Help the little ol CUDA developer

[nelisky]

Well, I want to continue to pursue this CUDA thing, but funds to diverge to that are scarce, so I thought I'd try and get a little help from the community, in the form of an auction of sorts.

I need to spend something around US$400 on hardware, which I will anyway, but in an attempt to reduce the impact of that in my very own personal economy, here's what I though; I have 2000 BTCs that I'll part of (though reluctantly) to help on paying for things, which at mtgox would net me ~US$120... very far from the target. Would you like to buy some, at a higher than market rate? I'll take paypal transfers for whatever amount you care to buy, assuming you are the highest bidder.

If there are no takers, no problem, I'll work on this still and will release all my work MIT licensed just like bitcoin client. If I get more than I ask for, I'll even develop an opencl version for all you ATI users out there. If I get less or just enough, I'll polish the CUDA version I have so it detects cards and works with more than one if available, so you can build your very own Tesla based super bitcoin miner

20 blocks of 100 coins, from here to Sunday, place your bid for nr of blocks and price per block! Obviously I'll also take coins as donation, but I will not give anyone earlier access to code, or any of that, and I do no promises as to how much time I'll be able to dedicate to this, of course.

[FreeMoney]

I'd be willing to buy the coins you will eventually generate right now at below market rates. 2000BTC on Christmas (date negotiable) for $100 right now? Will you generate or receive donations for plenty more than that?

[nelisky]

I'd be willing to buy the coins you will eventually generate right now at below market rates. 2000BTC on Christmas (date negotiable) for $100 right now? Will you generate or receive donations for plenty more than that?

Hmmm, that's an interesting concept... I don't really like futures negotiation, it's not in my culture if you will. But the thing that I wonder about is how to actually control that, I mean, I know I'm honest but if between now and Xmas I don't generate a single coin (unlikely but still a statistical possibility) what would happen then? I'm doing this because it's a lot of fun for me, I love the challenge, and I'm not trying to cut any profit, but while it is really hard to establish an honest reputation it takes little more that a little misunderstanding to be stuck with a dishonest one.

I'm open to such an agreement but only under the understanding that things may go sideways, I may have to stop working on this (though I already have a good part of the work done and disclosed) and I'm not going to take this as some kind of contract binding me to do this by a set date...

And as for generation, yes I hope to generate much more, based on my current stats. Donations I expect at least 10 times what I got so far with the CUDA code I've provided, which roughly adds up to... 0.0000 That is why I'm trying to sell the coins, man.

[FreeMoney]

I'd be willing to buy the coins you will eventually generate right now at below market rates. 2000BTC on Christmas (date negotiable) for $100 right now? Will you generate or receive donations for plenty more than that?

Hmmm, that's an interesting concept... I don't really like futures negotiation, it's not in my culture if you will. But the thing that I wonder about is how to actually control that, I mean, I know I'm honest but if between now and Xmas I don't generate a single coin (unlikely but still a statistical possibility) what would happen then? I'm doing this because it's a lot of fun for me, I love the challenge, and I'm not trying to cut any profit, but while it is really hard to establish an honest reputation it takes little more that a little misunderstanding to be stuck with a dishonest one.

I'm open to such an agreement but only under the understanding that things may go sideways, I may have to stop working on this (though I already have a good part of the work done and disclosed) and I'm not going to take this as some kind of contract binding me to do this by a set date...

And as for generation, yes I hope to generate much more, based on my current stats. Donations I expect at least 10 times what I got so far with the CUDA code I've provided, which roughly adds up to... 0.0000 That is why I'm trying to sell the coins, man.

I'll take the lottery site as collateral. I'd honor all tickets and deposits of course.

Hmm, Nelisky has 2000 to unload, and the pick6 is 2100.... j/k...

The graphics card code isn't valuable tome directly or I'd just give you a prepayment for it+support. It's hot here so I'm not going to set up a generating plant. ;-)

[ribuck]

I'll start the ball rolling. I'll buy at $7 per 100 coins (up to 2000 coins).

Hopefully there is someone who is able to improve on that offer, as it's for a good cause.

[TTBit]

I want 10% of all coins generated. oh wait....

[nelisky]

I'll take the lottery site as collateral. I'd honor all tickets and deposits of course.

Hmm, Nelisky has 2000 to unload, and the pick6 is 2100.... j/k...

The graphics card code isn't valuable tome directly or I'd just give you a prepayment for it+support. It's hot here so I'm not going to set up a generating plant. ;-)

Heh, yeah, lots of money to be made on the lottery. Anyone wants to buy me out? I'd disclose the total profit I made so far but... I' ashamed to admit it

And I do have a lot of coins bet on the pick6, a grand total of 121. I gotta say, don't know what I'll do with all these coins, they're burning my crypto pockets!

Anyway, the code is actually not worth much to any individual alone, unless it is kept closed as it has been by some fellow members of the community. What I'm trying to achieve is effectively neutralize the value of that code for the benefit of the community as a whole. It sounds a bit lame, saying out loud, but the simple fact remains; as difficulty increases the generation becomes much more valuable to the few that have the mega crunchers, and a money pit for everyone else. Doesn't feel right...

[nelisky]

I have received an anonymous donation from a forum member, a wooping $50. So $350 is my current target, maybe even less if I don't get outbid on some local auctions

[nelisky]

Ahhh, go opencl... not cuda

ATI Radeon 5770 for about $130.

NVIDIA GTS 250 for under $100.

What other parts do you need? A video card should be the only thing you need to develop on. $400 could get you a whole system to work in.

