GPU #0: bitcredit_setBlockTarget invalid device symbol about pow-miner-gpu HOT 35 OPEN

This is not the problem. Or just use the opencl miner which will work fine

from pow-miner-gpu.

The opencl version of tonminer is less then 7% lower in terms of hashrate compare to cuda. So threre is a way to avoid this bug just unsing opencl miner
Разница производительности майнера тон между opencl и cuda для описанных здесь видеокарт - около 7%, не следует считать, что это баг, выход есть. Он оправдан.

from pow-miner-gpu.

The opencl version of tonminer is less then 7% lower in terms of hashrate compare to cuda. So threre is a way to avoid this bug just unsing opencl miner Разница производительности майнера тон между opencl и cuda для описанных здесь видеокарт - около 7%, не следует считать, что это баг, выход есть. Он оправдан.

@DelphiR
Хватит вешать лапшу на уши людям. Разница в 4.4 раза, а не 7%, при этом gpu не загружена больше 1%. Баг есть. И люди не могут полноценно использовать gpu.

@tontechio When will the fix come out?
OpenCL speed: 17 Mhash\s
Gpu with 0,2% load: 74 Mhash\s

from pow-miner-gpu.

The opencl version of tonminer is less then 7% lower in terms of hashrate compare to cuda. So threre is a way to avoid this bug just unsing opencl miner Разница производительности майнера тон между opencl и cuda для описанных здесь видеокарт - около 7%, не следует считать, что это баг, выход есть. Он оправдан.

@DelphiR Хватит вешать лапшу на уши людям. Разница в 4.4 раза, а не 7%, при этом gpu не загружена больше 1%. Баг есть. И люди не могут полноценно использовать gpu.

@tontechio When will the fix come out? OpenCL speed: 17 Mhash\s Gpu with 0,2% load: 74 Mhash\s

I fully support it!
Developers, solve the problem!

from pow-miner-gpu.

Having the same issue on the latest release (https://github.com/tontechio/pow-miner-gpu/releases/tag/20211119-4).
[ 1][t 4][2021-11-19 20:23:04.7726548][cuda.cpp:209] [ GPU #0: bitcredit_setBlockTarget invalid device symbol ]

But have CUDA 11.5 installed:

`PS C:\miner> nvidia-smi
Fri Nov 19 23:28:58 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 496.76       Driver Version: 496.76       CUDA Version: 11.5     |
|-------------------------------+----------------------+----------------------+

Device: GTX850M.
Win 10x64.
Driver: 496.76.
CUDA 11.5.

Waiting for the fix to be addressed.

from pow-miner-gpu.

This is an issue related to Win 11 + mobile-grade GPU. Working on it

from pow-miner-gpu.

Helo!
I have the same error on Windows 10x64 Version 20H2.
NVIDIA GeForce GTX 850M

See line 70.

...
4: [ 1][t 1][2021-11-10 15:06:12.4775488][cuda.cpp:81][!console]	[ GPU #0: SM 5.0 NVIDIA GeForce GTX 850M ]
5: [ 3][t 1][2021-11-10 15:06:12.4799021][tonlib-cli.cpp:883][!console]	Miner #1 created
...
67: synchroni[ 3][t 4][2021-11-10 15:13:45.7972554][credits.cu:29]	[ START MINER, GPU ID: 0, boost factor: 32, throughput: 16777216 ]
68: [ 1][t 4][2021-11-10 15:13:45.7987026][cuda.cpp:205]	[ GPU #0: bitcredit_setBlockTarget invalid device symbol ]
69: [ 3][t 4][2021-11-10 15:22:53.2207998][credits.cu:29]	[ START MINER, GPU ID: 0, boost factor: 32, throughput: 16777216 ]
70: [ 1][t 4][2021-11-10 15:22:53.2212183][cuda.cpp:205]	[ GPU #0: bitcredit_setBlockTarget invalid device symbol ]
 DONE in 779.7ms
71: synchronization: 100%
72: synsynchronization: DONE in 505.6ms
73: synchronization: 10synsynchronization: DONE in 669.4ms
...

