The 2080 TI did. around 1500 mk

So what was the problem why it didn’t work b4?

Sorry 4 the late answer, i was gone..
and thank u very much! i will try this in the coming days..

Hey JeanLuc

btw. when i start the release version 1.19

GPU: GPU #0 GeForce RTX 3090 (82x0 cores) Grid(656x128)

82x0 ? what does that mean? is there some division by 0 happening and thats why there are misaligned addresses?

i'm trying to compile the code with VisualStudio19 for CCAP 8.6 (cuda 11.1 rtx 3090)
i adjustet the path in vanitysearch.vcxproj
do i need to add compute_80,sm_80;compute_86,sm_86; to the CodeGeneration section? or change 'ReleaseSM30|x64'">compute_30,sm_30
(read somewhere that the linux driver hast not jet implementet sm86 so i didnt try on linux jet)

would there be any other changes nessecary since every cuda core can now compute two FP32 simultaniously (or one INT32 and one FP32)

anyway since i dont know much about cuda (is there some book or course anyone can recommend?) i get the generic MSB3721 error and dont know how to fix it.

can anyone help?

Fehler   MSB3721   Der Befehl ""C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\bin\nvcc.exe" -gencode=arch=compute_30,code=\"sm_30,compute_30\" --use-local-env -ccbin "C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.28.29333\bin\HostX86\x64" -x cu   -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\include" -I"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\include"     --keep-dir x64\ReleaseSM30 -maxrregcount=0 --ptxas-options=-v --machine 64 --compile -cudart static     -DWITHGPU -D_CRT_SECURE_NO_WARNINGS -DWIN32 -DWIN64 -DNDEBUG -D_CONSOLE -D_MBCS -Xcompiler "/EHsc /W3 /nologo /O2 /Fdx64\ReleaseSM30\vc142.pdb /FS /Zi  /MD " -o x64\ReleaseSM30\GPUEngine.cu.obj "C:\Users\beiex\Desktop\VS\119\VanitySearch-1.19\VanitySearch-1.19\GPU\GPUEngine.cu"" wurde mit Code 1 beendet.   VanitySearch   C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\MSBuild\Microsoft\VC\v160\BuildCustomizations\CUDA 11.1.targets   785   

just to chime in here since i have the same error since 2 month ago... and yep it also craches from 50--100 addresses and upwards



>"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\bin\cuda-memcheck.exe" VanitySearch.exe -b -t 0 -gpu -gpuId 0 -g 540,512 -m 1520000 -r 44400 -o found_0.txt -i 20.txt
========= CUDA-MEMCHECK
[Loading input file 100.0%]
VanitySearch v1.19
[Building lookup16  98.9%]
[Building lookup32 100.0%]
Search: 67714 addresses (Lookup size 42081,[1,11]) [Compressed or Uncompressed]
Start Tue Sep  1 11:49:36 2020
Base Key: Randomly changed every 44400 Mkeys
Number of CPU thread: 0
GPU: GPU #0 GeForce RTX 2080 Ti (68x64 cores) Grid(540x512)
GPUEngine: Kernel: too many resources requested for launch
========= Program hit cudaErrorLaunchOutOfResources (error 701) due to "too many resources requested for launch" on CUDA API call to cudaLaunchKernel.
=========     Saved host backtrace up to driver entry point at error
=========     Host Frame:C:\Windows\system32\DriverStore\FileRepository\nv_dispui.inf_amd64_ae0a6307e8a28cc4\nvcuda64.dll (cuProfilerStop + 0x8ff3e) [0x2ad53e]
=========     Host Frame:C:\Windows\system32\DriverStore\FileRepository\nv_dispui.inf_amd64_ae0a6307e8a28cc4\nvcuda64.dll (cuProfilerStop + 0x928e3) [0x2afee3]
=========     Host Frame:C:\Windows\system32\DriverStore\FileRepository\nv_dispui.inf_amd64_ae0a6307e8a28cc4\nvcuda64.dll [0x86ebe]
=========     Host Frame:C:\Windows\system32\DriverStore\FileRepository\nv_dispui.inf_amd64_ae0a6307e8a28cc4\nvcuda64.dll (cuProfilerStop + 0x113e8a) [0x33148a]
=========     Host Frame:C:\Windows\system32\DriverStore\FileRepository\nv_dispui.inf_amd64_ae0a6307e8a28cc4\nvcuda64.dll (cuProfilerStop + 0x12c212) [0x349812]
=========     Host Frame:C:\Users\x\Desktop\VS\VanitySearch.exe [0x2199]
=========     Host Frame:C:\Users\x\Desktop\VS\VanitySearch.exe [0x4f214]
=========     Host Frame:C:\Users\x\Desktop\VS\VanitySearch.exe [0x4f0ca]
=========     Host Frame:C:\Users\x\Desktop\VS\VanitySearch.exe [0x4a922]
=========     Host Frame:C:\Users\x\Desktop\VS\VanitySearch.exe [0x4643f]
=========     Host Frame:C:\Windows\System32\KERNEL32.DLL (BaseThreadInitThunk + 0x14) [0x17bd4]
=========     Host Frame:C:\Windows\SYSTEM32\ntdll.dll (RtlUserThreadStart + 0x21) [0x6ce51]
=========
========= Program hit cudaErrorLaunchOutOfResources (error 701) due to "too many resources requested for launch" on CUDA API call to cudaGetLastError.
=========     Saved host backtrace up to driver entry point at error
=========     Host Frame:C:\Windows\system32\DriverStore\FileRepository\nv_dispui.inf_amd64_ae0a6307e8a28cc4\nvcuda64.dll (cuProfilerStop + 0x8ff3e) [0x2ad53e]
=========     Host Frame:C:\Windows\system32\DriverStore\FileRepository\nv_dispui.inf_amd64_ae0a6307e8a28cc4\nvcuda64.dll (cuProfilerStop + 0x928e3) [0x2afee3]
=========     Host Frame:C:\Windows\system32\DriverStore\FileRepository\nv_dispui.inf_amd64_ae0a6307e8a28cc4\nvcuda64.dll [0x86ebe]
=========     Host Frame:C:\Windows\system32\DriverStore\FileRepository\nv_dispui.inf_amd64_ae0a6307e8a28cc4\nvcuda64.dll (cuProfilerStop + 0x113e8a) [0x33148a]
=========     Host Frame:C:\Windows\system32\DriverStore\FileRepository\nv_dispui.inf_amd64_ae0a6307e8a28cc4\nvcuda64.dll (cuProfilerStop + 0x12c212) [0x349812]
=========     Host Frame:C:\Users\x\Desktop\VS\VanitySearch.exe [0x1e31]
=========     Host Frame:C:\Users\x\Desktop\VS\VanitySearch.exe [0x4f0cf]
=========     Host Frame:C:\Users\x\Desktop\VS\VanitySearch.exe [0x4a922]
=========     Host Frame:C:\Users\x\Desktop\VS\VanitySearch.exe [0x4643f]
=========     Host Frame:C:\Windows\System32\KERNEL32.DLL (BaseThreadInitThunk + 0x14) [0x17bd4]
=========     Host Frame:C:\Windows\SYSTEM32\ntdll.dll (RtlUserThreadStart + 0x21) [0x6ce51]
=========
========= ERROR SUMMARY: 2 errors

no but i just tried my ryzen3 3600
with -t 11 it gets 13,5 MH/s

ok, cool, post your finding here :-)

holy shit man.. :-)

how many kW is it consuming per hour?
20? 150?

search for btc puzzle on bitcoin forum...

in short someone produced a transaction where several btc are spread over 256 addresses
the first address has 1 bit private key
the secon 2 bit priv
the third 3 bit
...
the 120'th uses a priv key comprised of the lower 120 bit's
and so on until 256 bit

all leading bits are 0

it is surmised that this is as much a puzzle to solve as it is a way to see if the security of btc is compromised "suddenly and in a significant way" if the addresses all got cracked in a short time..

what r u'r rigs like?
really 256 x v100? that would be damn impressive ;-)

then you could use the kangaroo algo and try to calculate the private key to that public key and get access to the funds
or rather to use kagaroo instead of vanity search to do it faster

AI is not really an intelligence rather a system that is a little "smarter" than traditional algos..
imagine this..
for a neural network u use a set of data with wich u train your model and leave a bit at the end that is not used for training but testing.
the training of this model however is no magic it is most likely just a bunch of "fancy" numerical aglorithms that most likely still just try the stupidest random things.. but a lot lot lot of them
then after the first run it is determined what stupid random choices actually brought on the desired result and the model "remembers" on what grounds these positive results where based on and achieved.
all that is called an "epoch" and is repeated. in the second epoch almost the same random stupid choices run is repeated but almost all of them now factor in what randomly brought on results in the first epoch.
this goes on for however many epochs you want or maybe until another epoch can't refine or beat the results of the previous run.
at that point the learning is done and cant be improved without changes to the model.

the problem in relation to this EC task here is simply that the first epoch would never yield any result'SSSSSS if even one (most likely not) so there would never be any result from wich to learn from.
and even IF u found something:
1. it would take the same time or more TO FIND as if u just used JC's prog
2. multiple results that would be needed to "learn" something aka training the model again same or more time needed
3. the math used by the EC prevents extracting anything "learnable" by whatever model u use in the AI

at least "3." is the theory and if someone would find a way to produce or force some state in with you could "infer" some information you would have found a way to produce collisions in the prosess like some researchers a few years ago.. they found a flaw in "SHA1"?! if i remember correctly..

and if that happend btc would imedietly collapse and be worthless or some realy nice whitehat hacker would work with the team on changing the algo while keeping everything under wraps..

the point is that u can give vs a list of addresses with balance in them and if it finds a private key then it is one to an address with btc in it.

self compiled 1.17 with cuda 10.2 -> NOT working
your exe 1.17       with cuda 10.2 -> working

your exe 1.17       with cuda 11   -> NOT working
your exe 1.18       with cuda 10.2 -> NOT working but no surprise
your exe 1.18       with cuda 11   -> NOT working bit of a surprise

self compiled 1.18 with cuda 11   -> not tested as of now

i was a litte frustrated because i couldnt figure out why my complies didnt work on 10.2 with 1.17 (compilation went ok exe just didnt work.. same error as above) also because a few weeks ago there was someone posting here that they didn't have any problems compiling 1.17 to a running state after a small issue he had.

and now with 1.18 and cuda 11 i'm totaly confused as to why.

PS: 1.17 with cuda 10.0 self compiled on linux was working fine, if i want to use it though i have to run this on a windows system..

thx 4 .18 !

though having the same problem i had with cuda 10.2 when self compiled but this time with you'r realeased 18.exe (cuda 11 is installed and working fine)

-check all fine
prefix search all fine (though not testet over 50 prefixes)
searching very few adresses runs a while but even below 100 adresses:

VanitySearch v1.18
Search: 73 addresses (Lookup size 37,[1,7]) [Compressed or Uncompressed]
Start Tue Jun 23 13:24:08 2020
Base Key: Randomly changed every 44400 Mkeys
Number of CPU thread: 0
GPU: GPU #0 GeForce RTX 2080 Ti (68x64 cores) Grid(544x512)
[852.29 Mkey/s][GPU 852.29 Mkey/s][Total 2^31.67][Prob 0.0%][50% in 3.76903e+31y][Found 0]  GPUEngine: Launch: an illegal memory access was encountered

is there maybe some cuda environment setting i'm missing or havent set ?!

thx up front...

10.2 really? can u tell me the exact driver version? maybe nvidia patched something like a bug or so.. i would give it another try then..

did you also run -check and it was without errors ?

did you edit cuda config or used the one from source?

yes i know and i weight in bcause i also answered him and if you can reproduce the issue.. ok.. and then priv's could be the issue..
though i don't see why that should be neccesary and also i don't have that issue    on windows 10    on system volume    self compiled or prereleased..
 

What? no.. I might be wrong or misunderstand you but I get output with -o on windows and linux without elevated privileges..
if there is nothing special in that regard to his setup it should produce output..

I rather suspect that the test setup did not find a result ?! try something very simple like 1AB as a prefix...
 still no output?...

the command output also shows the number of results at the very end [Found 0] if this does show Found 1 or more and your output file is empty then maybe something to look at with the privileges

-i is only the IN file it will not be changed. Use -o OUT file for putting potential results in a file instead of the command line output.