CCminer(SP-MOD) Modded GPU kernels.

[pallas]

For you, guys. I have a screen of x17 sp mod.
Not bad but not one would expect for 0.8 btc ))))
All the trick is to mod mem allocation to support higher intensities ))

higher intensity means more rejected shares.
never trade 1% of hashrate for crazy high intensity.

[rednoW]

For you, guys. I have a screen of x17 sp mod.
Not bad but not one would expect for 0.8 btc ))))
All the trick is to mod mem allocation to support higher intensities ))

higher intensity means more rejected shares.
never trade 1% of hashrate for crazy high intensity.

alexis and tpruvot's miners (tpruvot uses one of sp_'s kernels for blake512_80) won't work with intensity higher than 21.3 on 10x0 series. Error  in buffer allocation routine. So sp_ in his private mod changed something to avoid it. Maybe larger number for second param in launch config ... don't know ... int math and gpu-programming is difficult for me as I always worked with simulations with double type.

I see that alexis uses 128 registers (reg count) in default config but tpruvot uses less (80 as far as I remember) with some custom numbers for some algos. What is your recommendation for this setting? Does it really have affect on performance?

[sp_]

higher intensity means more rejected shares.
never trade 1% of hashrate for crazy high intensity.

i 24 is not a high intensity. blake2s and blake256 is using -i 31 without any rejects or problems

[pallas]

For you, guys. I have a screen of x17 sp mod.
Not bad but not one would expect for 0.8 btc ))))
All the trick is to mod mem allocation to support higher intensities ))

higher intensity means more rejected shares.
never trade 1% of hashrate for crazy high intensity.

alexis and tpruvot's miners (tpruvot uses one of sp_'s kernels for blake512_80) won't work with intensity higher than 21.3 on 10x0 series. Error  in buffer allocation routine. So sp_ in his private mod changed something to avoid it. Maybe larger number for second param in launch config ... don't know ... int math and gpu-programming is difficult for me as I always worked with simulations with double type.

I see that alexis uses 128 registers (reg count) in default config but tpruvot uses less (80 as far as I remember) with some custom numbers for some algos. What is your recommendation for this setting? Does it really have affect on performance?

alexis can do >21 intensity.
it's tpruvot that can't, because of old simd code.
about regs, usually it's set by the launch config, not maxregs.

[pallas]

higher intensity means more rejected shares.
never trade 1% of hashrate for crazy high intensity.

i 24 is not a high intensity. blake2s and blake256 is using -i 31 without any rejects or problems

you need to relate that to the hashrate.
blake is much faster so the the gpu returns a lot quicker.

[bensam1231]

There are a lot of miner developers. They just don't sell to you.

Heh. All this time, and he still ain't realized venom doesn't attract folks.

Heh, you haven't apparently realized that your singular experience doesn't apply to all interactions from an individual you don't actually talk to. Also money doesn't seem to either.

I guess the clever retort here is 'Nor does butt raping the majority of miners by selling to a handful of giant farms', which you've done when it really mattered and may have something to do with your interaction with myself.

But you know I'm not interested in sucking other peoples dicks in order to have a 'proper' business relationship. There are plenty around here that are willing to do that though.

[disoman]

The x17 mod is best on the gtx 1060. The gtx 1060 aurus does 9MHASH with oc. the 1060 3gb cards are normally around 8MHASH
It is proven to be faster on all the other models as well. At least +5% faster than the alexis miner.

i want MOD C11 ? how to do ?

[sp_]

you need to relate that to the hashrate.
blake is much faster so the the gpu returns a lot quicker.

some pools accept stale shares. f.eks miningpoolhub.com so no work is lost. for coins with a fast blocktime you can change the scantime with the -s parameter

[Saingo]

For you, guys. I have a screen of x17 sp mod.
Not bad but not one would expect for 0.8 btc ))))
All the trick is to mod mem allocation to support higher intensities ))
https://i.imgur.com/sDq0dHX.png

Hi would you like to sell me this mod?

[sp_]

2126 MHz.  That's not what most people have in mind when they talk about gtx 1060s at PL 70%. I guess it's only possible in MSI land, with their high stock TDP.
Could you elaborate on "+ all  (/Ot) &  (/O2)" for us noobs, what is this?

if you remove the crazy clock and use 1913Mhz like the normal gtx 1060 6gb cards can do you end up with a hashrate below 8MHASH. That's slower than my mod does on the 1060 3gb cards with the same clock. You 70% tdp is probably because your card has a 8pin power connector (150watt of power) and other cards only have 6 pin power connector (75watt power)

[rednoW]

For you, guys. I have a screen of x17 sp mod.
Not bad but not one would expect for 0.8 btc ))))
All the trick is to mod mem allocation to support higher intensities ))

Hi would you like to sell me this mod?