Hope it helps.

from pow-miner-gpu.

Hi there,
I have the same issue on Windows 10*64
NVIDIA GeForce GTX 780

C:\minertools-cuda-windows-x86-64>tonlib-cuda-cli.exe -V
tonlib-cli build information: [ Commit: 81ca0b840e8cf1b438361e14c2c94d0988f09b76, Date: 2021-11-08 00:19:18 +0300]

C:\minertools-cuda-windows-x86-64>tonlib-cuda-cli.exe -v 3 -C global.config.json -e "pminer start kf91o4NNTryJ-Cw3sDGt9OTiafmETdVFUMvylQdFPoOxIsLm EQCfw5MgLzwIEkJZwTltEnXWvIvOka1snM35yAjJAQ-njP5c 0 64 0"
[ 3][t 1][2021-11-11 13:47:40.9179265][TonlibClient.cpp:2108][!Tonlib]  Use blockchain name from MASTER config: 'mainnet' (was '')
Tonlib is inited
[ 1][t 1][2021-11-11 13:47:41.0207920][cuda.cpp:81][!console]   [ GPU #0: SM 3.5 NVIDIA GeForce GTX 780 ]
[ 1][t 1][2021-11-11 13:47:41.0209110][cuda.cpp:81][!console]   [ GPU #1: SM 3.5 NVIDIA GeForce GTX 780 ]
[ 3][t 1][2021-11-11 13:47:41.0212588][tonlib-cli.cpp:883][!console]    Miner #1 created
synchronization: ???
synchronization: 0%
synchronization: DONE in 515.6ms
[ 3][t 1][2021-11-11 13:47:41.6917309][tonlib-cli.cpp:796][!PowMiner]   pminer: got new options from kf91o4NNTryJ-Cw3sDGt9OTiafmETdVFUMvylQdFPoOxIsLm, seed=33884584255273417740275354792098277267, complexity=187403080593944191722212052688801659771644975614754686049472402
[ 3][t 1][2021-11-11 13:47:41.6919514][tonlib-cli.cpp:801][!PowMiner]   [ expected required hashes for success: 617877192148238 ]
[ 3][t 1][2021-11-11 13:47:41.6925224][tonlib-cli.cpp:674][!PowMiner]   pminer: start workers
[ 3][t 4][2021-11-11 13:47:41.7775496][credits.cu:29]   [ START MINER, GPU ID: 0, boost factor: 64, throughput: 33554432 ]
[ 1][t 4][2021-11-11 13:47:41.7779222][cuda.cpp:205]    [ GPU #0: bitcredit_setBlockTarget invalid device symbol ]
synchronization: 100%
synchronization: DONE in 369.9ms

Everything works perfect on my RTX 3080 Ti but I wish to have a chance to run it on my old one station.

from pow-miner-gpu.

The same error on Ubuntu.
[cuda.cpp:205] [ GPU #0: bitcredit_setBlockTarget invalid device symbol ]

System: Ubuntu 20.04
Video Card: GeForce GTX 770
Driver: 470.42.01
Cuda: 11.4

from pow-miner-gpu.

[ 1][t 3][2021-11-12 03:04:52.729423524][cuda.cpp:205] [ GPU #1: bitcredit_setBlockTarget invalid device symbol ]
Linux Mint 20
1060
Driver Version: 450.51.06
CUDA Version: 11.0

from pow-miner-gpu.

Windows 10 x64, Geforce GTX950m
Same problem

from pow-miner-gpu.

Windows 10 x64, Geforce 950M
AND
Windows 10 x64, Geforce 670
Same problem

from pow-miner-gpu.

Hey guys, problem has been solved by running OpenCL build instead of CUDA. However hashrate on 780GTX is very dissapointing.

from pow-miner-gpu.

