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

[pluMmet]

Which of  the Cryptos will the mine?

[GG9000]

Testing this now, looks good so far!

Regards

[CryptoLego]

I wanted to give my stats:

GTX 580 using 331.82 drivers

Using Cudaminer 11/20/2013

Win 7 X64

options: -H 1 -l F16x14

Auto-tuning was using F32x8 which was averaging lower than what I am getting using F16x14.

avg about 230 khash/s

I was using the 64 bit version but I was only averaging about 200 khash/s

Using about 95% of GPU running at a temp of about 83 degrees Celsius.

I believe this is the best I can expect running a stock GTX 580 with no overclocking.

I think you should get more than that as i get 230Kh/s from a 560ti 448 core. Maybe try F16x16? oh and -i 0.

Hey fruittool:

Thanks for the tip. GPU usage has increased to 99%.

I am now hitting a max of 240 khash/s.

GPU Core running at 85 degrees Celsius.

Update:

I started using -C 1 option.

My lowly vanilla GTX 580 now maxes at 259 khash/s! 

[inevitable7]

As a total newb, I damn nearly pulled my hair out with this thing but I eventually got it to work with my 2x780 EVGA FTW's. I poured over this entire thread TWICE to find a config that worked and NOTHING WORKED until I came across someone who posted this extension:

 -H 1 -l K48x6 -C 2

That worked great, but only one card functioned and with weird lag issues. That all went away after disabling SLI and adding:

 -d1,0

It did an autotune for the second card but then both fired into life! YAAAAAY I didn't just waste FOUR HOURS!!!! I'm so happy! Currently hashing away at a combined speed of 620kh/s. If anyone has any other suggestions to tweak these bad boys, i'd be super grateful to hear it. Anyway, just thought I would share my findings.

EDIT: Dropped to 560kh/s now. The next mission: MORE POWWWWEEERRRRR!

my advice?  sell those cards on ebay and get 3 r9 280x for 2100 khash/s

[xtrem3x]

I have a 780 and use:
cudaminer.exe -i 0 -l K42x6 -C 2 -H 1 -o

Use it on x86 for a ~5% boost over x64

My 780 is water cooled so max temp after 20 hours was 45c

Non overclocked - 330Kh/s
Mild overclock - 360Kh/s
Max overclock - not tried yet but 400 Kh/s should be obtainable without issue.

This is my gaming rig, mining is just something it'll do at night time and when I'm out.
only sees a max of 10% CPU so everything is responsive still so may keep it mining all the time unless I'm playing a game.

I know an AMD card would be better for mining, but gaming is still priority so I'll stick with Nvidia.

[cbuchner1]

Non overclocked - 330Kh/s
Mild overclock - 360Kh/s
Max overclock - not tried yet but 400 Kh/s should be obtainable without issue.

I am working on some secret sauce to make Kepler generation devices even faster.

[CryptoLego]

Non overclocked - 330Kh/s
Mild overclock - 360Kh/s
Max overclock - not tried yet but 400 Kh/s should be obtainable without issue.

I am working on some secret sauce to make Kepler generation devices even faster.

Hi Christian:

I apologize in advance if this question has been asked before.

Do you recommend running cpuminer concurrently with cudaminer?

Thanks in advance for your input and all  your hard work.

[xtrem3x]

Non overclocked - 330Kh/s
Mild overclock - 360Kh/s
Max overclock - not tried yet but 400 Kh/s should be obtainable without issue.

I am working on some secret sauce to make Kepler generation devices even faster.

This would be awesome!!!

[DuckDodgers]

Here's what my old rusty GTX580 can do:

Code:

           *** CudaMiner for nVidia GPUs by Christian Buchner ***
                     This is version 2013-11-20 (alpha)
        based on pooler-cpuminer 2.3.2 (c) 2010 Jeff Garzik, 2012 pooler
               Cuda additions Copyright 2013 Christian Buchner
           My donation address: LKS1WDKGED647msBQfLBHV3Ls8sveGncnm

[2013-11-30 18:32:05] Starting Stratum on stratum+tcp://coinotron.com:3334
[2013-11-30 18:32:05] 1 miner threads started, using 'scrypt' algorithm.
[2013-11-30 18:32:05] Stratum detected new block
[2013-11-30 18:32:06] GPU #0: GeForce GTX 580 with compute capability 2.0
[2013-11-30 18:32:06] GPU #0: interactive: 1, tex-cache: 2D, single-alloc: 1
[2013-11-30 18:32:06] GPU #0: using launch configuration F16x16
[2013-11-30 18:32:06] GPU #0: GeForce GTX 580, 8192 hashes, 35.84 khash/s
[2013-11-30 18:32:14] GPU #0: GeForce GTX 580, 2154496 hashes, 275.91 khash/s
[2013-11-30 18:32:20] GPU #0: GeForce GTX 580, 1597440 hashes, 272.02 khash/s
[2013-11-30 18:32:20] accepted: 1/1 (100.00%), 272.02 khash/s (yay!!!)
[2013-11-30 18:33:20] GPU #0: GeForce GTX 580, 16326656 hashes, 274.50 khash/s
[2013-11-30 18:33:50] Stratum detected new block
[2013-11-30 18:33:50] GPU #0: GeForce GTX 580, 8495104 hashes, 275.91 khash/s

GPU overclock: 825/4200MHz
OS: Windows 8 x64

[cbuchner1]

Do you recommend running cpuminer concurrently with cudaminer?

