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

[leofur]

Well if it depends on the algo, I guess the new version won't fix it... Now the question is if it worth getting a strong CPU now for my rig...

[zelante]

Well if it depends on the algo, I guess the new version won't fix it... Now the question is if it worth getting a strong CPU now for my rig...

intel i7 3770k@4500MHz 1 instance of ccminer with x11 use 6,5% of cpu. Using 19,5% now for 660ti + 2x 750ti (660ti take same percents of usage cpu as 750ti)

[leofur]

Well if it depends on the algo, I guess the new version won't fix it... Now the question is if it worth getting a strong CPU now for my rig...

intel i7 3770k@4500MHz 1 instance of ccminer with x11 use 6,5% of cpu. Using 19,5% now for 660ti + 2x 750ti (660ti take same percents of usage cpu as 750ti)

On an i3 750ti uses 9% of CPU, on celeron 19%.
Maybe Christian can give a better answer what causes the high CPU usage.

[gpuminer]

Well if it depends on the algo, I guess the new version won't fix it... Now the question is if it worth getting a strong CPU now for my rig...

intel i7 3770k@4500MHz 1 instance of ccminer with x11 use 6,5% of cpu. Using 19,5% now for 660ti + 2x 750ti (660ti take same percents of usage cpu as 750ti)

On an i3 750ti uses 9% of CPU, on celeron 19%.
Maybe Christian can give a better answer what causes the high CPU usage.

maybe the CPU is adding to the hashrate. For certain algo among the 11 algo to make it more efficient, just guessing , I am no programmer.

[cbuchner1]

X11 perf on one 750 Ti (development version)...  Be afraid AMD.

Code:

[2014-05-12 00:43:33] thread 0: 3670017 hashes, 2698 khash/s
[2014-05-12 00:43:33] Total: 2698 khash/s
[2014-05-12 00:43:42] thread 0: 13107201 hashes, 2646 khash/s
[2014-05-12 00:43:42] Total: 2646 khash/s

Brutal.
I'm shocked.

I think I've just 1up'ed this. This is one 780Ti EVGA FTW on X11 with "Killer Groestl" and improved SIMD hash.

Code:

[2014-05-14 10:39:36] thread 0: 4194305 hashes, 2828 khash/s
[2014-05-14 10:39:36] Total: 2828 khash/s
[2014-05-14 10:39:41] thread 0: 14680065 hashes, 2821 khash/s
[2014-05-14 10:39:41] Total: 2821 khash/s

A GTX 780 (non Ti) registered at 5 MHash/s.


Because the new SIMD hash is to be considered unportable to AMD (just look  at all the nVidia specific warp shuffles!) I have committed
this to github master. Note that our "killer groestl" is still absent from this version, so you won't quite see above cited performance
figures - but still a huge improvement over the v1.0 release.

Enjoy the X11 speed upgrade - but be aware that Compute 2.0 devices definitely won't like this code.

I measure slightly above 2 MHash for a 750Ti now with the code in github (MAXWELL_OR_FERMI defined as #1)
and 4 MHash on the 780 (non Ti) with MAXWELL_OR_FERMI defined as #0

Christian

[sobad]

So, nonTi is doing better?

[leofur]

With the problem with CPU usage, I guess I'm back to talk coin for now. Anyone else still mining it?

[izum286]

A GTX 780 (non Ti) registered at 5 MHash/s.

HOW? I have 780ti - 3 MHash only...

[cayars]

Awesome Chris(s)!

[cbuchner1]

A GTX 780 (non Ti) registered at 5 MHash/s.

HOW? I have 780ti - 3 MHash only...

duh, check out github source code and compile. This one does 4 MHash then.

[cayars]

cbuchner1, given what you know.

If you had X money to spend on buying GPUs.  What would you buy/purchase or would you wait for the next wave of maxwell?

X could be 1K/us,  5K or 10K just to make it interesting.

Love to get your feedback on this.

Carlo

[cbuchner1]

Well if it depends on the algo, I guess the new version won't fix it... Now the question is if it worth getting a strong CPU now for my rig...

intel i7 3770k@4500MHz 1 instance of ccminer with x11 use 6,5% of cpu. Using 19,5% now for 660ti + 2x 750ti (660ti take same percents of usage cpu as 750ti)

On an i3 750ti uses 9% of CPU, on celeron 19%.
Maybe Christian can give a better answer what causes the high CPU usage.

the "throughput" value was set pretty low (i.e. the number of threads thrown at the GPUs at once). The current github code has it increased 8-fold.
This should lower CPU use.

[leofur]

So anyone can build it for us all...

[zelante]

Well if it depends on the algo, I guess the new version won't fix it... Now the question is if it worth getting a strong CPU now for my rig...

intel i7 3770k@4500MHz 1 instance of ccminer with x11 use 6,5% of cpu. Using 19,5% now for 660ti + 2x 750ti (660ti take same percents of usage cpu as 750ti)

On an i3 750ti uses 9% of CPU, on celeron 19%.
Maybe Christian can give a better answer what causes the high CPU usage.

the "throughput" value was set pretty low (i.e. the number of threads thrown at the GPUs at once). The current github code has it increased 8-fold.
This should lower CPU use.

Great work!
750ti now uses ~1,5% (previously 6,5%) and got 2150 khash/s (previously 1700 khash/s)!!
660ti got 2300 khash/s (previously 1900 khash/s)

[bigjme]

You mean khash/s?

[zelante]

You mean khash/s?

yup. wishful thinking 

[sp00lin9]

Well if it depends on the algo, I guess the new version won't fix it... Now the question is if it worth getting a strong CPU now for my rig...

intel i7 3770k@4500MHz 1 instance of ccminer with x11 use 6,5% of cpu. Using 19,5% now for 660ti + 2x 750ti (660ti take same percents of usage cpu as 750ti)

On an i3 750ti uses 9% of CPU, on celeron 19%.
Maybe Christian can give a better answer what causes the high CPU usage.

the "throughput" value was set pretty low (i.e. the number of threads thrown at the GPUs at once). The current github code has it increased 8-fold.
This should lower CPU use.

Great work!
750ti now uses ~1,5% (previously 6,5%) and got 2150 mhash/s (previously 1700 mhash/s)!!
660ti got 2300 mhash/s (previously 1900 mhash/s)

Outstanding work Christian! Can't wait to try all the fancy new x11 coins with nvidia power! Zelante, I was wondering if you would please share the compiled windows version, it's quite late over here and I don't want to deal with VS tonight. Thanks 

[bigjme]

I think i best move over to x11 soon down to 0.007btc a day on jackpot

[valuxin]

It seems strange, but cgminer do better work on my Quadro 5010m (Fermi arch) then cudaminer. For example, max speed on blake256 algo with cudaminer - 78 Mh/s, cgminer - 574 Mh/s. Cudaminer use 3700Mb of video memory, cgminer - 200Mb.

[cbuchner1]

If you had X money to spend on buying GPUs.  What would you buy/purchase or would you wait for the next wave of maxwell?

Maybe I wouldn't actually invest in GPUs, considering the uncertainity about making ROI

Christian