Hey guys, problem has been solved by running OpenCL build instead of CUDA. However hashrate on 780GTX is very dissapointing.

So this is a bad decision.
The problem is not solved.

from pow-miner-gpu.

Hey guys, problem has been solved by running OpenCL build instead of CUDA. However hashrate on 780GTX is very dissapointing.

So this is a bad decision. The problem is not solved.

Could it be because of weak GPU not OpenCL problem?

from pow-miner-gpu.

Hey guys, problem has been solved by running OpenCL build instead of CUDA. However hashrate on 780GTX is very dissapointing.

can you tell us more about the solution?

from pow-miner-gpu.

Hey guys, problem has been solved by running OpenCL build instead of CUDA. However hashrate on 780GTX is very dissapointing.

can you tell us more about the solution?

Use OpenCL zip from here instead of CUBA one. The rest settings are the same.

from pow-miner-gpu.

Hey guys, problem has been solved by running OpenCL build instead of CUDA. However hashrate on 780GTX is very dissapointing.

"Основной проблемой реализации OpenCL от NVidia является низкая производительность по сравнению с CUDA" (https://habr.com/ru/post/72247), так что вряд-ли это эффективный метод решения проблемы.

from pow-miner-gpu.

The same problem, win10 64, 750TI.

from pow-miner-gpu.

The same problem on Windows 10.
Video Card: GeForce GTX 770
Driver: 472.12
Cuda: 11.4.3

from pow-miner-gpu.

This is not the problem. Or just use the opencl miner which will work fine

It's not suliton if performance is low

from pow-miner-gpu.

Hey guys, problem has been solved by running OpenCL build instead of CUDA. However hashrate on 780GTX is very dissapointing.

"Основной проблемой реализации OpenCL от NVidia является низкая производительность по сравнению с CUDA" (https://habr.com/ru/post/72247), так что вряд-ли это эффективный метод решения проблемы.

Outdated (2009) post. For now OpenCL slightly slower than CUDA

from pow-miner-gpu.

Outdated (2009) post. For now OpenCL slightly slower than CUDA

Will there be an update for working CUDA?

from pow-miner-gpu.

I fully support it! Developers, solve the problem!

Это не решение, а костыль, который не использует все ресурсы видеокарты

This is not a solution, but a crutch that does not use all the resources of the video card

from pow-miner-gpu.

D:\Minertools>tonlib-cuda-cli.exe -V
tonlib-cli build information: [ Commit: b44bffbffa900c56489cc9586c873adaa5b0bd25, Date: 2021-11-16 19:38:32 +0300]

D:\Minertools>tonlib-cuda-cli.exe -v 3 -C global.config.json -e "pminer start kf8SYc83pm5JkGt0p3TQRkuiM58O9Cr3waUtR9OoFq716lN- EQBv1grmRESid7mMmTQDd_lGJSZ1dh7QlAlG23DeKc30g5kj 0 32 0"
[ 3][t 1][2021-11-16 23:48:28.0467758][TonlibClient.cpp:2108][!Tonlib]  Use blockchain name from MASTER config: 'mainnet' (was '')
[ 2][t 1][2021-11-16 23:48:28.0687556][tonlib-cli.cpp:285][!console]    Tonlib is inited
[ GPU #0: SM 5.0 NVIDIA GeForce GTX 950M ]
[ 3][t 1][2021-11-16 23:48:28.1009003][tonlib-cli.cpp:888][!console]    Miner #1 created
[ 2][t 1][2021-11-16 23:48:28.6860850][tonlib-cli.cpp:801][!PowMiner]   pminer: got new options from kf8SYc83pm5JkGt0p3TQRkuiM58O9Cr3waUtR9OoFq716lN-, seed=145842340426808058230022360004472202638, complexity=71488823370706088821063697944787930557331696836802571304284960
[ 3][t 1][2021-11-16 23:48:28.6865739][tonlib-cli.cpp:806][!PowMiner]   [ expected required hashes for success: 1619722968958028 ]
[ 3][t 1][2021-11-16 23:48:28.6883586][tonlib-cli.cpp:674][!PowMiner]   pminer: start workers
[ 2][t 4][2021-11-16 23:48:28.7553183][credits.cu:29]   [ START MINER, GPU ID: 0, boost factor: 32, throughput: 16777216 ]
[ 1][t 4][2021-11-16 23:48:28.7561439][cuda.cpp:209]    [ GPU #0: bitcredit_setBlockTarget invalid device symbol ]
[pid 9632] [time 1637106508] Signal: 22

