CCminer(SP-MOD) Modded GPU kernels.

[tbearhere]

i've heard the hash per watt on a tesla card is dismal. they werent really made for mining, or the miners aren't coded to be efficient on tesla cards

Tesla is Kepler based but a Maxwell version would be interesting. I wonder how long Nvidia
would let someone mine using their test drive...

http://www.nvidia.com/object/gpu-test-drive.html

less than a minute

Did you see an improvement in hashrate djm ?

over what ? I don't remember what I tried to run... (that was last year...)

Ok thx 

[dga]

So some of you guys though even for a moment that the article benchmarking the supposed new dual chip cards on minesweeper was true? LoL!!!!

Such a nice results on solitaire also

Solitaire is serious business

I take it you tried.

Cloud computing services adapted a long time ago and they monitor everything. If you stress the CPU or GPU(s) or network bandwidth too much to their liking they will cut you off.
Unless of course that's exactly what you pay for but that is pretty much the equivalent of burning money unless you're name is David Andersen.

There are people who were much better at it than I.  There are a few folks around who managed in the range of $1mm net profit across a bunch of different altcoins between, say, 2012-2014.

In fairness to the cloud providers:  They're *very* happy about being used for mining if you're paying for their services.  I had a long chat with the director of Amazon's spot market program, and he was interested in knowing how he could better support people doing large-scale mining on Amazon.  Paying customers are paying customers.

But you're right - most of the time it's not profitable, and you have to be very, very careful when you do it, because you *will* get a bill at the end of the month.

[chrysophylax]

i've heard the hash per watt on a tesla card is dismal. they werent really made for mining, or the miners aren't coded to be efficient on tesla cards

i would really like to see a tesla oriented miner and optimizations made for the card ...

i have seen it in action with 3d modelling and rendering - all in real time ... its a powerhouse ...

#crysx

[chrysophylax]

i've heard the hash per watt on a tesla card is dismal. they werent really made for mining, or the miners aren't coded to be efficient on tesla cards

Tesla is Kepler based but a Maxwell version would be interesting. I wonder how long Nvidia
would let someone mine using their test drive...

http://www.nvidia.com/object/gpu-test-drive.html

less than a minute

Did you see an improvement in hashrate djm ?

over what ? I don't remember what I tried to run... (that was last year...)

i wonder how much it would cost to make a 'test' system worth the look-in for devs to have a play with ...

interesting ...

gonna look into this later this week ...

i do know that they are EXPENSIVE units - but if the mining could be specifically designed for the card - then it would be worth doing in my opinion ...

will see what figures we get on costings ...

if it could be done ... would you devs be interested in designing and testing the tesla system? ...

yes - that means djm34 - sp - epsylon and others ... ...

my personal interest would be the x11 optimizations that could be had with the card ... but that MY interest ...

#crysx

[bensam1231]

even if this were completely true AND available right now - it would be a very difficult task to implement such cards in a mining environment due to the logistics of space - cost of cards and cost of consumption of electricity - and shelving space for the machines ...

Entertaining the idea, I don't see your point. If we disregard the obviously super high hashrate per initial card price and the terribly inefficient looking cooling I'm sure there would be a sweet spot (guessing 3 cards per rig or 6 GPUs).
One of these imaginary rigs would do as much as 4 full 750 Ti rigs if we scale the numbers off the Titan X and the density would surely be welcomed by most miners so I don't see any logistics of space issue. On the contrary, less space, less fans and probably less noise, less components so in overall less points of failure.

Yup, they would just be inefficient as far as hash/$. Everything else would be a benefit and even hash/$ can be more thoroughly worked out through overhead of the host machine or machines via density.

Either way, too expensive even if they have the density. 970/980 still have growing issues that aren't ironed out for some algos compared to 750tis, I'm sure this card wouldn't have a better day as it's rarer. Still neat to think about though.

