[ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX]

[sp_]

Nice job on Lyra DJM34. The uint2 Rotate can be used to speedup many of the algo's in the X series as well.


Blake512,BMW512,Keccak512  etc. I will try to have a look toninght.

[Amph]

hi!!!

it is possible to know if cudaMiner implements load balance?
i need it to use one gpu, 970, over more accounts/miner instance!!

thanks!!!

just run multiple istance

ccminer will split the hash

hi!!!

it is possible to know if cudaMiner implements load balance?
i need it to use one gpu, 970, over more accounts/miner instance!!

thanks!!!

it may work as long as you have enough vram, but if you run out of memory expect a crash...
(also it is't obvious if the algo are too different... was testing lyra while running on something else, the other miner did crash)

4gb should be plenty of ram

[djm34]

hi!!!

it is possible to know if cudaMiner implements load balance?
i need it to use one gpu, 970, over more accounts/miner instance!!

thanks!!!

just run multiple istance

ccminer will split the hash

hi!!!

it is possible to know if cudaMiner implements load balance?
i need it to use one gpu, 970, over more accounts/miner instance!!

thanks!!!

it may work as long as you have enough vram, but if you run out of memory expect a crash...
(also it is't obvious if the algo are too different... was testing lyra while running on something else, the other miner did crash)

4gb should be plenty of ram

you may need to watch FB (frame buffer) usage too... if it is goes higher than 100%... well it cant' 
(if the FB is saturated, you will get lags if the card is connected to a monitor)

[djm34]

Nice job on Lyra DJM34. The uint2 Rotate can be used to speedup many of the algo's in the X series as well.


Blake512,BMW512,Keccak512  etc. I will try to have a look toninght.

That improves for sure keccak and skein, but in x11 context, you won't see almost anything as these algo are already pretty fast.
The main problem with uint2 is that it does not work at all with compute 30 (the logic should work but it seems there is a problem with register allocation...)

[bathrobehero]

Sorry to see that.

Just to add one more thing, do not use more than 2 on each daisy chain. Those daisy chains aren't designed to pull a lot of power through them.

Yeah, even on a quality PSU if I put 3 cards on one chain the cable gets warm after a while. I'm sticking with these cables, have dozens of them but I'll be limiting the TDP for the most power hungry algos that's for sure.

If anyone still decides to invest in 750 Ti's for mining for some reason, do not buy cards which doesn't have 6-pin power inputs. I have both MSI and ASUS 6-pin-less cards and they are a pain compared to Gigabyte cards with 6-pin.

had the same problem with one , however it happened at the other end of the cable (and there was only one gpu on the whole psu, my 780ti and obviously nothing else on that cable).
I think the culprit are those small cables.  
Best move: remove those piece of crap and plug the riser directly into one molex of the psu.
(and as you can see it has nothing to do with the presence of 6pin or not...)

My PSU's have 2 x molex chains but if I put 3 cards on one chain the cable gets warm and hot with 4. But if one or two of the cards have 6-pins on the same cable, it remains cold which led me to believe cards barely pull anything from the molex when they can pull from the 6-pin instead.

[djm34]

Sorry to see that.

Just to add one more thing, do not use more than 2 on each daisy chain. Those daisy chains aren't designed to pull a lot of power through them.

Yeah, even on a quality PSU if I put 3 cards on one chain the cable gets warm after a while. I'm sticking with these cables, have dozens of them but I'll be limiting the TDP for the most power hungry algos that's for sure.

If anyone still decides to invest in 750 Ti's for mining for some reason, do not buy cards which doesn't have 6-pin power inputs. I have both MSI and ASUS 6-pin-less cards and they are a pain compared to Gigabyte cards with 6-pin.

had the same problem with one , however it happened at the other end of the cable (and there was only one gpu on the whole psu, my 780ti and obviously nothing else on that cable).
I think the culprit are those small cables. 
Best move: remove those piece of crap and plug the riser directly into one molex of the psu.
(and as you can see it has nothing to do with the presence of 6pin or not...)