D:\Minertools>goto _minerstart

D:\Minertools>tonlib-cuda-cli.exe -V
tonlib-cli build information: [ Commit: b44bffbffa900c56489cc9586c873adaa5b0bd25, Date: 2021-11-16 19:38:32 +0300]

D:\Minertools>tonlib-cuda-cli.exe -v 3 -C global.config.json -e "pminer start kf8SYc83pm5JkGt0p3TQRkuiM58O9Cr3waUtR9OoFq716lN- EQBv1grmRESid7mMmTQDd_lGJSZ1dh7QlAlG23DeKc30g5kj 0 32 0"
[ 3][t 1][2021-11-16 23:48:28.8505122][TonlibClient.cpp:2108][!Tonlib]  Use blockchain name from MASTER config: 'mainnet' (was '')
[ 2][t 1][2021-11-16 23:48:28.8714742][tonlib-cli.cpp:285][!console]    Tonlib is inited
[ GPU #0: SM 5.0 NVIDIA GeForce GTX 950M ]
[ 3][t 1][2021-11-16 23:48:28.9082582][tonlib-cli.cpp:888][!console]    Miner #1 created

After update miner

from pow-miner-gpu.

Hi,all

Please look at compatibility matrix for CUDA/GPU: https://en.wikipedia.org/wiki/CUDA#GPUs_supported

For example, to build this project for GeForce GTX 770 GPU you'll need CUDA SDK 10.x or earlier versions (support compute 3.0 level). Current version of CUDA SDK 11.5 support only cards at compute 3.5 or higher levels.

Also this way to set compute level is not working for me:

This patch seems to work (compute level 3.0):

diff --git a/crypto/CMakeLists.txt b/crypto/CMakeLists.txt
index 02d3c7b..57af07b 100644
--- a/crypto/CMakeLists.txt
+++ b/crypto/CMakeLists.txt
@@ -311,12 +311,12 @@ if(MINERCUDA)
   enable_language(CUDA)
 
   find_package(CUDA REQUIRED)
-  SET(CUDA_NVCC_FLAGS “-arch=sm_11”)
   include_directories("${CUDA_INCLUDE_DIRS}")
 
   add_library(pow-miner-cuda-lib util/Miner.cpp util/MinerCuda.cpp util/Miner.h util/cuda/miner.h util/cuda/sha256.cpp util/cuda/sha256.h
     util/cuda/cuda.hpp util/cuda/cuda.cpp util/cuda/cuda_vector.h util/cuda/credits.cu util/cuda/cuda_credits.cu)
   target_include_directories(pow-miner-cuda-lib PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}>)
+  target_compile_options(pow-miner-cuda-lib PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:-arch=sm_30>)
   target_link_libraries(pow-miner-cuda-lib PUBLIC ton_crypto ton_block ${CUDA_LIBRARIES})
 
   add_executable(pow-miner-cuda util/pow-miner.cpp)

from pow-miner-gpu.

Hi,all

Please look at compatibility matrix for CUDA/GPU: https://en.wikipedia.org/wiki/CUDA#GPUs_supported

For example, to build this project for GeForce GTX 770 GPU you'll need CUDA SDK 10.x or earlier versions (support compute 3.0 level). Current version of CUDA SDK 11.5 support only cards at compute 3.5 or higher levels.

