CPUs start mining, GPUs fail instantly

[Sleinous]

I'm using exactly the same config as I was on the previous xmr-stak-nvidia :

`"gpu_threads_conf" : [ // gpu: GeForce GTX 580 architecture: 20 // memory: 1211/1536 MiB { "index" : 0, "threads" : 16, "blocks" : 32, "bfactor" : 8, "bsleep" : 100, "affine_to_cpu" : false, }, // gpu: GeForce GTX 560 Ti architecture: 21 // memory: 1211/1536 MiB { "index" : 1, "threads" : 8, "blocks" : 24, "bfactor" : 8, "bsleep" : 100, "affine_to_cpu" : false, }, // gpu: GeForce GTX 560 Ti architecture: 21 // memory: 1211/1536 MiB { "index" : 2, "threads" : 8, "blocks" : 24, "bfactor" : 8, "bsleep" : 100, "affine_to_cpu" : false, }, // gpu: GeForce GTX 560 Ti architecture: 21 // memory: 1211/1536 MiB { "index" : 3, "threads" : 8, "blocks" : 24, "bfactor" : 8, "bsleep" : 100, "affine_to_cpu" : false, },

],`

But am unable to start GPU mining, force closes xmr-stak exe

[fireice-uk]

@Sleinous Can you try lowering threads and blocks (start by halving), and give us a config that works?

[Sleinous]

@fireice-uk Config above works fine on xmr-stak-nvidia latest release (Apr 24) : I can certainly try lowering some more, but that setup was working fine for my fermi cards :) FYI previous config.txt :

"gpu_threads_conf" : [
  { "index" : 0,
    "threads" : 16, "blocks" : 32,
    "bfactor" : 8, "bsleep" :  100,
    "affine_to_cpu" : true,
  },
  { "index" : 1,
    "threads" : 8, "blocks" : 24,
    "bfactor" : 8, "bsleep" :  100,
    "affine_to_cpu" : true,
  },
  { "index" : 2,
    "threads" : 8, "blocks" : 24,
    "bfactor" : 8, "bsleep" :  100,
    "affine_to_cpu" : true,
  },
  { "index" : 3,
    "threads" : 8, "blocks" : 24,
    "bfactor" : 8, "bsleep" :  100,
    "affine_to_cpu" : true,
  },
],

I've tried halving the above config, still getting same error message. When compiling I noticed it skipped Open_CL_2.0 and used Open_CL_1.2, not sure if that's an issue though..

[fireice-uk]

Ok, that's a different issue to what used to be the problem last time then. I will let the gpu wizard @psychocrypt handle it =)

[Sleinous]

Ok cheers @fireice-uk let me know if you need any more info for testing purposes Just tried compiling a brandnew one on my GTX 970 rig and ran into this : So... I think I may just be compiling all of this wrongly.. I have everything installed though as per the compile guide

[psychocrypt]

For the gtx970 windows compile problem: The environment variables to Cuda are not set. Provide the path to the cuda installation via set CUDA_ROOT=.... or set CMAKE_PREFIX_PATH=PATHTOCUDA

Fermi: The config shown in the first post, is this the auto suggestion or du you changed the values ro the same than you used for xmr-stak-nvidia. If you changed the values please post post the config suggested by the miner. You can trigger the generation by removing the nvidia.txt file.

[psychocrypt]

please open a new issue for the compile if you still have problems. Let us not mix two issues.

[Sleinous]

@psychocrypt For original issue with fermi setup, config shown was with values that worked on xmr-stak-nvidia, autosugggest values didnt work either and here they are below :

`"gpu_threads_conf" : [ // gpu: GeForce GTX 580 architecture: 20 // memory: 1211/1536 MiB { "index" : 0, "threads" : 10, "blocks" : 32, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, // gpu: GeForce GTX 560 Ti architecture: 21 // memory: 1211/1536 MiB { "index" : 1, "threads" : 20, "blocks" : 16, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, // gpu: GeForce GTX 560 Ti architecture: 21 // memory: 1211/1536 MiB { "index" : 2, "threads" : 20, "blocks" : 16, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, // gpu: GeForce GTX 560 Ti architecture: 21 // memory: 1211/1536 MiB { "index" : 3, "threads" : 20, "blocks" : 16, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, },

],`

Does fermi require version CUDA 8 max ?