[chrysophylax]

even if this were completely true AND available right now - it would be a very difficult task to implement such cards in a mining environment due to the logistics of space - cost of cards and cost of consumption of electricity - and shelving space for the machines ...

Entertaining the idea, I don't see your point. If we disregard the obviously super high hashrate per initial card price and the terribly inefficient looking cooling I'm sure there would be a sweet spot (guessing 3 cards per rig or 6 GPUs).
One of these imaginary rigs would do as much as 4 full 750 Ti rigs if we scale the numbers off the Titan X and the density would surely be welcomed by most miners so I don't see any logistics of space issue. On the contrary, less space, less fans and probably less noise, less components so in overall less points of failure.

Yup, they would just be inefficient as far as hash/$. Everything else would be a benefit and even hash/$ can be more thoroughly worked out through overhead of the host machine or machines via density.

Either way, too expensive even if they have the density. 970/980 still have growing issues that aren't ironed out for some algos compared to 750tis, I'm sure this card wouldn't have a better day as it's rarer. Still neat to think about though.

would density 'really' be the main issue? ...

i ask purely on the basis that we have just put an order in for another 18 x gigabyte 750ti oc lp cards and they are small cards ...

BUT we can only fit 6 to a single motherboard at a time ( unless of course we go with the larger server motherboards - some having up to 12 x gpu in one system ) ...

so density really isnt an issue here - but i think you are spot on with the fiat cost ...

power an heat displacement are the main factors for what we have here - but also density if we were to try and pack in as many in a smaller space ...

i guess im speaking in circles at the moment - as im unsure which way to go with the farm ...

its a large nvidia to amd split at the moment and judging by what is going on in the market - this doesnt look like its going to change in the very near future ... especially concerning power draw ...

in terms of the higher grade cards - the 980ti seems to be getting more attention at the moment - and what seems to be more focus for development ...

i guess we will see over the next 6-12 months ...

may stick with the 750ti cards for the time being - and have an exit plan for an upgrade at a later date to the 980ti when its a reasonable price ... i cant see amd coming out with a low power card in the very near future - unless ive missed something here ...

anyway - back to work for me ...

#crysx

[sp_]

my personal interest would be the x11 optimizations that could be had with the card ... but that MY interest ...

The price of X11 is going up  Nicehash is now paying 0.5125BTC/GHASH. Last year they only payed 0.13BTC/GHASH.

Makes it worth it optimize, but getting more and more difficult.. The easy pickings are done.

[bensam1231]

even if this were completely true AND available right now - it would be a very difficult task to implement such cards in a mining environment due to the logistics of space - cost of cards and cost of consumption of electricity - and shelving space for the machines ...

Entertaining the idea, I don't see your point. If we disregard the obviously super high hashrate per initial card price and the terribly inefficient looking cooling I'm sure there would be a sweet spot (guessing 3 cards per rig or 6 GPUs).
One of these imaginary rigs would do as much as 4 full 750 Ti rigs if we scale the numbers off the Titan X and the density would surely be welcomed by most miners so I don't see any logistics of space issue. On the contrary, less space, less fans and probably less noise, less components so in overall less points of failure.

Yup, they would just be inefficient as far as hash/$. Everything else would be a benefit and even hash/$ can be more thoroughly worked out through overhead of the host machine or machines via density.

Either way, too expensive even if they have the density. 970/980 still have growing issues that aren't ironed out for some algos compared to 750tis, I'm sure this card wouldn't have a better day as it's rarer. Still neat to think about though.

would density 'really' be the main issue? ...

750ti require a host machine to run. If they scaled appropriately (which I doubt they ever will), one of these cards would be the equivalent of like 8 750tis.

So that means you have to build eight systems to get the same amount of hash rate as one system with these cards on it. That would be your system overhead cost. Eight systems is quite a chunk of change, why I bought 970s instead of 750tis. A 970 is about three 750tis depending on the algo. Even though these would offer almost double the density of my 970s, they still wouldn't be worth it.