Also this way to set compute level is not working for me:
...

Yeah. Last build solves the problem (possibly) on linux but unfortunately not affected windows version. Hope next release will be more consistent 🤗

from pow-miner-gpu.

The opencl version of tonminer is less then 7% lower in terms of hashrate compare to cuda. So threre is a way to avoid this bug just unsing opencl miner Разница производительности майнера тон между opencl и cuda для описанных здесь видеокарт - около 7%, не следует считать, что это баг, выход есть. Он оправдан.

Not fully understand it. The fact we have a much less efficient workaround doesn't mean that there is no bug.

from pow-miner-gpu.

Hi guys, does it possible to use GTX 950M for mining?
I use Ubuntu with CUDA

lorfindeil@Glor-home-laptop:/opt/ton-miner$ /opt/ton-miner/pow-miner-cuda -vv -g 0 -p 0 -B -F 16 -t 10 kQBWkNKqzCAwA9vjMwRmg7aY75Rf8lByPA9zKXoqGkHi8SM7 229760179690128740373110445116482216837 5391989333430127958933403017403926134727428884508114496220722049840 10000000000000000000
[ GPU #0: SM 5.0 GeForce GTX 950M ]
[ 3][t 0][2021-11-24 14:31:41.665005546][pow-miner.cpp:374] [ expected required hashes for success: 21474836480 ]
[ 2][t 0][2021-11-24 14:31:41.693725071][credits.cu:29] [ START MINER, GPU ID: 0, boost factor: 1, throughput: 524288 ]
[ 1][t 0][2021-11-24 14:31:41.693924971][cuda.cpp:209] [ GPU #0: bitcredit_setBlockTarget invalid device symbol ]

glorfindeil@Glor-home-laptop:/opt/ton-miner$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2020 NVIDIA Corporation
Built on Wed_Jul_22_19:09:09_PDT_2020
Cuda compilation tools, release 11.0, V11.0.221
Build cuda_11.0_bu.TC445_37.28845127_0

from pow-miner-gpu.

I research this error and it is not related to compute capability of 3.5- and cuda 11.4+. I check this issue using build with 10.2 CUDA and compare to windows build with 11.3+ CUDA. The error was reproduced in both way for some Kepler's GPU and even with the 472+ driver too.

from pow-miner-gpu.

As the time goes, the hash rate of the network is increasing and the value of this fix is decreasing. Since still there is no activity for this fix, seems that we shouldn't even wait for it.

from pow-miner-gpu.

the problem repdroduce even under the Linux OS under similar systems [ 1][t 3][2021-11-28 20:14:57.870103171][cuda.cpp:209] [ GPU #0: bitcredit_setBlockTarget invalid device symbol ]
So. You should use the openCL miner version in such cases.

from pow-miner-gpu.

This is a issues with incorrect initialized CMAKE_CUDA_ARCHITECTURES.
Solution:

if(${CMAKE_VERSION} VERSION_GREATER "3.22")
include(FindCUDA/select_compute_arch)
CUDA_DETECT_INSTALLED_GPUS(INSTALLED_GPU_CCS_1)
string(STRIP "${INSTALLED_GPU_CCS_1}" INSTALLED_GPU_CCS_2)
string(REPLACE " " ";" INSTALLED_GPU_CCS_3 "${INSTALLED_GPU_CCS_2}")
string(REPLACE "." "" CUDA_ARCH_LIST "${INSTALLED_GPU_CCS_3}")
SET(CMAKE_CUDA_ARCHITECTURES ${CUDA_ARCH_LIST})
message("-- CUDA_ARCHITECTURES: ${CUDA_ARCH_LIST}")
set_property(GLOBAL PROPERTY CUDA_ARCHITECTURES "${CUDA_ARCH_LIST}")
else()
SET(CUDA_NVCC_FLAGS “-arch=sm_11”)
endif()

from pow-miner-gpu.

