Author

Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 969. (Read 3426989 times)

newbie
Activity: 37
Merit: 0
Hold on? Are NVidia cards becoming viable on price/performance vs ATI?  Shocked

Can one install Nvidia and ATI on the same Windows rig? Anyone experimented NVidia cards on PCI-e risers?

I ran GTX670 as a primary and R290X as secondary on Windows 7 for a month. Cgminer is a pain in the ass to tweak though, wrong settings would either stall the computer or reboot it spontaneously, I had days were it ran well for hours and days were it rebooted every 5min on same settings.
Getting your actual accepted hash was a pain also, you might be mining at 900Kh but W/U would only be 750-800

I swapped for the Gigabyte 780 Ghz edition. To be honest lost maybe like 100-150kh (accepted) (regular scrypt), but:
1) Runs much more stable
2) Cudaminer keeps boosting performance, specially for like Vertcoin, YAcoin
3) you can actually take advantage of the second nVidia card if you decide to play games
4) produces less heat

R290x doesn't work unless you plug a monitor to it or use a dummy plug.

Orange is rejected
https://i.imgur.com/4isfWxA.png
newbie
Activity: 2
Merit: 0
@cbuchner1
Thanks Smiley this work!
hero member
Activity: 756
Merit: 502
Hello Smiley
I have a little problem. My cudaminer crash when i run this.
screen
I suppose that this is problem with configuration, am i right ?
Can you help me ?

I think the autotune fails for you.

so let's do a manual tuning Wink

8800 GT.. 112 cores

112 / 8 = 14 multiprocessors

try -l L14x3   possibly in combination with -C 1 or -C 2

and don't expect miracles.
hero member
Activity: 756
Merit: 502
Hold on? Are NVidia cards becoming viable on price/performance vs ATI?  Shocked

Can one install Nvidia and ATI on the same Windows rig? Anyone experimented NVidia cards on PCI-e risers?

price/performance... not so much for scrypt mining. GTX 780 (Ti) cards are f...in' expensive

but very much for scrypt-jane coins with high N-factor. Here a mid range card (660Ti) can beat the AMD high end offerings.

I will experiment with PCI-e risers soon. I've ordered a 1x->1x riser, a 4x->4x riser, and a 1x->16x riser (unpowered) to play with. With a bit of Dremel magic, the card will fit into the 1x and 4x slots on the risers Wink

Christian
sr. member
Activity: 840
Merit: 255
SportsIcon - Connect With Your Sports Heroes
Hold on? Are NVidia cards becoming viable on price/performance vs ATI?  Shocked

Can one install Nvidia and ATI on the same Windows rig? Anyone experimented NVidia cards on PCI-e risers?
newbie
Activity: 2
Merit: 0
Hello Smiley
I have a little problem. My cudaminer crash when i run this.
screenhttp://img.pl/GMwf.png
I suppose that this is problem with configuration, am i right ?
Can you help me ?
sr. member
Activity: 350
Merit: 250
Ahh ok. I will stick with T till you say how well MRC is doing on returns :-)
hero member
Activity: 756
Merit: 502
Would you recommend trying the Z kernel on my 780 to try it christian? For yacoin at the moment that is

The Z kernel is good for low N factor only. For yacoin better stick with T.

hero member
Activity: 756
Merit: 502
Using the Z kernel I just broke 900 khash/s with a water cooled 780 TI.

Nice!! I put your posting up on reddit...

http://redd.it/1wmtjh  and http://redd.it/1wmxoi

upvote for visibility.
sr. member
Activity: 350
Merit: 250
So everyone is finding an improvement using the Z kernel?
That is on newer cards (780, 780 ti)? If so I may give it a go. How do I auto tune in a given kernel?
It is just -l Z?
newbie
Activity: 2
Merit: 0
Hi all

I'm new to mining but had a go setting up a linux box with CUDAminer today. I tested with a GTX470 but plan to upgrade to a GTX780Ti (or two Smiley ) if it all goes well. Getting a fresh linux box up and going was a little tricky so I jotted down instructions. Christian may want to includes these as a doc on github or add them to the first post.

Using this command line I get about 165kh/s:
 
Code:
./cudaminer --benchmark -H0 -d 0 -i 0 -C 1 -l F28x10

Enjoy
Helium3
--

Code:
Install Xubuntu 13.10 (it's the same as Ubuntu but uses Xfce window manager) - http://xubuntu.org/

[Install base packages - open a terminal window from X]
sudo apt-get install openssh-server
sudo apt-get install vim
sudo apt-get install screen

[Download latest NVIDIA driver from nvidia.com - get the .run file]

[Optional: ssh into box - failing that, press CTRL-ALT-F1 to switch to a terminal window from X]

sudo service lightdm stop
sudo ./NVIDIA-Linux-x86_64-331.38.run  [[change for current driver version]]

sudo vim /etc/modprobe.d/nvidia-graphics-drivers.conf

[Populate the following file with following information: ]
# Ensure the manual driver and kernel module is used
blacklist nouveau
blacklist lbm-nouveau
blacklist nvidia-173
blacklist nvidia-96
blacklist nvidia-current
blacklist nvidia-173-updates
blacklist nvidia-96-updates
alias nvidia nvidia_current_updates
alias nouveau off
alias lbm-nouveau off