My PSU's have 2 x molex chains but if I put 3 cards on one chain the cable gets warm and hot with 4. But if one or two of the cards have 6-pins on the same cable, it remains cold which led me to believe cards barely pull anything from the molex when they can pull from the 6-pin instead.

actually, it is the inverse (but it is a cheap 2rails psu, don't want to buy a new psu every 6 months...) for me, the 6+8pin of the 780ti is rather cold and the cable powering the riser is rather warm/hot

[sp_]

That improves for sure keccak and skein, but in x11 context, you won't see almost anything as these algo are already pretty fast.  The main problem with uint2 is that it does not work at all with compute 30 (the logic should work but it seems there is a problem with register allocation...)

The uint2 keccak made x11 20-25 KHASH faster. on a 750ti

[djm34]

That improves for sure keccak and skein, but in x11 context, you won't see almost anything as these algo are already pretty fast.  The main problem with uint2 is that it does not work at all with compute 30 (the logic should work but it seems there is a problem with register allocation...)

The uint2 keccak made x11 20-25 KHASH faster. on a 750ti

that's what I call "almost anything"

[ivanlabrie]

Guys, cheap psu and psu cable adapters are a big no no xD
They like to catch fire at odd moments...
How is everyone doing?
No more farm administration for me...Investors sold all the Nvidia and AMD cards :/

[Dotcommie]

Guys, cheap psu and psu cable adapters are a big no no xD
They like to catch fire at odd moments...
How is everyone doing?
No more farm administration for me...Investors sold all the Nvidia and AMD cards :/

Yeah I'm about to sell off a rig or two of mine.  Especially because it has randomly been crashing since I've tried to use Miningcontrol and autoswitching.  I miss the days I could leave them mining 1 or 2 coins for months straight with no problems and still be profitable.

[tbearhere]

Guys, cheap psu and psu cable adapters are a big no no xD
They like to catch fire at odd moments...
How is everyone doing?
No more farm administration for me...Investors sold all the Nvidia and AMD cards :/

Yes, ivanlabrie exactly, that's why I buy the best psu's
And waiting till March for the new ti Nvidia 18 nm. Maybe a comeback with hash to electricity costs.  

[djm34]

Guys, cheap psu and psu cable adapters are a big no no xD
They like to catch fire at odd moments...
How is everyone doing?
No more farm administration for me...Investors sold all the Nvidia and AMD cards :/

Yes, ivanlabrie exactly, that's why I buy the best psu's
And waiting till March for the new ti Nvidia 18 nm. Maybe a comeback with hash to electricity costs.   

yep, would be better, the problem is that I changed already my psu once, I didn't wanted to upgrade the second time... I use a spare one (bronze+) to power the 780ti, but it isn't the best quality, the problem though came from the raiser spare cable on which, I am pretty sure, there is no quality control at all...

Are you sure that it will be in march ?

[tbearhere]

Guys, cheap psu and psu cable adapters are a big no no xD
They like to catch fire at odd moments...
How is everyone doing?
No more farm administration for me...Investors sold all the Nvidia and AMD cards :/

Yes, ivanlabrie exactly, that's why I buy the best psu's
And waiting till March for the new ti Nvidia 18 nm. Maybe a comeback with hash to electricity costs.  

yep, would be better, the problem is that I changed already my psu once, I didn't wanted to upgrade the second time... I use a spare one (bronze+) to power the 780ti, but it isn't the best quality, the problem though came from the raiser spare cable on which, I am pretty sure, there is no quality control at all...

Are you sure that it will be in march ?

No I thought you maybe said that in the thread way back. ie  There was a so called leak that nvidia is working on it. I hope.  