>tonlib-cuda-cli.exe -v 3 -C global.config.json -e "pminer start kf-FV4QTxLl-7Ct3E6MqOtMt-RGXMxi27g4I645lw6MTWraV EQBv1grmRESid7mMmTQDd_lGJSZ1dh7QlAlG23DeKc30g5kj 0 4 1"
[ 3][t 1][2021-12-05 18:59:31.5428451][TonlibClient.cpp:2108][!Tonlib]  Use blockchain name from MASTER config: 'mainnet' (was '')
[ 2][t 1][2021-12-05 18:59:31.5996617][tonlib-cli.cpp:293][!console]    Tonlib is inited
[ GPU #0: SM 5.0 NVIDIA GeForce GTX 950M ]
[ 3][t 1][2021-12-05 18:59:31.6306885][tonlib-cli.cpp:1072][!console]   Miner #1 created
[ 1][t 0][2021-12-05 18:59:33.8714359][adnl-ext-connection.cpp:129][!outconn]   Client got error [WindowsError : ╨Я╨╛╨┤╨║╨╗╤О╤З╨╡╨╜╨╕╨╡ ╨╜╨╡ ╤Г╤Б╤В╨░╨╜╨╛╨▓╨╗╨╡╨╜╨╛, ╤В.╨║. ╨║╨╛╨╜╨╡╤З╨╜╤Л╨╣ ╨║╨╛╨╝╨┐╤М╤О╤В╨╡╤А ╨╛╤В╨▓╨╡╤А╨│ ╨╖╨░╨┐╤А╨╛╤Б ╨╜╨░ ╨┐╨╛╨┤╨║╨╗╤О╤З╨╡╨╜╨╕╨╡. : 10061 : Error on [fd:00000000000001FC]]
[ 1][t 2][2021-12-05 18:59:41.6391605][tonlib-cli.cpp:1093][!console]   [Error : 500 : LITE_SERVER_NETWORKadnl query timeout] (#-1136338705)
[ 2][t 1][2021-12-05 18:59:41.6399391][LastBlock.cpp:261][!LastBlock]   Failed liteServer_getMasterchainInfo [Error : 500 : LITE_SERVER_NETWORKadnl query timeout]
[ 1][t 1][2021-12-05 18:59:41.6427313][tonlib-cli.cpp:838][!PowMiner]   pminer: [Error : 500 : LITE_SERVER_NETWORKadnl query timeout] (#-1136338705)
[ 3][t 1][2021-12-05 18:59:41.6433774][tonlib-cli.cpp:706][!PowMiner]   pminer: stopped

from pow-miner-gpu.

This is a issues with incorrect initialized CMAKE_CUDA_ARCHITECTURES. Solution:

if(${CMAKE_VERSION} VERSION_GREATER "3.22") include(FindCUDA/select_compute_arch) CUDA_DETECT_INSTALLED_GPUS(INSTALLED_GPU_CCS_1) string(STRIP "${INSTALLED_GPU_CCS_1}" INSTALLED_GPU_CCS_2) string(REPLACE " " ";" INSTALLED_GPU_CCS_3 "${INSTALLED_GPU_CCS_2}") string(REPLACE "." "" CUDA_ARCH_LIST "${INSTALLED_GPU_CCS_3}") SET(CMAKE_CUDA_ARCHITECTURES ${CUDA_ARCH_LIST}) message("-- CUDA_ARCHITECTURES: ${CUDA_ARCH_LIST}") set_property(GLOBAL PROPERTY CUDA_ARCHITECTURES "${CUDA_ARCH_LIST}") else() SET(CUDA_NVCC_FLAGS “-arch=sm_11”) endif()

It seems no. Seem to be checked already. Did you run after this fix on Kepler GPU and does it work? Post you binaries please.

from pow-miner-gpu.