I do this on two machines. Works for me

[bugilt]

I've started to have a weird issue over the last few days. My ethernet dies when I run cudaminer on my 780. I reinstalled the driver for the ethernet, but it still happens. When I get more time I'm going to try to troubleshoot it down. A new GPU driver was the latest change to my system. That will be my next step.

[boldwin]

How can I tweak to maximize my khash?
My setup: Nvidia GTX 570 (OC'ed)
Core CLock: 850 Mhz
Memory Clock 2000 Mhz
Core Voltage: 1063 mV

Win 7, 64 bit.
Using Cuduminer 64 bit I get 215 khash/s with this command line: C:\Users\AA\Desktop\cudaminer-2013-11-20\x64\cudaminer.exe -o stratum+tcp://global.wemineltc.com:3335 -O Workername:pass

Auto-tuning yields: F15x14 (sometimes F28x7) , interactive:1, tex-cash: 0, single-alloc: 0.
One problem is my GPU changes from between 70 - 97% usage. With temperatures at 65 *C after 30mins.

How can I make my GPU stay at 97% usage to maximize khash?

How else can I tweak to maximize my khash?

[cbuchner1]

-i 0 maximizes GPU use. It may get laggy if you also run a display on the same GPU.

try -C 1 or -C 2 for a few extra kHash/s

otherwise, consider overclocking.

[cbuchner1]

thanks for the 2 LTC donation that just popped in. That is quite an amount these days (buys me a lot of Pampers for my son, lol)

[ilostcoins]

Would it be hard to adapt this CUDA miner to "scrypt-jane" coins, like YACoin? It "uses the scrypt + chacha20/8 (N,1,1) hashing algorithm, with N gradually rising over time to increase memory requirements."

The performance differential between CPU and AMD GPU right now for YACoin is quite a bit smaller than for normal scrypt coins. May be the performance gap between Nvidia and AMD GPU will also be smaller than scrypt coins. In other words, Nvidia cards may be better off with mining "scrypt-jane" coins.

Thanks for considering the question.

[boldwin]

-i 0 maximizes GPU use. It may get laggy if you also run a display on the same GPU.

try -C 1 or -C 2 for a few extra kHash/s

otherwise, consider overclocking.

Yeah Im OCed right now.

I do run my display on the same GPU. Is there a way to set that -i 0     command to a percentage of GPU usage? like 98%? instead of 100?

[cbuchner1]

I do run my display on the same GPU. Is there a way to set that -i 0     command to a percentage of GPU usage? like 98%? instead of 100?

Nope, no fine granular control over GPU utilization is available in the current release.

What -i 1 actually does is as follows:

-It lets autotune only choose launch configurations that run for less than 1/30th of a second.

-because the scrypt kernel consists of two parts (A and B) and interactive mode performs
 a sleep of 1 millisecond between each invocation, the GPU is able to perform draw calls
 at least twice per 1/30th second -> so you should reach a screen update rate of 60 FPS.

So you can watch movies, TV in interactive mode without noticeable stutter. But the 60
FPS is only guaranteed if you use a launch config that autotune actually found in interactive
mode. This especially matters on slow GPUs where big kernel launches take considerable time.

My 560Ti 448 core edition (not overclocked) beats your card. 228 kHash/s here.
Must be the interactive mode?

I run it like this: -l F28x8 -C 1 -i 0 -m 1 -H 1

Also: try the cudaminer 32 bit binary for comparison. With some kernels it does run quite a bit
faster.

[cbuchner1]

Would it be hard to adapt this CUDA miner to "scrypt-jane" coins, like YACoin? It "uses the scrypt + chacha20/8 (N,1,1) hashing algorithm, with N gradually rising over time to increase memory requirements."

So far I haven't considered any other hashing algorithms. It's just too much work to adapt the code.

But the thing is open source. Hack away!

[cbuchner1]

CryptoLego: I don't get it ... you have basically the same set up as me yet I get the dreaded "Unable to query CUDA driver" ... what have you installed that I have not?? 

Any suggestions would be great!

does the free utility cuda-z show any available cuda devices?

[cbuchner1]

I am working on some secret sauce to make Kepler generation devices even faster.

The secret sauce is currently in github in the test_kernel.cu (which now requires Compute 3.5 to build). Be sure to run autogen.sh to update the Makefile before compiling.

However the secret sauce seems to be an undesiredlly "hot sauce".  The card runs a bit hotter and slower than the old Titan kernel. You get the experimental kernel by using a launch config prefixed with -X

I get 417 kHash with the experimental kernel and up to 450 kHash with the old Titan kernel (cudaminer 32 bit version). The experimental stuff is interesting from a technical standpoint. It gets rid of shared memory entirely, replacing it with Kepler's warp shuffle functionality.