[ANN] cudapts - GPU mining for Protoshares (PTS)

[dga]

Hi, all - as discussed in a prior thread, I've open sourced my code for mining ProtoShares (PTS) on Nvidia GPUs.

The code is still a little rough around the edges but is functional and decently fast - and, at least until the difficulty catches up, a very good mining-per-watt option.

https://github.com/dave-andersen/cudapts

Tested on MacOS and Linux and known to run on GT 650m, GTX 690, Tesla K20c, and GTX 650Ti.

The code is based on ptsminer and is tied to beeeeer right now, but should be straightforward to extend.

No binaries yet, and building it requires some experience with getting CUDA running on your system.  This is still advanced-intermediate terrain, but should get better as the kinks get worked out.

Happy mining!

  -Dave

[dga]

Reserving some space for updates:

 - Restructured the code and made it easier to run from the command line this morning.  Now just:

    ./cudapts <pay-addr>

    should work for most single-GPU systems.

 - Confirmed to work on a GTX 570 by changing makefile.unix, looking for 'arch' and changing it to sm_20

[unsoindovo]

Reserving some space for updates:

 - Restructured the code and made it easier to run from the command line this morning.  Now just:

    ./cudapts <pay-addr>

    should work for most single-GPU systems.

 - Confirmed to work on a GTX 570 by changing makefile.unix, looking for 'arch' and changing it to sm_20

hi dga!
how many cpm whit this miner/GPU???

tx

[dga]

Reserving some space for updates:

 - Restructured the code and made it easier to run from the command line this morning.  Now just:

    ./cudapts <pay-addr>

    should work for most single-GPU systems.

 - Confirmed to work on a GTX 570 by changing makefile.unix, looking for 'arch' and changing it to sm_20

hi dga!
how many cpm whit this miner/GPU???

tx

It's currently similar to a fast CPU but at lower power draw.  My internal version is a little faster.  The last time I tried to put together a set of numbers, I got this:

Dual GTX 690s:  2400 c/m at 115W at 10% CPU.  (600 c/s per GPU core, 2 per card, 2 cards in the box)
650Ti:  330-340 c/m at 1% CPU.
Macbook Pro with GT 650M:  200 c/m.
Tesla K20c:  ~800c/m.

Someone reported 500-600 on an older GTX 570.

(UPDATE:  I wrote c/s in a brain-dead moment.  Meant c/m, of course.)

[cbuchner1]

Weird that you haven't blogged about this yet

Christian

[dga]

Weird that you haven't blogged about this yet

Christian

grin.  I'm going to, of course.

I kind of like the two-bit counting filter I built for the GPU scanner.  It's far from optimal, but I've got one more clever idea to optimize it more, and then I'll blog about it.

  -Dave

[SpeedDemon13]

I'm interested in complying it into a workable binary, but I don't have the experience. I'm willing to learn if you can be directed to proper sites and/or books that can help me. I learn pretty fast and be come efficient once I learn a skill.

[matt4054]

Thanks for this.

Compiling on Linux / OSX wouldn't be a problem for me but my only nVidia is on Windows right now... would definitely be interested in a Windows binary.

[dga]

Thanks for this.

Compiling on Linux / OSX wouldn't be a problem for me but my only nVidia is on Windows right now... would definitely be interested in a Windows binary.

Heard.  I'm slowly trickling in some compilation fixes that should let things work on Windows under Cygwin.  Just committed another - but I'm pretty sure there are remaining bugs.  If anyone wants to grab the latest version from git and try, LMK.

(For those who tried before by manually compiling gpuhash.cu and got an error about "restrict", that bug should be fixed.  I hope.)

  -Dave

[Sy]

A 680 GTX drawing 115W?

Besides that, very nice - now we need openCL

[cbuchner1]

A 680 GTX drawing 115W?

Besides that, very nice - now we need openCL

Power draw is very much reduced when you run memory intensive stuff that is not at all compute bound. For example my script-jane (Yacoin) implementation in CUDA makes my mining rig run at 400 Watts, whereas mining scrypt (Litecoin) will make it run at 850 Watts. Yet the mining returns are bigger with jane, so I am not complaining

nVidia doesn't play nice with OpenCL (nVidia's support is outdated and half-assed. CUDA will in most cases get you better performance over an equivalent OpenCL implementation). So don't expect an OpenCL port from someone who mainly owns nVidia gear....

Christian

[Sy]

Thanks for the info, didnt expect one from you - just wondering, will it be faster as everything else or this one finally better with cuda?

[dga]

Thanks for the info, didnt expect one from you - just wondering, will it be faster as everything else or this one finally better with cuda?

Not sure.  I'm gong to try an OpenCL one next week if nobody beats me to it.  I have a fairly different algorithmic approach in mind for AMD.  We'll see.

  -Dave

[MrMirkin]

Thanks for this.

Compiling on Linux / OSX wouldn't be a problem for me but my only nVidia is on Windows right now... would definitely be interested in a Windows binary.

Same here, would love to try this on my 570.  Please let us know when you can!

[zvs]

I have a GTX 780 that I received yesterday

so I'll make a windows binary sometime tonight, after I get around to setting it up (have to swap some power supplies too)

(beh, i'll post the other stuff on marketplace later maybe)

[dga]

I have a GTX 780 that I received yesterday

so I'll make a windows binary sometime tonight, after I get around to setting it up (have to swap some power supplies too)

(beh, i'll post the other stuff on marketplace later maybe)

I believe Archit has one working (binary only and donations to him, but works on windows! .  But help getting my version to work cleanly on windows would be appreciated - we'll have a higher-quality open source base for everyone.

[dga]

A 680 GTX drawing 115W?

Besides that, very nice - now we need openCL

I was slightly off on my power measurements - I was giving single card numbers earlier, not double, because I had the monitor on a different line than I thought I did.  Oops.  WIth the newer code, it's using more.  Dev version to be released next week if things goes well uses about 175W per card.  Rats.   (It's also faster.)

[dga]

Decent-ish update 2014-01-10:  I've committed some new changes to the repository.  Some are cosmetic, but three are important:

  1)  Memory use on the host side is reduced by about 500MB.  This may or may not matter for you.

  2)  Speed is boosted by 10-20% on a lot of platforms.  I have another speed boost patch coming next week once I've made it not horrible, but this one gets a decent chunk of the gains.

  3)  There's now a developer fee that goes to me.  Kinda. 

I'm doing an experiment with this code release in the developer fee:  It's easy to disable.  It's not hidden.  But it's also just a list of addresses that share the dev fee equally.

So here's my proposal:  If you port this software to another platform or release a binary, don't remove my address.  Instead, add yours to the list -- I've tried to make it super easy for you to get your own share.  If this works out, I'll continue to release improvements and try to make it even easier for other developers who improve upon the code, because we'll all have a reason to make software that remains open source and which is user-friendly and high performance.

If you think this is horrible, let me know and let's try to find a way to make it work better.

If you're a user who hates the idea of a dev fee, the source is yours and you can delete the addresses listed there and/or add your own.

  -Dave

[mmihai1978]

hi dga

is this guy using your source code? The project name in his screenshot looks just like yours.

https://bitcointalk.org/index.php?topic=410111.0

thanks

[relm9]

hi dga

is this guy using your source code? The project name in his screenshot looks just like yours.

https://bitcointalk.org/index.php?topic=410111.0

thanks

It seems to be the same. Gotta say that's pretty scummy, no credit to the original author and asks for donations