fenomenhaa]

Thanks , mates , I couldn't get more than 820kh/s with my 750TIs.
Which clock do you using with it..

1510/1610@1.125v fan 65% temp 51C (room temp 26C)

Is this with Sp released ccminer?

[rednoW]

Is this with Sp released ccminer?

yes. #52 std

[fenomenhaa]

Is this with Sp released ccminer?

yes. #52 std

With SP@52  1350 core /1400 mem 750k hashing what's wrong in my set u?

[rednoW]

With SP@52  1350 core /1400 mem 750k hashing what's wrong in my set u?

memclock maybe to low ... try to increase, lyra algo doesn't stress to much and allows to use frequencies above stable )))
Also check in gpu-z if your card is not throttling due to tdp ...

[fenomenhaa]

With SP@52  1350 core /1400 mem 750k hashing what's wrong in my set u?

memclock maybe to low ... try to increase, lyra algo doesn't stress to much and allows to use frequencies above stable )))
Also check in gpu-z if your card is not throttling due to tdp ...

Already edited to bios to 65w TPD,According to Gpuz Just %30 tpd.. i'll try to incerease mem clock but i'm not sure that all cards can go to 1600 mem clock,Most of cards with samsung vram some hynix , anyway i'll try ans see  ) Thanks for your help.. I had old nv driver ,maybe couse of this.. i'll will update.

[bensam1231]

Is this with Sp released ccminer?

yes. #52 std

With SP@52  1350 core /1400 mem 750k hashing what's wrong in my set u?

Hmmm it'd be nice to get comparable rates out of a 970. I think this is currently the most profitable algo (more the quark) based on your hash rates for you 750ti if it worked properly for everyone else.

[go6ooo1212]

@rednoW
Unfortunately my cards (Palit 750ti stormX) couldn't gain suck OC. Which brand are yours ?

[rednoW]

@rednoW
Unfortunately my cards (Palit 750ti stormX) couldn't gain suck OC. Which brand are yours ?

It was the cheapest thing I was able to find Palit stormx oc 750 with only 1gb of Samsung ddr5. I was lucky or maybe only 1gb mem and only 512 shaders let the card oc so well under std conditions

[rednoW]

I had old nv driver ,maybe couse of this.. i'll will update.

i will certainly recommend 352.86 driver, it works very good for me ...

[sp_]

klaus_T has optimized lyra2 on the gtx970. I merged his changes to my fork. (+60khash) The performance on the 750ti is the same.