Heat can be dealt with with fans and ventilation, that's not a problem. You don't need AC to deal with it if it's done right either.

my personal interest would be the x11 optimizations that could be had with the card ... but that MY interest ...

The price of X11 is going up  Nicehash is now paying 0.5125BTC/GHASH. Last year they only payed 0.13BTC/GHASH.

Makes it worth it optimize, but getting more and more difficult.. The easy pickings are done.

Yup, Quark is going down too. I sense a shift of balance in the force. It's almost worth mining x11... almost... That dashcoin. Still miss YAAMP though.

Conversely, perhaps you've looked at it this way already, instead of increasing hash, have you tried reducing power usage instead? Not quite the same thing, but still very useful.

BTW if you guys are looking for a good pool for dash, definitely check out suchpool. I've been mining on there off and on since the beginning and even after their fall, very stable, fair, and reliable service.

[chrysophylax]

my personal interest would be the x11 optimizations that could be had with the card ... but that MY interest ...

The price of X11 is going up  Nicehash is now paying 0.5125BTC/GHASH. Last year they only payed 0.13BTC/GHASH.

Makes it worth it optimize, but getting more and more difficult.. The easy pickings are done.

and one of the reasons why i have been asking for x11 optimizations since we took over granitecoin ...

definitely worth donating hash there for you ...

which we will anyway ...

#crysx

[chrysophylax]

even if this were completely true AND available right now - it would be a very difficult task to implement such cards in a mining environment due to the logistics of space - cost of cards and cost of consumption of electricity - and shelving space for the machines ...

Entertaining the idea, I don't see your point. If we disregard the obviously super high hashrate per initial card price and the terribly inefficient looking cooling I'm sure there would be a sweet spot (guessing 3 cards per rig or 6 GPUs).
One of these imaginary rigs would do as much as 4 full 750 Ti rigs if we scale the numbers off the Titan X and the density would surely be welcomed by most miners so I don't see any logistics of space issue. On the contrary, less space, less fans and probably less noise, less components so in overall less points of failure.

Yup, they would just be inefficient as far as hash/$. Everything else would be a benefit and even hash/$ can be more thoroughly worked out through overhead of the host machine or machines via density.

Either way, too expensive even if they have the density. 970/980 still have growing issues that aren't ironed out for some algos compared to 750tis, I'm sure this card wouldn't have a better day as it's rarer. Still neat to think about though.

would density 'really' be the main issue? ...

750ti require a host machine to run. If they scaled appropriately (which I doubt they ever will), one of these cards would be the equivalent of like 8 750tis.

So that means you have to build eight systems to get the same amount of hash rate as one system with these cards on it. That would be your system overhead cost. Eight systems is quite a chunk of change, why I bought 970s instead of 750tis. A 970 is about three 750tis depending on the algo. Even though these would offer almost double the density of my 970s, they still wouldn't be worth it.

Heat can be dealt with with fans and ventilation, that's not a problem. You don't need AC to deal with it if it's done right either.

my personal interest would be the x11 optimizations that could be had with the card ... but that MY interest ...

The price of X11 is going up  Nicehash is now paying 0.5125BTC/GHASH. Last year they only payed 0.13BTC/GHASH.

Makes it worth it optimize, but getting more and more difficult.. The easy pickings are done.

Yup, Quark is going down too. I sense a shift of balance in the force. It's almost worth mining x11... almost... That dashcoin. Still miss YAAMP though.

Conversely, perhaps you've looked at it this way already, instead of increasing hash, have you tried reducing power usage instead? Not quite the same thing, but still very useful.

BTW if you guys are looking for a good pool for dash, definitely check out suchpool. I've been mining on there off and on since the beginning and even after their fall, very stable, fair, and reliable service.

some good points there bensam ...

i think in the event that the cost of the 980ti ( or equivalent ) drops to a reasonable level - that will be when we move across to them and upgrade ...

till then - the systems that are currently our farm will stay - mainly due to the cheap outlay of the systems ...

we have been mining x11 since its inception - various coins and various pools ... suchpool is one of the ones we mainly mined on with dedicated pool ( before they closed ) and coinking ( which is a ghost town at the moment - missing / quiet devs ) as part of the rotating pool list ...

but x11 has always had a soft spot in our mining heart ...

