search

JeanLucPons / VanitySearch

Bitcoin Address Prefix Finder
GNU General Public License v3.0
422 stars 203 forks source link

How to compile RTX 3090 with vanity search on ubuntu(VM)? #97

Closed siupune closed 3 years ago

siupune commented 3 years ago

please guide me if anybody has done it.....

Jolly-Pirate commented 3 years ago

First, install the CUDA toolkit https://developer.nvidia.com/cuda-downloads , the download and install instructions are given there, I recommend you do the "deb (local)" intallation, it's easier and less problematic than the "runfile (local) "method. Then do this too sudo apt -y install nvidia-cuda-toolkit, and reboot (you have to do that). Check the nvcc version with nvcc --version and run nvidia-smi to check CUDA and the drivers are all good. Clone the VanitySearch repo and build with:

git clone https://github.com/JeanLucPons/VanitySearch
cd VanitySearch
make gpu=1 CCAP=8.0 CUDA=/usr/local/cuda-11.2 CXXCUDA=/usr/bin/g++ all

or make gpu=1 CCAP=8.6 CUDA=/usr/local/cuda-11.2 CXXCUDA=/usr/bin/g++ all I'm not sure which CCAP version is the one for RTX 3090, and of course use the appropriate CUDA version you installed, e.g. CUDA=/usr/local/cuda-11.2, then you should have a VanitySearch binary that you can run with ./VanitySearch. You can copy it to your $HOME/bin folder so it will be accessible from any other folder, and you can run it directly by typing VanitySearch (no need for the ./

There's also another useful tool you can compile:

cd /usr/local/cuda-11.2/samples/1_Utilities/deviceQuery
sudo make
sudo cp deviceQuery /usr/bin/

And type deviceQuery for a detailed info about your GPU.

siupune commented 3 years ago

prabably some bug as i delete few letters from address, it works fine....

siupune commented 3 years ago

Thank you sooooooooo much for the help......

On Fri, Mar 19, 2021 at 12:21 PM drakos @.***> wrote:

First, install the CUDA toolkit https://developer.nvidia.com/cuda-downloads , the download and install instructions are given there, I recommend you do the "deb (local)" intallation, it's easier and less problematic than the "runfile (local) "method. Then do this too sudo apt -y install nvidia-cuda-toolkit, and reboot (you have to do that). Check the nvcc version with nvcc --version and run nvidia-smi to check CUDA and the drivers are all good. Clone the VanitySearch repo and build with:

git clone https://github.com/JeanLucPons/VanitySearch cd VanitySearch make gpu=1 CCAP=8.0 CUDA=/usr/local/cuda-11.2 CXXCUDA=/usr/bin/g++ all

or make gpu=1 CCAP=8.6 CUDA=/usr/local/cuda-11.2 CXXCUDA=/usr/bin/g++ all I'm not sure which CCAP version is the one for RTX 3090, and of course use the appropriate CUDA version you installed, e.g. CUDA=/usr/local/cuda-11.2, then you should have a VanitySearch binary that you can run with ./VanitySearch. You can copy it to your $HOME/bin folder so it will be accessible from any other folder, and you can run it directly by typing VanitySearch (no need for the ./

There's also another useful tool you can compile:

cd /usr/local/cuda-11.2/samples/1_Utilities/deviceQuery sudo make sudo cp deviceQuery /usr/bin/

And type deviceQuery for a detailed info about your GPU.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/JeanLucPons/VanitySearch/issues/97#issuecomment-802599537, or unsubscribe https://github.com/notifications/unsubscribe-auth/ARMHY2YWKQYYVM74WY4NZLTTELX57ANCNFSM4ZOCPI6Q .

Jolly-Pirate commented 3 years ago

Thank you sooooooooo much for the help......

Actually, I should have mentioned to install deviceQuery first, since it gives you the exact CCAP version e.g. CUDA Capability Major/Minor version number: 7.5. From what I've read on forums and comments, it's 8.6 for the RTX 3090.

siupune commented 3 years ago

i checked kangaroo with ubuntu 20.04 with rtx 3090 and showed these errors.... kangaroo error

siupune commented 3 years ago

First, install the CUDA toolkit https://developer.nvidia.com/cuda-downloads , the download and install instructions are given there, I recommend you do the "deb (local)" intallation, it's easier and less problematic than the "runfile (local) "method. Then do this too sudo apt -y install nvidia-cuda-toolkit, and reboot (you have to do that). Check the nvcc version with nvcc --version and run nvidia-smi to check CUDA and the drivers are all good. Clone the VanitySearch repo and build with:

git clone https://github.com/JeanLucPons/VanitySearch
cd VanitySearch
make gpu=1 CCAP=8.0 CUDA=/usr/local/cuda-11.2 CXXCUDA=/usr/bin/g++ all

or make gpu=1 CCAP=8.6 CUDA=/usr/local/cuda-11.2 CXXCUDA=/usr/bin/g++ all I'm not sure which CCAP version is the one for RTX 3090, and of course use the appropriate CUDA version you installed, e.g. CUDA=/usr/local/cuda-11.2, then you should have a VanitySearch binary that you can run with ./VanitySearch. You can copy it to your $HOME/bin folder so it will be accessible from any other folder, and you can run it directly by typing VanitySearch (no need for the ./

There's also another useful tool you can compile:

cd /usr/local/cuda-11.2/samples/1_Utilities/deviceQuery
sudo make
sudo cp deviceQuery /usr/bin/

And type deviceQuery for a detailed info about your GPU.

