CCminer(SP-MOD) Modded GPU kernels.

[sp_]

Agreed. In his efforts to advertise higher hashrates for certain algos on certain cards he has detsabilized
other algos and other cards.

I run my farm on my modded and optimized kernals and I make a profit. If you get crashes it might be because your powersupply is not good enough. Higher intensities can increase the power usage with 5%.

[tbearhere]

Agreed. In his efforts to advertise higher hashrates for certain algos on certain cards he has detsabilized
other algos and other cards.

I run my farm on my modded and optimized kernals and I make a profit. If you get crashes it might be because your powersupply is not good enough. Higher intensities can increase the power usage with 5%.

I don't use intensities!
And I bet your farm doesn't use 2 core cpu's.
And has nothing to do with psu because I took that card out... only using 650 watts on 1300 watt psu.

[joblo]

Agreed. In his efforts to advertise higher hashrates for certain algos on certain cards he has detsabilized
other algos and other cards.

I run my farm on my modded and optimized kernals and I make a profit. If you get crashes it might be because your powersupply is not good enough. Higher intensities can increase the power usage with 5%.

Wrong. You've optimized your kernels for your farm and tell your paying customers to expect the same results with different configs.

[Amph]

i think sp is too much worried about his roi right, now, i think you should relax sp, even if etehreum anc other crap coin, will be unprofitable, other good profitable time will come, you will eventually roi one day

[tbearhere]

Agreed. In his efforts to advertise higher hashrates for certain algos on certain cards he has detsabilized
other algos and other cards.

I run my farm on my modded and optimized kernals and I make a profit. If you get crashes it might be because your powersupply is not good enough. Higher intensities can increase the power usage with 5%.

Wrong. You've optimized your kernels for your farm and tell your paying customers to expect the same results with different configs.

Exactly joblo exactly.

And I solved the problem .... I upped the time for switching algo's to 12 seconds...10 seconds it crashed yesterday after 4 hrs  .... why this works I don't know ....on my 2 core rigs only I have this problem.
Both 2 core rigs are stable now.
This was going on for 6 months. And now to see if it is stable with light oc'ing.

[tbearhere]

Agreed. In his efforts to advertise higher hashrates for certain algos on certain cards he has detsabilized
other algos and other cards.

I run my farm on my modded and optimized kernals and I make a profit. If you get crashes it might be because your powersupply is not good enough. Higher intensities can increase the power usage with 5%.

Wrong. You've optimized your kernels for your farm and tell your paying customers to expect the same results with different configs.

You're damned right, joblo - further, his time is not spent writing elegant/clean code, nor infrastructure that isn't buggy. I like tpruvot's work often, because he pays attention to shit. He wouldn't do like sp_ and be like, "Well, it might crash once every two days, but who the hell cares."

Exactly Wolf0... tpruvot's, and djm's work are always stable on all my rigs.... never tried yours because you do AMD but I know your work is good too.

@Sp if your doing a final build .. qubit needs intensity setting updated sp5 private.
On this computer only I have to tap into one card and and Lower the intensity to be able to use this....  so I put  -d 0,1,2 -i 16 and it lowers all cards hashrates... it should be  -d 0,1,2 -i 0,16,0 but that doesn't work. I'm tapping into card 1 or device 1. On all other algo's setting intensity to each card works like this -d 0,1,2 -i 0,16,0 or like this -d 0,1,2 -i 0,14,0 and works great. Setting is dependent on the algo.

[tbearhere]

Agreed. In his efforts to advertise higher hashrates for certain algos on certain cards he has detsabilized
other algos and other cards.

I run my farm on my modded and optimized kernals and I make a profit. If you get crashes it might be because your powersupply is not good enough. Higher intensities can increase the power usage with 5%.

Wrong. You've optimized your kernels for your farm and tell your paying customers to expect the same results with different configs.

You're damned right, joblo - further, his time is not spent writing elegant/clean code, nor infrastructure that isn't buggy. I like tpruvot's work often, because he pays attention to shit. He wouldn't do like sp_ and be like, "Well, it might crash once every two days, but who the hell cares."

Exactly Wolf0... tpruvot's, and djm's work are always stable on all my rigs.... never tried yours because you do AMD but I know your work is good too.

@Sp if your doing a final build .. qubit needs intensity setting updated sp5 private.
On this computer only I have to tap into one card and and Lower the intensity to be able to use this....  so I put  -d 0,1,2 -i 16 and it lowers all cards hashrates... it should be  -d 0,1,2 -i 0,16,0 but that doesn't work. I'm tapping into card 1 or device 1. On all other algo's setting intensity to each card works like this -d 0,1,2 -i 0,16,0 or like this -d 0,1,2 -i 0,14,0 and works great. Setting is dependent on the algo.

I started doing CUDA - likely those builds have at least some of my code in them :3

If sp quits will you do pascal?

[pallas]

I might do some.

[Amph]

that's comforting, at least we will not be left without hope, but it is yet to be seen if pascal will be a real game changer, or just vaporware

i would be happy with a 2x increase with the same comsuption, not the other way around...

[joblo]

that's comforting, at least we will not be left without hope, but it is yet to be seen if pascal will be a real game changer, or just vaporware

i would be happy with a 2x increase with the same comsuption, not the other way around...

I'm hoping cuda 8 will be better than 7.5 out of the box. Most of the 7.5 work has been to recover lost performance
from 6.5. I'm also hoping pascal will perform well with existing code compiled with 7.5 or 6.5. Then the pacal
optimizing can begin.

[bensam1231]

Wolf0 charges a arm and a leg and Pallas codes on occasion... Not looking too hot if SP quits.

[sp_]

My git is still active. And I am working on a refork, but it takes time. Too busy at work.

[sp_]

@Sp if your doing a final build .. qubit needs intensity setting updated sp5 private.
On this computer only I have to tap into one card and and Lower the intensity to be able to use this....  so I put  -d 0,1,2 -i 16 and it lowers all cards hashrates... it should be  -d 0,1,2 -i 0,16,0 but that doesn't work. I'm tapping into card 1 or device 1. On all other algo's setting intensity to each card works like this -d 0,1,2 -i 0,16,0 or like this -d 0,1,2 -i 0,14,0 and works great. Setting is dependent on the algo.

I use 8gb memory on all my 6card rigs. And you should add 8-16 gb of virtual memory as well. High intensity and 6 cards require alot of system memory. When you add virtual memory remember to reeboot to get the  effect. Windows 10 is no good. Use windows 8.1 or windows 7. I have run for weeks without a reeboot or crash in the sp-mod private, but I use a failover script/ miner control software as well.

[Longsnowsm]

I am in the process of building a 6 gpu rig with some 750ti's.  SP_ can you tell me what your hash rates are for the various algo's with your private builds?  I would love to get an idea where I can get the best bang for my buck with these 750ti's and CCminer.  The current plan is to use Linux, but looking for recommendations on the OS.  Is there a preference on which Linux distro works best for CCminer?  I see in the last few posts something about going with more memory for the 6 gpu rigs.  I bought 8gb, but sounds like I should get 8gb more?  Anything else I need to optimize the rig?  It will be running an i3 Haswell cpu, 128gb SSD, 1000w EVGA Platinum P2 psu, 6 Gigabyte GTX 750ti's.  Thoughts and feedback would be great on how to make these cards sing in this rig.  I am leaning toward Neoscrypt at the moment for this rig, but I would love to have some hash rates, and power consumption figures so I can do the comparisons.  Thanks!

[sp_]

750ti cards have a coreclock of around 1100-1400 and the hashrate depends on the clock.

Optimal settings(oc) (core of around 1390)

Quark: 7mhash
Qubit: 5mhash
Lyra2v2: 5.3 mhash
X11: 3.45mhash
X13:2.6mhash
X15:2.4Mhash
Lyra2re: 1.2mhash
Neoscrypt: 240mhash (cuda 6.5 build)
Decred: 640mhash
Vcash: 1000mhash
Pentablake:32 mhash

The algos use from 40-60 watt per card if you have a platinum powersupply. You can run a 6x 750ti rig on a 500w psu, but I recomend 700+. To get the best hashrates you need to increase the tdp of your card from 38w to 65w. In the Bios or use the nvsmi tool


I normally don't release linux versions, only windows. But the decred and vcash sourcecode is available for a extra fee

[sp_]

I'm hoping cuda 8 will be better than 7.5 out of the box. Most of the 7.5 work has been to recover lost performance from 6.5. I'm also hoping pascal will perform well with existing code compiled with 7.5 or 6.5. Then the pacal
optimizing can begin.

Compute 6.0 will probobly need some re-tuning. A new instructionset? Or just the maxwell shrinked to 16nm..

[tbearhere]

@Sp if your doing a final build .. qubit needs intensity setting updated sp5 private.
On this computer only I have to tap into one card and and Lower the intensity to be able to use this....  so I put  -d 0,1,2 -i 16 and it lowers all cards hashrates... it should be  -d 0,1,2 -i 0,16,0 but that doesn't work. I'm tapping into card 1 or device 1. On all other algo's setting intensity to each card works like this -d 0,1,2 -i 0,16,0 or like this -d 0,1,2 -i 0,14,0 and works great. Setting is dependent on the algo.

I use 8gb memory on all my 6card rigs. And you should add 8-16 gb of virtual memory as well. High intensity and 6 cards require alot of system memory. When you add virtual memory remember to reeboot to get the  effect. Windows 10 is no good. Use windows 8.1 or windows 7. I have run for weeks without a reeboot or crash in the sp-mod private, but I use a failover script/ miner control software as well.

I have them maxed out with Hyperx 16 gig.... what exactly do you mean by adding virtual memory and how to do it plz?

[sp_]

Google it

http://www.tech-recipes.com/rx/36092/windows-7-and-8-change-the-size-of-virtual-memory/

[tbearhere]

Google it

http://www.tech-recipes.com/rx/36092/windows-7-and-8-change-the-size-of-virtual-memory/

K reading thanks.
EDIT:I already did that but looking again.
ps lower the intensity on the next build...

[sp_]

Are you using gtx 950 or gtx 960 cards?

Defauls intensity qubit:

Gtx 970,980,980ti: -X 24
Gtx 750 ti: -X 12
Gtx 750: -X 10
Gtx 950,960: -X 23


                if (strstr(props.name, "970"))
      {
         intensity = 256 * 256 * 24;
      }
      else if (strstr(props.name, "980"))
      {
         intensity = 256 * 256 * 24;
      }
      else if (strstr(props.name, "750 Ti"))
      {
         intensity = 256 * 256 * 12;
      }
      else if (strstr(props.name, "750"))
      {
         intensity = 256 * 256 * 10;
      }
      else if (strstr(props.name, "960"))
      {
         intensity = 256 * 256 * 23;
      }
      else if (strstr(props.name, "950"))
      {
         intensity = 256 * 256 * 23;
      }
      throughput = device_intensity(device_map[thr_id], __func__, intensity);