[psychocrypt]

Yes you need to use CUDA 8 because NVIDIA is not supporting fermi anymore with cuda 9. If you use cuda 9 than it is the reason. Never the less I found an issue in the auto auggestion which I will fix today. I will check if I can create a check to detect the wrong compiled architecture.

[Sleinous]

Ah, gotcha! So I guess the latest version of xmr-stak-nvidia was compiled by yourselves with cuda 8 to ensure backward compatibility with fermi & that was why it worked ? If so ill retry again later today cheers!

Télécharger Outlook pour Androidhttps://aka.ms/ghei36

---

From: psychocrypt notifications@github.com Sent: Tuesday, November 14, 2017 7:36:49 AM To: fireice-uk/xmr-stak Cc: Sleinous; Mention Subject: Re: [fireice-uk/xmr-stak] CPUs start mining, GPUs fail instantly (#110)

Yes you need to use CUDA 8 because NVIDIA is not supporting fermi anymore with cuda 9. If you use cuda 9 than it is the reason. Never the less I found an issue in the auto auggestion which I will fix today. I will check if I can create a check to detect the wrong compiled architecture.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/fireice-uk/xmr-stak/issues/110#issuecomment-344160489, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AgFLR4T21Ix2AkMIzewCTYR5Q7v9MPGsks5s2TUBgaJpZM4QcnQS.

[Sleinous]

So, been trying to install cuda 8 to get ready to compile :

Have successfully managed to install CUDA8 now, recompiling as I type this and will let you know how it goes.

EDIT : @psychocrypt it's still crashing like it was with cuda 9, i've noticed that the GTX 560 ti's are reporting 1.5GB RAM when they only have 1GB, cause for concern?

"gpu_threads_conf" :
[
  // gpu: GeForce GTX 580 architecture: 20
  //      memory: 1211/1536 MiB
  { "index" : 0,
    "threads" : 10, "blocks" : 32,
    "bfactor" : 6, "bsleep" :  25,
    "affine_to_cpu" : false,
  },
  // gpu: GeForce GTX 560 Ti architecture: 21
  //      memory: 1211/1536 MiB
  { "index" : 1,
    "threads" : 20, "blocks" : 16,
    "bfactor" : 6, "bsleep" :  25,
    "affine_to_cpu" : false,
  },
  // gpu: GeForce GTX 560 Ti architecture: 21
  //      memory: 1211/1536 MiB
  { "index" : 2,
    "threads" : 20, "blocks" : 16,
    "bfactor" : 6, "bsleep" :  25,
    "affine_to_cpu" : false,
  },
  // gpu: GeForce GTX 560 Ti architecture: 21
  //      memory: 1211/1536 MiB
  { "index" : 3,
    "threads" : 20, "blocks" : 16,
    "bfactor" : 6, "bsleep" :  25,
    "affine_to_cpu" : false,
  },

[psychocrypt]

should be solved by #116

[Sleinous]

@psychocrypt Ok cool, let me know if & when you want me to retest

[psychocrypt]

@Sleinous If you like you can test #116 (direkt download: https://github.com/psychocrypt/xmr-stak/archive/topic-checkIfBinarySupportsGPUArch.zip) -

[Sleinous]

@psychocrypt So, I can confirm your change works (it detects correctly) :

I'm guesisng this means i'm doing something wrong when compiling for Fermi SM20/21?

\xmr-stak-backend.vcxproj]
  c:\xmr-stak-topic-checkifbinarysupportsgpuarch\xmrstak\misc\executor.hpp(68):
 warning C4244: 'initializing': conversion from 'double' to 'std::size_t', poss
ible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-sta
k-backend.vcxproj]
  c:\xmr-stak-topic-checkifbinarysupportsgpuarch\xmrstak\misc\executor.hpp(68):
 warning C4244: 'initializing': conversion from 'double' to 'const std::size_t'
, possible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\x
mr-stak-backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\misc\telemetry.cpp(93)
: warning C4244: '=': conversion from 'uint64_t' to 'double', possible loss of
data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak-backend.vcx
proj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\misc\telemetry.cpp(94)
: warning C4244: '=': conversion from 'uint64_t' to 'double', possible loss of
data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak-backend.vcx
proj]
  c:\xmr-stak-topic-checkifbinarysupportsgpuarch\xmrstak\net\socks.hpp(38): war
ning C4267: 'argument': conversion from 'size_t' to 'DWORD', possible loss of d
ata [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak-backend.vcxp
roj]
  c:\xmr-stak-topic-checkifbinarysupportsgpuarch\xmrstak\net\socks.hpp(51): war
ning C4267: 'argument': conversion from 'size_t' to 'DWORD', possible loss of d
ata [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak-backend.vcxp
roj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\.\xmrstak/misc/executor.hpp(68
): warning C4244: 'initializing': conversion from 'double' to 'std::size_t', po
ssible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-s
tak-backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\.\xmrstak/misc/executor.hpp(68
): warning C4244: 'initializing': conversion from 'double' to 'const std::size_
t', possible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build
\xmr-stak-backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\net\jpsock.cpp(234): w
arning C4267: 'argument': conversion from 'size_t' to 'unsigned int', possible
loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak-bac
kend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\net\jpsock.cpp(252): w
arning C4244: 'initializing': conversion from '__int64' to 'int', possible loss
 of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak-backend
.vcxproj]
  c:\xmr-stak-topic-checkifbinarysupportsgpuarch\xmrstak\net\socks.hpp(38): war
ning C4267: 'argument': conversion from 'size_t' to 'DWORD', possible loss of d
ata [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak-backend.vcxp
roj]
  c:\xmr-stak-topic-checkifbinarysupportsgpuarch\xmrstak\net\socks.hpp(51): war
ning C4267: 'argument': conversion from 'size_t' to 'DWORD', possible loss of d
ata [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak-backend.vcxp
roj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\.\xmrstak/misc/executor.hpp(68
): warning C4244: 'initializing': conversion from 'double' to 'std::size_t', po
ssible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-s
tak-backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\.\xmrstak/misc/executor.hpp(68
): warning C4244: 'initializing': conversion from 'double' to 'const std::size_
t', possible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build
\xmr-stak-backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\net\socket.cpp(150): w
arning C4267: 'initializing': conversion from 'size_t' to 'int', possible loss
of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak-backend.
vcxproj]

"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\install.vcxproj" (default
 target) (1) ->
"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\ALL_BUILD.vcxproj" (defau
lt target) (3) ->
"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak.vcxproj" (defaul
t target) (4) ->
"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak-backend.vcxproj"
 (default target) (5) ->
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\backend\cpu\jconf.cpp(
106): warning C4267: 'argument': conversion from 'size_t' to 'rapidjson::SizeTy
pe', possible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\buil
d\xmr-stak-backend.vcxproj]

"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\install.vcxproj" (default
 target) (1) ->
"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\ALL_BUILD.vcxproj" (defau
lt target) (3) ->
"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak.vcxproj" (defaul
t target) (4) ->
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\.\xmrstak/misc/executor.hpp(68
): warning C4244: 'initializing': conversion from 'double' to 'std::size_t', po
ssible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-s
tak.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\.\xmrstak/misc/executor.hpp(68
): warning C4244: 'initializing': conversion from 'double' to 'const std::size_
t', possible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build
\xmr-stak.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\cli\cli-miner.cpp(269)
: warning C4244: 'argument': conversion from 'time_t' to 'unsigned int', possib
le loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr-stak.
vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\cli\cli-miner.cpp(532)
: warning C4244: 'initializing': conversion from 'unsigned __int64' to 'double'
, possible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\x
mr-stak.vcxproj]

"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\install.vcxproj" (default
 target) (1) ->
"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\ALL_BUILD.vcxproj" (defau
lt target) (3) ->
"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmrstak_cuda_backend.vcxp
roj" (default target) (7) ->
(CustomBuild target) ->
  CUSTOMBUILD : nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architect
ures are deprecated, and may be removed in a future release (Use -Wno-deprecate
d-gpu-targets to suppress warning). [C:\xmr-stak-topic-checkIfBinarySupportsGPU
Arch\build\xmrstak_cuda_backend.vcxproj]
  CUSTOMBUILD : nvcc warning : The -std=c++11 flag is not supported with the co
nfigured host compiler. Flag will be ignored. [C:\xmr-stak-topic-checkIfBinaryS
upportsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  CUSTOMBUILD : nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architect
ures are deprecated, and may be removed in a future release (Use -Wno-deprecate
d-gpu-targets to suppress warning). [C:\xmr-stak-topic-checkIfBinarySupportsGPU
Arch\build\xmrstak_cuda_backend.vcxproj]
  CUSTOMBUILD : nvcc warning : The -std=c++11 flag is not supported with the co
nfigured host compiler. Flag will be ignored. [C:\xmr-stak-topic-checkIfBinaryS
upportsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_core.cu(22): warning C4146: unary minus operator applied to unsigned ty
pe, result still unsigned [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build
\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_core.cu(42): warning C4018: '<': signed/unsigned mismatch [C:\xmr-stak-
topic-checkIfBinarySupportsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  CUSTOMBUILD : nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architect
ures are deprecated, and may be removed in a future release (Use -Wno-deprecate
d-gpu-targets to suppress warning). [C:\xmr-stak-topic-checkIfBinarySupportsGPU
Arch\build\xmrstak_cuda_backend.vcxproj]
  CUSTOMBUILD : nvcc warning : The -std=c++11 flag is not supported with the co
nfigured host compiler. Flag will be ignored. [C:\xmr-stak-topic-checkIfBinaryS
upportsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  CUSTOMBUILD : nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architect
ures are deprecated, and may be removed in a future release (Use -Wno-deprecate
d-gpu-targets to suppress warning). [C:\xmr-stak-topic-checkIfBinarySupportsGPU
Arch\build\xmrstak_cuda_backend.vcxproj]
  CUSTOMBUILD : nvcc warning : The -std=c++11 flag is not supported with the co
nfigured host compiler. Flag will be ignored. [C:\xmr-stak-topic-checkIfBinaryS
upportsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(177): warning C4297: 'cryptonight_extra_cpu_set_data': functio
n assumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySu
pportsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(205): warning C4297: 'cryptonight_extra_cpu_init': function as
sumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppor
tsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(206): warning C4297: 'cryptonight_extra_cpu_init': function as
sumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppor
tsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(207): warning C4297: 'cryptonight_extra_cpu_init': function as
sumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppor
tsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(208): warning C4297: 'cryptonight_extra_cpu_init': function as
sumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppor
tsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(209): warning C4297: 'cryptonight_extra_cpu_init': function as
sumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppor
tsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(210): warning C4297: 'cryptonight_extra_cpu_init': function as
sumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppor
tsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(211): warning C4297: 'cryptonight_extra_cpu_init': function as
sumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppor
tsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(212): warning C4297: 'cryptonight_extra_cpu_init': function as
sumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppor
tsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(213): warning C4297: 'cryptonight_extra_cpu_init': function as
sumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppor
tsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(214): warning C4297: 'cryptonight_extra_cpu_init': function as
sumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppor
tsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(226): warning C4297: 'cryptonight_extra_cpu_prepare': function
 assumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySup
portsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(252): warning C4018: '<': signed/unsigned mismatch [C:\xmr-sta
k-topic-checkIfBinarySupportsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(238): warning C4297: 'cryptonight_extra_cpu_final': function a
ssumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppo
rtsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(239): warning C4297: 'cryptonight_extra_cpu_final': function a
ssumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppo
rtsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(241): warning C4297: 'cryptonight_extra_cpu_final': function a
ssumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppo
rtsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(243): warning C4297: 'cryptonight_extra_cpu_final': function a
ssumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppo
rtsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(244): warning C4297: 'cryptonight_extra_cpu_final': function a
ssumed not to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySuppo
rtsGPUArch\build\xmrstak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(461): warning C4267: '=': conversion from 'size_t' to 'int', p
ossible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmrs
tak_cuda_backend.vcxproj]
  C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/xmrstak/backend/nvidia/nvcc_co
de/cuda_extra.cu(413): warning C4297: 'cuda_get_deviceinfo': function assumed n
ot to throw an exception but does [C:\xmr-stak-topic-checkIfBinarySupportsGPUAr
ch\build\xmrstak_cuda_backend.vcxproj]

"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\install.vcxproj" (default
 target) (1) ->
"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\ALL_BUILD.vcxproj" (defau
lt target) (3) ->
"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmrstak_cuda_backend.vcxp
roj" (default target) (7) ->
(ClCompile target) ->
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\backend\nvidia\jconf.c
pp(121): warning C4267: 'argument': conversion from 'size_t' to 'rapidjson::Siz
eType', possible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\b
uild\xmrstak_cuda_backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\.\xmrstak/misc/executor.hpp(68
): warning C4244: 'initializing': conversion from 'double' to 'std::size_t', po
ssible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmrst
ak_cuda_backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\.\xmrstak/misc/executor.hpp(68
): warning C4244: 'initializing': conversion from 'double' to 'const std::size_
t', possible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build
\xmrstak_cuda_backend.vcxproj]

"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\install.vcxproj" (default
 target) (1) ->
"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\ALL_BUILD.vcxproj" (defau
lt target) (3) ->
"C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmrstak_opencl_backend.vc
xproj" (default target) (8) ->
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\backend\amd\amd_gpu\gp
u.cpp(33): warning C4267: 'argument': conversion from 'size_t' to 'DWORD', poss
ible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmrstak
_opencl_backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\backend\amd\amd_gpu\gp
u.cpp(458): warning C4018: '<': signed/unsigned mismatch [C:\xmr-stak-topic-che
ckIfBinarySupportsGPUArch\build\xmrstak_opencl_backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\backend\amd\amd_gpu\gp
u.cpp(507): warning C4018: '<': signed/unsigned mismatch [C:\xmr-stak-topic-che
ckIfBinarySupportsGPUArch\build\xmrstak_opencl_backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\backend\amd\amd_gpu\gp
u.cpp(622): warning C4267: 'argument': conversion from 'size_t' to 'cl_uint', p
ossible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmrs
tak_opencl_backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\backend\amd\amd_gpu\gp
u.cpp(659): warning C4267: '=': conversion from 'size_t' to 'cl_int', possible
loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmrstak_open
cl_backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\backend\amd\amd_gpu\gp
u.cpp(941): warning C4267: '+=': conversion from 'size_t' to 'uint32_t', possib
le loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmrstak_o
pencl_backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\backend\amd\jconf.cpp(
101): warning C4267: 'argument': conversion from 'size_t' to 'rapidjson::SizeTy
pe', possible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\buil
d\xmrstak_opencl_backend.vcxproj]
  c:\xmr-stak-topic-checkifbinarysupportsgpuarch\xmrstak\backend\amd\autoAdjust
.hpp(60): warning C4267: 'initializing': conversion from 'size_t' to 'int', pos
sible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmrsta
k_opencl_backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\.\xmrstak/misc/executor.hpp(68
): warning C4244: 'initializing': conversion from 'double' to 'std::size_t', po
ssible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmrst
ak_opencl_backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\.\xmrstak/misc/executor.hpp(68
): warning C4244: 'initializing': conversion from 'double' to 'const std::size_
t', possible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build
\xmrstak_opencl_backend.vcxproj]
  C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\xmrstak\backend\amd\minethd.cp
p(205): warning C4267: 'initializing': conversion from 'size_t' to 'uint32_t',
possible loss of data [C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build\xmr
stak_opencl_backend.vcxproj]

    76 Warning(s)
    0 Error(s)

Time Elapsed 00:11:43.40

C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build>

[psychocrypt]

Could you please post the output of cmake -L . from within the build folder

[Sleinous]

@psychocrypt and here we go :

Microsoft Windows [Version 10.0.15063]
(c) 2017 Microsoft Corporation. All rights reserved.

C:\Users\Jon>cd C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build

C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build>cmake -L
CMake Error: The source directory "C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/build" does not appear to contain CMakeLists.txt.
Specify --help for usage, or press the help button on the CMake GUI.
-- Cache values
CMAKE_BUILD_TYPE:STRING=Release
CMAKE_CONFIGURATION_TYPES:STRING=Debug;Release;MinSizeRel;RelWithDebInfo
CMAKE_INSTALL_PREFIX:PATH=C:/xmr-stak-topic-checkIfBinarySupportsGPUArch/build
CMAKE_LINK_STATIC:BOOL=OFF
CPU_ENABLE:BOOL=ON
CUDA_ARCH:STRING=30;35;37;50;52;20;60;61;62
CUDA_COMPILER:STRING=nvcc
CUDA_ENABLE:BOOL=ON
CUDA_HOST_COMPILER:FILEPATH=$(VCInstallDir)bin
CUDA_KEEP_FILES:BOOL=OFF
CUDA_SDK_ROOT_DIR:PATH=CUDA_SDK_ROOT_DIR-NOTFOUND
CUDA_SHOW_CODELINES:BOOL=OFF
CUDA_SHOW_REGISTER:BOOL=OFF
CUDA_TOOLKIT_ROOT_DIR:PATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0
CUDA_USE_STATIC_CUDA_RUNTIME:BOOL=ON
HWLOC:FILEPATH=C:/xmr-stak-dep/hwloc/lib/libhwloc.lib
HWLOC_ENABLE:BOOL=ON
HWLOC_INCLUDE_DIR:PATH=C:/xmr-stak-dep/hwloc/include
LIB_EAY_DEBUG:FILEPATH=LIB_EAY_DEBUG-NOTFOUND
LIB_EAY_RELEASE:FILEPATH=C:/xmr-stak-dep/openssl/lib/libeay32.lib
MHTD:FILEPATH=C:/xmr-stak-dep/libmicrohttpd/lib/libmicrohttpd.lib
MICROHTTPD_ENABLE:BOOL=ON
MTHD_INCLUDE_DIR:PATH=C:/xmr-stak-dep/libmicrohttpd/include
OpenCL_ENABLE:BOOL=ON
OpenSSL_ENABLE:BOOL=ON
SSL_EAY_DEBUG:FILEPATH=SSL_EAY_DEBUG-NOTFOUND
SSL_EAY_RELEASE:FILEPATH=C:/xmr-stak-dep/openssl/lib/ssleay32.lib
WIN_UAC:BOOL=ON
XMR-STAK_CURRENCY:STRING=all
XMR-STAK_LARGEGRID:BOOL=ON
XMR-STAK_THREADS:STRING=0

C:\xmr-stak-topic-checkIfBinarySupportsGPUArch\build>

EDIT: Silly me... problem stems from : CUDA_ARCH:STRING=30;35;37;50;52;20;60;61;62 ??

Should maybe read CUDA_ARCH:STRING=20;21;30;35;37;50;52;20;60;61;62 ?

[psychocrypt]

Could it be that you have not cleaned you build directory after you switched from cuda 9 to cuda 8? CUDA_ARCH must have the value 20 included. cmake -DCUDA_ARCH=20 <PATH TO THE SOURCECODE> will compile for fermi only.

[Sleinous]

I think you're right, facepalm moment here, my bad, going to correct and add in SM20 and 21 and recompile for retest Cheers

[psychocrypt]

Oh I missed the 20 is included in 30;35;37;50;52;20;60;61;62 , let me check if it is an bug in my pull request.

[psychocrypt]

please do not set to 21 because it is only a virtual architecture. 20 will support also 21

[Sleinous]

@psychocrypt build folder was empty before switch to cuda8 fyi, ok standing by, I wont add 21

[psychocrypt]

There was an bug in my architecture detection. This is now fixed. If you download the code again it should work now even if you compile for all architectures.

[psychocrypt]

Could you please post the new auto suggestion and the hash rate with it.

[Sleinous]

@psychocrypt Sure thing, am recompiling now, 11 minutes till retest, will keep you posted!

[Sleinous]

@psychocrypt Here's the requested hash rates :

Bug fix works! 1350 H/s on a dual Xeon E5-2630 with 1x GTX 580 and 3x GTX 560 ti. Compare that with 1550 H/s on my old core i7 970 with 3x GTX 970 and i'd say Fermi is still a viable option, i'm still making money from these cards given current electricity prices in France.

Please find below the auto suggestion for those 4 fermi cards :

"gpu_threads_conf" :
[
  // gpu: GeForce GTX 580 architecture: 20
  //      memory: 1210/1536 MiB
  //      smx: 16
  { "index" : 0,
    "threads" : 10, "blocks" : 32,
    "bfactor" : 6, "bsleep" :  25,
    "affine_to_cpu" : false,
  },
  // gpu: GeForce GTX 560 Ti architecture: 21
  //      memory: 803/1024 MiB
  //      smx: 8
  { "index" : 1,
    "threads" : 20, "blocks" : 16,
    "bfactor" : 6, "bsleep" :  25,
    "affine_to_cpu" : false,
  },
  // gpu: GeForce GTX 560 Ti architecture: 21
  //      memory: 803/1024 MiB
  //      smx: 8
  { "index" : 2,
    "threads" : 20, "blocks" : 16,
    "bfactor" : 6, "bsleep" :  25,
    "affine_to_cpu" : false,
  },
  // gpu: GeForce GTX 560 Ti architecture: 21
  //      memory: 803/1024 MiB
  //      smx: 8
  { "index" : 3,
    "threads" : 20, "blocks" : 16,
    "bfactor" : 6, "bsleep" :  25,
    "affine_to_cpu" : false,
  },

],

Logically I should be able to optimise that GTX 580 a little more as I was getting 300+ H/s on xmr-stak-nvidia. Will look into that later. Hope all of this helps you and don't hesitate if you need anything else tested on fermi. Thanks for all the great dev work you two are putting into this!

EDIT : Bumping threads from 10 to 16 results in a 110 H/s increase on GTX 580 :

[mendo-inspired]

Just setup a 2nd we rig with a gtx 560 ti and it peaks at around 135h/s but drops out continuously. Any ideas what could be causing this?

[mendo-inspired]

Sorry should say im mining xmr via nanopool

[stefono9]

[2019-02-26 21:11:41] : Starting AMD GPU (OpenCL) thread 0, no affinity. [2019-02-26 21:11:41] : Starting AMD GPU (OpenCL) thread 1, no affinity. [2019-02-26 21:11:41] : Starting AMD GPU (OpenCL) thread 2, no affinity. [2019-02-26 21:11:41] : Starting AMD GPU (OpenCL) thread 3, no affinity. [2019-02-26 21:11:41] : NVIDIA: try to load library 'xmrstak_cuda_backend_cuda10_0' WARNING: NVIDIA cannot load backend library: xmrstak_cuda_backend_cuda10_0.dll [2019-02-26 21:11:41] : NVIDIA: try to load library 'xmrstak_cuda_backend_cuda9_2' WARNING: NVIDIA cannot load backend library: xmrstak_cuda_backend_cuda9_2.dll [2019-02-26 21:11:41] : NVIDIA: try to load library 'xmrstak_cuda_backend' CUDA [10.1/8.0] GPU#0, device architecture 35: "GeForce GT 730"... device init succeeded [2019-02-26 21:11:42] : NVIDIA: GPU configuration stored in file 'nvidia.txt' NVIDIA: found 1 potential device's [2019-02-26 21:11:42] : Starting NVIDIA GPU thread 0, no affinity. CUDA [10.1/8.0] GPU#0, device architecture 35: "GeForce GT 730"... device init succeeded [2019-02-26 21:11:42] : NVIDIA: use library 'xmrstak_cuda_backend' [2019-02-26 21:11:42] : Starting 1x thread, no affinity. [2019-02-26 21:11:42] : Fast-connecting to trtl.muxdux.com:7777 pool ... [2019-02-26 21:11:42] : Pool trtl.muxdux.com:7777 connected. Logging in... [2019-02-26 21:11:43] : Difficulty changed. Now: 1000225. [2019-02-26 21:11:43] : Pool logged in. [CUDA] Error gpu 0: C:/xmr-stak/xmrstak/backend/nvidia/nvcc_code/cuda_core.cu:766 any ideas?

[Spudz76]

Downgrade driver until it works, or just go back to 386.28 which will be perfect match for CUDA 8.0.

Outlined in docs

Best when it says CUDA [8.0/8.0] whereas you are running a recent driver with CUDA 10.1 inside. 8.0 apps (required by your Fermi level GPU) didn't even do well on 9.2 drivers, so I doubt it will begin working until you get back to 9.0 or 9.1 -- but still ideal is for driver to match the Toolkit (SDK). I run a laptop with NVS-5200M on a 9.1 providing driver and it works well. Earlier drivers had various issues with features other than the CUDA on this laptop, or weren't for mobile series chips so I had to make a compromise. Mining costs more if the backlight never shuts off, etc...

Also don't comment in old, closed, threads (open new if there isn't a recent/open issue similar).