...

#crysx

[hashbrown9000]

what are you guys getting on X11 with the EVGA Superclocked 750ti ?

[joblo]

what are you guys getting on X11 with the EVGA Superclocked 750ti ?

I get about 3070 KH/s.

[DirtyUniverse]

what are you guys getting on X11 with the EVGA Superclocked 750ti ?

I get about 3070 KH/s.

same for me, can you post your configs. you have any optimization for GPU or Software?

[sp_]

Use ccminer sp-mod release 53. Run with -g 2. should be closer to 3,2MHASH. But it uses 2x the memory and a little more power around 45-50watt.

[sp_]

fast(er) lyra algo has been released:

windows binaries: https://mega.co.nz/#!5ZkVmLZb!qgTTz4OJFsAtJNLqwYqf6mGvaSWwHiVMBoDUlOk1sDc
source code: https://github.com/djm34/ccminer-lyra

1140kh/s on 750ti
2500kh/s on 980
2900kh/s on 780ti (and 3.3MH/s if you are lucky enough to screw your compilation... got that once, mostly random  )

Just added some of the modules to my fork. Lyra is more than 40% faster.

Nicely done by the Cuda expert djm34

[djm34]

fast(er) lyra algo has been released:

windows binaries: https://mega.co.nz/#!5ZkVmLZb!qgTTz4OJFsAtJNLqwYqf6mGvaSWwHiVMBoDUlOk1sDc
source code: https://github.com/djm34/ccminer-lyra

1140kh/s on 750ti
2500kh/s on 980
2900kh/s on 780ti (and 3.3MH/s if you are lucky enough to screw your compilation... got that once, mostly random  )

Just added some of the modules to my fork. Lyra is more than 40% faster.

Nicely done by the Cuda expert djm34

thanks, it goes up to 90% faster for the big guns

[bensam1231]

fast(er) lyra algo has been released:

windows binaries: https://mega.co.nz/#!5ZkVmLZb!qgTTz4OJFsAtJNLqwYqf6mGvaSWwHiVMBoDUlOk1sDc
source code: https://github.com/djm34/ccminer-lyra

1140kh/s on 750ti
2500kh/s on 980
2900kh/s on 780ti (and 3.3MH/s if you are lucky enough to screw your compilation... got that once, mostly random  )

Just added some of the modules to my fork. Lyra is more than 40% faster.

Nicely done by the Cuda expert djm34

thanks, it goes up to 90% faster for the big guns

Appreciate it... Hopefully Lyra will have a upturn again.

[sp_]

My gtx 970 is up 60% in lyra nice.
did you change some other kernals as well? I just merged the lyra kernal and kept the others..

[djm34]

My gtx 970 is up 60% in lyra nice.
did you change some other kernals as well? I just merged the lyra kernal and kept the others..

cuda_vector.h might have changed as well
btw you also need to change opt_scantime in miner_thread and put it to 60 (or something like that)  gives more stable mining in many algo

[scryptr]

My gtx 970 is up 60% in lyra nice.
did you change some other kernals as well? I just merged the lyra kernal and kept the others..

cuda_vector.h might have changed as well
btw you also need to change opt_scantime in miner_thread and put it to 60 (or something like that)  gives more stable mining in many algo

@SP_,  for Lyra2 Merge--

Please get it to a Windows release.  I think it is time for a new round of donations!

@DJM34--  I keep getting "Cuda error in func 'scanhash_Lyra2' at line 93 : out of memory.", on launch.  This is on NiceHash with "--diff 1", using a 2GB GTX 960 SSC on Win 7 x64, and your latest binary.  The Win 7 system has 8GB of memory.

Thanks!       --scryptr