[djm34]

Guys, cheap psu and psu cable adapters are a big no no xD
They like to catch fire at odd moments...
How is everyone doing?
No more farm administration for me...Investors sold all the Nvidia and AMD cards :/

Yes, ivanlabrie exactly, that's why I buy the best psu's
And waiting till March for the new ti Nvidia 18 nm. Maybe a comeback with hash to electricity costs.   

yep, would be better, the problem is that I changed already my psu once, I didn't wanted to upgrade the second time... I use a spare one (bronze+) to power the 780ti, but it isn't the best quality, the problem though came from the raiser spare cable on which, I am pretty sure, there is no quality control at all...

Are you sure that it will be in march ?

No I thought you maybe said that in the thread way back. ie  There was a so called leak that nvidia is working on it. I hope. 

I read recently that they aren't going to 18nm but to 16nm actually... (don't remember where) and a 960 is planned for january

[mmmaybe]

What kind of performance/price will the 960s have compared to the 970s and 980s?

[fredeq]

Check out this new version for VTC http://cryptomining-blog.com/4049-updated-windows-binary-of-the-ccminer-1-5-1-git-fork-by-sp-for-maxwell/

[djm34]

Check out this new version for VTC http://cryptomining-blog.com/4049-updated-windows-binary-of-the-ccminer-1-5-1-git-fork-by-sp-for-maxwell/

It isn't a new version it is (again) the old test version which doesn't support compute 3.0  and he wonder why I was unhappy that he released the test code...
"include some djm34 code addition"  I like the irony: "some addition" here consists in the whole lyra2re code. 
(band of vultures   and next time cryptomining will say I haven't done anything in a long time like the last time... )

newest version here: Lyra2RE with support to compute 3.0 https://github.com/djm34/ccminer
include m7 (xcn) code optimized for 9xx serie: 980: 12.8MH/s 780: 13.5MH/s

[tbearhere]

Guys, cheap psu and psu cable adapters are a big no no xD
They like to catch fire at odd moments...
How is everyone doing?
No more farm administration for me...Investors sold all the Nvidia and AMD cards :/

Yes, ivanlabrie exactly, that's why I buy the best psu's
And waiting till March for the new ti Nvidia 18 nm. Maybe a comeback with hash to electricity costs.  

yep, would be better, the problem is that I changed already my psu once, I didn't wanted to upgrade the second time... I use a spare one (bronze+) to power the 780ti, but it isn't the best quality, the problem though came from the raiser spare cable on which, I am pretty sure, there is no quality control at all...

Are you sure that it will be in march ?

No I thought you maybe said that in the thread way back. ie  There was a so called leak that nvidia is working on it. I hope.  

I read recently that they aren't going to 18nm but to 16nm actually... (don't remember where) and a 960 is planned for january

Excellent    http://gigjets.com/11/nvidia-geforce-gtx-960-specs-price/

[sp_]

That improves for sure keccak and skein, but in x11 context, you won't see almost anything as these algo are already pretty fast.  The main problem with uint2 is that it does not work at all with compute 30 (the logic should work but it seems there is a problem with register allocation...)

The uint2 keccak made x11 20-25 KHASH faster. on a 750ti

that's what I call "almost anything"

I implemented it now, Bmw +32%, keccac + 40%, skein +30% x11 is 70-100 KHASH faster on the 750ti. Blake512 is running slower with uint2.

[fredeq]

Check out this new version for VTC http://cryptomining-blog.com/4049-updated-windows-binary-of-the-ccminer-1-5-1-git-fork-by-sp-for-maxwell/

It isn't a new version it is (again) the old test version which doesn't support compute 3.0  and he wonder why I was unhappy that he released the test code...
"include some djm34 code addition"  I like the irony: "some addition" here consists in the whole lyra2re code. 
(band of vultures   and next time cryptomining will say I haven't done anything in a long time like the last time... )

newest version here: Lyra2RE with support to compute 3.0 https://github.com/djm34/ccminer
include m7 (xcn) code optimized for 9xx serie: 980: 12.8MH/s 780: 13.5MH/s

Sorry for the confusion djm. It yielded better results than miner listed in VTC thread, so I assumed its indeed new

Thanks for the explanation.