Jolly-Pirate commented 3 years ago

i checked kangaroo with ubuntu 20.04 with rtx 3090 and showed these errors.... kangaroo error

Kangaroo uses the old ccap make switch, so change CCAP=8.6 to ccap=86. You can determine that by looking at the Makefile, and it's also in the README of Kangaroo.

ZielarPL commented 3 years ago

You must edit file named Makefile Go to the lines 62 and 65 and change "$(ccap)" to "86"

Jolly-Pirate commented 3 years ago

You must edit file named Makefile Go to the lines 62 and 65 and change "$(ccap)" to "86"

No need to edit the Makefile, you can pass it as a parameter when you run the make command. Besides, it can be tedious to edit everytime if you have multiple and different GPUs.

ZielarPL commented 3 years ago

When i try compile few weeks ago - set CCAP=86 give that same error what when use 8.6 so i write checked solution to solve that problem

siupune commented 3 years ago

its not setting full cuda cores( 82x 0) image

siupune commented 3 years ago

Any suggestions please?????????

Jolly-Pirate commented 3 years ago

I've seen that GeForce RTX 3090 (82x0 cores) problem before https://bitcointalk.org/index.php?topic=5112311.msg55773362#msg55773362 Not sure if it was resolved. Can you try downgrading to CUDA 11.1 and recompile it? You can get it here https://developer.nvidia.com/cuda-toolkit-archive

Also, in later responses on the forum https://bitcointalk.org/index.php?topic=5112311.msg56537112#msg56537112, it seems many are having problems with RTX 30xx. Clearly the code of VanitySearch and Kangaroo need proper updating/optimization for those cards.

siupune commented 3 years ago

Somebody already changed codes for Kangaoo here : https://gitea.datahoarding.agency/ZenulAbidin/Kangaroo-256 just change last line in GPUEngine.cu 86 , 128 not 64. thats all but i came across Segmentation broken(core dumped)

siupune commented 3 years ago

i get these error after using these updated kangaroo-256 : https://gitea.datahoarding.agency/ZenulAbidin/Kangaroo-256

Kangaroo-256 core dumped witout -g

ZenulAbidin commented 3 years ago

Somebody already changed codes for Kangaoo here : https://gitea.datahoarding.agency/ZenulAbidin/Kangaroo-256 just change last line in GPUEngine.cu 86 , 128 not 64. thats all but i came across Segmentation broken(core dumped)

Fork author here.

If modification to 128 SMs doesn't work then just leave it at the default 64, some A100 cards have less than 128 SMs: https://www.guru3d.com/news-story/nvidia-tesla-a100-with-ga100-ampere-gpu-spotted.html

Now for further debugging I need you to upload a core dump for me.

Run this command as root:

echo "/tmp/cores/core.%e.%p.%h.%t" > /proc/sys/kernel/core_pattern

This will temporarily enable core dumps until you reboot.

Then compile Kangaroo again in debug mode (make gpu=1 debug=1) and run the program again and then look for a file called "/tmp/cores/core.something.something.something.something". Upload it so that I can inspect it locally.

Radar-7528 commented 3 years ago

я получаю эту ошибку после использования этих обновленных kangaroo-256: https://gitea.datahoarding.agency/ZenulAbidin/Kangaroo-256

Ядро Kangaroo-256 выгружено без -g

я получаю эту ошибку после использования этих обновленных kangaroo-256: https://gitea.datahoarding.agency/ZenulAbidin/Kangaroo-256

Ядро Kangaroo-256 выгружено без -g

8 GPU, cool. Need to brute-force passwords wallet.dat....no need to waste time on collisions. Write to me by mail.

siupune commented 3 years ago

please send me your telegram id and will message you

On Thu, Apr 1, 2021 at 5:42 AM Radar-7528 @.***> wrote:

я получаю эту ошибку после использования этих обновленных kangaroo-256: https://gitea.datahoarding.agency/ZenulAbidin/Kangaroo-256

[image: Ядро Kangaroo-256 выгружено без -g] https://user-images.githubusercontent.com/72907883/112721333-4a6caf80-8f29-11eb-8903-da7ad85c73d2.JPG

я получаю эту ошибку после использования этих обновленных kangaroo-256: https://gitea.datahoarding.agency/ZenulAbidin/Kangaroo-256

[image: Ядро Kangaroo-256 выгружено без -g] https://user-images.githubusercontent.com/72907883/112721333-4a6caf80-8f29-11eb-8903-da7ad85c73d2.JPG

8 GPU, cool. Need to brute-force passwords wallet.dat....no need to waste time on collisions. Write to me by mail.

— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/JeanLucPons/VanitySearch/issues/97#issuecomment-811546248, or unsubscribe https://github.com/notifications/unsubscribe-auth/ARMHY25GKZNWWD4OKYP3WG3TGO25TANCNFSM4ZOCPI6Q .

Radar-7528 commented 3 years ago

please send me your telegram id and will message you ok. https://t.me/G_Rodary