[Recreate the initrd file: ]
sudo update-initramfs -u
[Reboot ]
sudo shutdown -r now

[Check you're running the right driver. Once booted into X, from a terminal run:
nvidia-settings
(or from a pure command shell "nvidia-settings --version")

[Get the CUDA toolkit from the developer.nvidia.com website]
mkdir cuda-extract
./cuda_5.5.22_linux_64.run --extract=~/cuda-extract/
cd cuda-extract
sudo sh ./cuda-linux64-rel-5.5.22-16488124.run

sudo vim /etc/ld.so.conf.d/cuda_settings.conf
[Add these two lines: ]
/usr/local/cuda-5.5/lib64
/usr/local/cuda-5.5/lib

[Rebuild the library data config]
sudo ldconfig

[Add the CUDA bin directory to your path]
echo PATH="$PATH:/usr/local/cuda-5.5/bin" >> ~/.profile


[Install all the various bits you'll need to build CUDAminer]
sudo apt-get install automake
sudo apt-get install curl
sudo apt-get install libcurl3
sudo apt-get install libcurl3:i386
sudo apt-get install libcurl3-gnutls-dev
sudo apt-get install gcc
sudo apt-get install g++
[NOTE: If you install Xubuntu 32bit, you can skip the libcurl3:i386. ]
[NOTE: If you install Xubuntu 64bit, you might also need libc6:i386 & libc++6:i386 but I don't think so]


[Grab Cudaminer from github via git or get the zip and unzip it. Then cd into the root Cudaminer dir]
./autogen.sh
./configure
make

[All of the above should hopefully work]
./cudaminer --help
./cudaminer --benchmark


[*** Optional, disable X from running at start-up ***]
sudo vim /etc/default/grub
[look for this line: ]
GRUB_CMDLINE_LINUX_DEFAULT="quiet splash"
[change it to: ]
GRUB_CMDLINE_LINUX_DEFAULT="quiet splash text"
[save and exit, then update grub ]
sudo update-grub
sudo shutdown -r now
newbie
Activity: 16
Merit: 0
Could someone remind me what the -m 0 and -m 1, and -H 0, 1, and 2 switches do? I'm finding it difficult to search for single letters.
From the readme:

"--single-memory  [-m] list of flags (0 or 1) to make the devices
                 allocate their scrypt scratchpad in a single,
                 consecutive memory block. On Windows Vista, 7/8
                 this may lead to a smaller memory size being used.
                 When using the texture cache this option is implied."

"--hash-parallel  [-H] scrypt also has a small SHA256 component to it:
                      0 hashes this single threaded on the CPU.
                      1 to enable multithreaded hashing on the CPU.
                      2 offloads everything to the GPU (default)"

I have the most efficient results with m1 and h2 with my 780 ti in windows.
newbie
Activity: 16
Merit: 0
Using the Z kernel I just broke 900 khash/s with a water cooled 780 TI. This kernel ended up being 11% faster than the 12-18 releases T kernel. I had to use -H 1 for the extra 8 hashes over -H 2, at the cost of 25w. A terrible trade, but it put me over 900. Cheesy

N-factor 9 Microcoin is running at @ 834 khash/s. That's a much smaller loss than a R9 290 takes. They are now yielding 666 vs 880 on scrypt. The memory overclock actually helps on Microcoin, but it isn't an efficient trade. 8~ hashes for 8~ watts. Also, solo mining MRC was terrible for me in the few hours I tried it. I had several times better results in the blocksolved pool.

https://i.imgur.com/BEONlhE.jpg
full member
Activity: 812
Merit: 102
Could someone remind me what the -m 0 and -m 1, and -H 0, 1, and 2 switches do? I'm finding it difficult to search for single letters.
newbie
Activity: 7
Merit: 0
Again the same - lame question - can you upload please binary win x64 please? ;-)

I believe cbuchner1 is targeting the end of the month for another official release.
legendary
Activity: 1400
Merit: 1050
On the 780ti, it clearly helps. Just made a rapid check on MRC:
On default setting: 500khash/s, overclocked (+130% TDP) 600khash/h.
(however on stock bios you might get only half of that, I modded my bios to increase the power limit.
but I didn't touch the voltage...)

The overclock is a bit tricky though because of the boost which may or may not kick-in...
sr. member
Activity: 350
Merit: 250
Most likely yes. As we dont know the pool I cant say for sure
jr. member
Activity: 63
Merit: 4
ok thanks.hay in the pool it show i have 14 credits does that =14 coins. thanks
sr. member
Activity: 350
Merit: 250
Some people have seen that it does.
Personally I see no change atall, but as I have a 780 is clocks itself anyway. Mine tends to just hit voltage limits
jr. member
Activity: 63
Merit: 4
thanks does overclocking you card help
Jump to: