Topic: Create vanity bitcoin addresses four times as fast - page 2. (Read 32015 times)

You'll want to read this: https://bitcointalksearch.org/topic/m.2606437
and this: http://bitcoin.stackexchange.com/questions/22748/if-bitcoin-addresses-are-truely-random-then-why-are-some-more-difficult-to-gene/22786#22786

tl;dr: some prefixes are simply less likely to occur.  Wait 'til you try generating '1111111......'

Hi all

I was playing with oclvanitygen64 and vanitygen64 for a day or so,
and all works fine so far. I like it a lot. Thanks to the devs !

Here is my QUESTION, if someone could respond please with some insights?

I want to create 2 BTC addresses both starting with a prefix of the same length(7) using the same parameters:

(1)  oclvanitygen64.exe -v -o output.txt 1BTCUSD    (Prefix difficulty: 15058417127)
(2)  oclvanitygen64.exe -v -o output.txt 1TRADER    (Prefix difficulty: 888446610538) = ca. 60x higher

Now, to create 1TRADER it takes around 60x longer (WTF?!?) than to create 1BTCUSD
I was assuming, calculating prefixes with the same length ( here: length=7) using the same parameters
"1ABCDEF"
"1234567"
"1TRADER"
"1BTCUSD"
etc....
would have the same difficulty and thus the same probability to finish the task after more or less the same time
But, that is not the case for my example.

So why should it be more difficult and would take around 60 times longer to find 1TRADER than 1BTCUSD ?
Here are the detailed screen dumps
(test for both was done on the same machine:
windows 8.1 64bit i7 8core each 3.5GHz  sporting a NVIDIA GeForce GTX 650)
-----------------------------------------------------
oclvanitygen64.exe -v -o output.txt 1BTCUSD

Prefix difficulty:          15058417127 1BTCUSD
Difficulty: 15058417127
Device: GeForce GTX 650
Vendor: NVIDIA Corporation (10de)
Driver: 331.65
Profile: FULL_PROFILE
Version: OpenCL 1.1 CUDA
Max compute units: 2
Max workgroup size: 1024
Global memory: 1073741824
Max allocation: 268435456
OpenCL compiler flags: -DPRAGMA_UNROLL -cl-nv-verbose
Loading kernel binary 00473ecd85a40ab3481d30c8ef3aa08c.oc
Grid size: 2048x1024
Modular inverse: 2048 threads, 1024 ops each
Using OpenCL prefix matcher
GPU idle: 3.19%
[6.72 Mkey/s][total 81788928][Prob 0.5%][50% in 25.7min]



oclvanitygen64.exe -v -o output.txt 1TRADER

Prefix difficulty:         888446610538 1TRADER
Difficulty: 888446610538
Device: GeForce GTX 650
Vendor: NVIDIA Corporation (10de)
Driver: 331.65
Profile: FULL_PROFILE
Version: OpenCL 1.1 CUDA
Max compute units: 2
Max workgroup size: 1024
Global memory: 1073741824
Max allocation: 268435456
OpenCL compiler flags: -DPRAGMA_UNROLL -cl-nv-verbose
Loading kernel binary 00473ecd85a40ab3481d30c8ef3aa08c
Grid size: 2048x1024
Modular inverse: 2048 threads, 1024 ops each
Using OpenCL prefix matcher
GPU idle: 3.30%
[6.74 Mkey/s][total 90177536][Prob 0.0%][50% in 1.1d]


Thanks in advance

Did that, but still not work.
win 7 64bit msi r9 280x

I was able to change pattern.c, but am having so much trouble getting it to compile correctly.

I use vanitygen purely for fun. I never use the addresses but I will give this a go in the interest of fun.
Not a programmer but curious.. how did you manage to do it 4 times faster

anyways downloadingf now!

I have same problem. If you find workaround - please share!

https://bitcointalksearch.org/topic/m.7171109

Does this support vanitygen's -r?

vanitygen -r "^1[a-z]$" -k

Where it searches for things like only lower cases?

thanks for heads up, i'll have to look into regex.  I was thinking of doing with a c++ code, but haven't coded in a while, so don't know how quickly i could get that done.

Tried that, but the GPUs were not detected. Tried the 14.4 beta drivers, and the speed is still the same. Is the SDK version 2.7?

Nope, you are wasting your GPUs work, try to downgrade your driver to 12.10 (they contain a compatible SDK version)

I also get awful luck, going to 98% probability 4 times with still no match. I also get these messages:

Match idx: 5
CPU hash: etc.
GPU hash: etc.

The hashes and match idx numbers change everytime. Is this normal?

Not without modifying the code, no - and I haven't tried compiling it in a very long time.  Shouldn't be any issue running a pre-parser over the output file to strip out everything but the private keys, though?  A regex search-and-replace should even do the job.

Why you little!

I should finish that post I was going to make before sh*t hit the fan left right and center.  At least VAMP is coming back, I guess.

These GPUs should get a better rate (a 280x is equally to an HD 7970), unfortunatly they aren't already optimized as the hd 79xx series, so you need to hopefully wait a new version from lifeboat.

I downgraded to AMD SDK 2.7, but my R9 290 is still getting just 15 Mkeys/s, with my R9 290, R9 280X and AMD A10 4-Core processor getting a total of 27 MKeys/s. Is this normal, or should I be getting a better rate?

Maybe you're not honourable enough for him to ask you. 

Is there any way to change what the output to a file is? like is there a way to just print the priv keys to file so its quicker easier to import?

I was able to change pattern.c, but am having so much trouble getting it to compile correctly.  Any help would be greatly appreciated, thx.

Tried on my gtx 770, quite impressed about the performance gain!  (29 mhash w/o overclock)

It works. Use it offline.

What would you like to know?

Waiting to hear from honorable member of this forum who has used it? I am thinking about it. So it would be great to hear from any honorable member who has already used it.