No, I don't have it, only screen.
Second, you'll get 22.2mhs with normal intensity with public available miner at this high clocks like on the screen.
So, this is useless ... totally
SP_ scammed again his buyers.

[rednoW]

2126 MHz.   That's not what most people have in mind when they talk about gtx 1060s at PL 70%. I guess it's only possible in MSI land, with their high stock TDP.
Could you elaborate on "+ all  (/Ot) &  (/O2)" for us noobs, what is this?

if you remove the crazy clock and use 1913Mhz like the normal gtx 1060 6gb cards can do you end up with a hashrate below 8MHASH. That's slower than my mod does on the 1060 3gb cards with the same clock. You 70% tdp is probably because your card has a 8pin power connector (150watt of power) and other cards only have 6 pin power connector (75watt power)

your x17 sp mod is only 0.5mhs faster at gtx1080ti thanks for higher intensity your miner allows. But higher intensity means instability and worse results on the pool. So it is useless.
Your words about your speed with gtx1060 are only words, no proof given.

[sp_]

Mining x17 with the 1080ti is stupid. The 1080ti is only 2.5 times faster than the 1060. If you instead mine Lux, Bitcore, tribus or Keccak with the sp-mod you get more than 3x the speed of a 1060 and nice returns.


The 1080ti is also slow in the  Cryptonight, ethereum and a number of other algos. My modded cryptonight miner is only doing 1MHASH on the 1080ti and 0.6MHASH on the 1060, so that's ony 1.6 times the speed of a 1060 card. The 1060 6gb is also almost identical to the P104-6gb mining card, and with my mod and 10000 cards ++ the 1080ti is outperformed and not profitable in this algo.

[rednoW]

Mining x17 with the 1080ti is stupid. The 1080ti is only 2.5 times faster than the 1060. If you instead mine Lux, Bitcore or Keccak with the sp-mod you get more than 3x the speed of a 1060.


The 1080ti is also slow in the  Cryptonight, ethereum and a number of other algos. My modded cryptonight miner is only doing 1MHASH on the 1080ti and 0.6MHASH on the 1060, so that's ony 1.6 times the speed of a 1060 card.

we are talking about speed of your miner comparing to public ones. Your 0.8btc miner suxx hard on gtx1080ti, and you even didn't provide proof of gtx1060 speed with x17mod.
Don't tell me whattomine with my cards ))

[sp_]

As I have stated before, my x17 mod is at least 5% faster on all models compute 5.0 or newer (750,750ti,950,960,970,980,980ti,1050,1050ti,1060 3gb, 1060 6gb,1070,1070ti, 1080,1080ti,P102,P104,P106), but often more than that. More work can be done to improve the performance on specific models, and I am working on a x16r mod that could result in a faster x17 mod update.

[rednoW]

x17 kernels are well-optimized by alexis and previous contributors including you. x16r new 80-bit kernels are SLOWWWWW and need to be rewritten. I think krnlx or palgin could do that but not you ))

[pallas]

you need to relate that to the hashrate.
blake is much faster so the the gpu returns a lot quicker.

some pools accept stale shares. f.eks miningpoolhub.com so no work is lost. for coins with a fast blocktime you can change the scantime with the -s parameter

A pool accepting stale shares doesn't matter: if everybody submits stale shares, the coins you get are the same than if the pool didn't accept them, because stale shares don't produce blocks.
If intensity is too high, scantime doesn't matter (because the work on gpu will take more time that scantime itself); and besides it only works for solo mining, not stratum.

[sp_]

x16r new 80-bit kernels are SLOWWWWW and need to be rewritten.

The 80bit kernels are only used once (1/16) of the time of a hash. To improve the speed you need to optimize all the kernels.

[rednoW]

x16r new 80-bit kernels are SLOWWWWW and need to be rewritten.

The 80bit kernels are only used once (1/16) of the time of a hash. To improve the speed you need to optimize all the kernels.

did you ever look at the code?

[sp_]

But there is a lot of miners in the ravencoin. Yiimp.eu just opened up a new port on the server.

Yeah, which makes it even less likely to earn big when it hits exchanges. Profit from mining unlisted coins involves luck and obscurity, Ravencoin is no longer obscure.

I don't feel so lucky with this one... 

There is still not a single exchange on this coin..