I could send a SATA hard drive and DDR2 memory like a gig. If you're in the states.

Edit - Damnit, I know who you are nelisky, we've chatted and you aren't in the states. What is on your hardware list? I'd hate to ship it to you and get dinged for customs more than it's worth.

That's the exact problem, I'm not in the states

I don't even need a GPU, I've got a mac, and a very modest cuda enabled laptop running linux. The thing is, there are multiple things I want to achieve:
- Proper hw detection, need to get a completely different platform (even though I'm only working on CUDA currently). I have my mining machine all set up, a good quad core, 6GB of ram, lots of hard drive and all my linux dev environment. What it lacks? A CUDA/OpenCL GPU
- Proper tuning functions, as simple config changes yield a huge difference. But manually recompiling sucks, and I end up tuning to the one card I'm developing for.
- Multiple card support. This is really important to scale. Unfortunately my motherboard supports a single PCI-e x16 interface, I need to get another one for two (that's €75 for a used one, at least)
- OpenCL I can do even on nvidia only, but I'd sure like to get this running on ATIs. But as it stands, doing OpenCL means redoing a lot of work, and frankly I don't own a single ATI card, and while some community members offered to send me some, it always gets dropped half way through.

My current plan is to get a dual pci-e mobo and a couple of fairly good nvidia cards, aiming at one 8800GTX for €85 and a GTS250 for €60, that's current bid, we'll see how high it gets. Add S&H of about €10 for each and you have my $400 plan (including paypal's fees for EUR exchange).

But I could be convinced to tackle the opencl path, if there are enough people interested. It is really not a question of money, but rather of time to market. The cuda I have done, and needs some TLC to be widely used, whereas the opencl will take a fair amount of time to be usable, and on top of that I just need the cuda. But hey, I will do opencl sooner or later, that's for sure! I might even get myself a 5770 just to force myself into doing it!

Please do not send any hw from the states, it will get stopped at customs and I'll have to pay the ridiculous fees that usually are higher than the cost of the hw. Not to mention it takes 5 weeks to receive the damn packages

Everyone: I'm not trying to rip people off, I'm loosing a LOT of income by dedicating time to this, but only because I feel it is worth it. Having said that, if anyone wants to donate a high-end tesla GPU I will not refuse :p

[BitLex]

I'll start the ball rolling. I'll buy at $7 per 100 coins (up to 2000 coins).

Hopefully there is someone who is able to improve on that offer, as it's for a good cause.

I'm in at $7.5/100coins for up to 1000coins.

[nelisky]

I feel like puddinpop has an excellent GPU client for the CUDA. running on .3.12 an old 8800gt does just fine with it.

What the community needs is an openCL version. Both of my new machines both have ATI cards in it. You said yourself that openCL runs on both nvidia and ati, so it could definitely be the best supported across multiple platforms.

Best of luck though.

Duly noted. I'll certainly consider that as I move on. I do think that OpenCL is the way to go, I'm really just not seeing enough interest in it other than a selected few. As a matter of fact, I'm not seeing much interest in the whole GPU mining, awkwardly enough. It is, for me, a question of what I can best do with my time, as I have very little of it, but every now and then I take a stab at OpenCL, so I'm moving that way already!

I think puddinpop was working on opencl already, so maybe he'll just beat me to it. I won't share my technical view of his approach here, that has been done elsewhere already, but if you are satisfied with his code, maybe we should just bump his thread and ask for updates?

[nelisky]

Ok, Sunday is here and I think that I'm just going to let this sit for now. USD/EUR exchange is not favorable for me and the offers aren't that high so I'm going to do this step by step instead of just throwing myself head first. I have bought nVidia card for my development desktop, which will help me both dig into OpenCL and have a better notion of how to work with a more common (i.e. non laptop) setup.

I'll use the very generous donation I got to help in that, but will just put my own money for the rest (FYI a great deal, I believe; €65/$90 for a GTS250, in the box as new). I will then buy the new MoBo and an ATI card, but sometime in the future.

Thanks for the offers, though, I really appreciate it.

[tashlan]

I hate to open such an old thread, but are you still interested in this project or have an update?

I'm running Linux on a box with a beefy Nvidia card and crying my eyes out due to the dearth of miners for such boxes.  I'm not going to buy an ATI card just so I can run a miner and don't think I'm going to attempt getting OpenCL installed on RedHat.

I have limited resources to assist you, but I'm very interested.  Anything I can do to help get a decent CUDA miner released for Linux?

[nelisky]

I really don't have the bandwidth anymore, but frankly you can't expect miracles here, you may get a slight increase from using CUDA instead of OpenCL with your card, but no bit shocking increase. The problem with nVidia cards is not limited to OpenCL, it's the actual card design. So why don't you use one of the OpenCL miners instead?

[tashlan]

The miracle I was hoping for was to see an increase over running off my CPU.  I've just discovered Bitcoin and haven't run on my GPU before.  I've looked into the OpenCL miners and what it would take to get one running under Fedora.  From what I've read, it isn't worth that much pain at this time.  I also have another project about to start seriously depleting my available free time and it has priority.  I need my box (intact and working) for that project and am not going to risk breaking my setup right before I start this other project.  If it were easy to get up and running, I'd give it a quick go and test it.  Otherwise, I have to lay that idea aside for the moment.

Thank you for the quick response.

[Syke]

OpenCL on Ubuntu is easy to setup. Check out my instructions at:

http://bitcointalk.org/index.php?topic=2636.0

[Stephen Gornick]

I need my box (intact and working) for that project and am not going to risk breaking my setup right before I start this other project.  If it were easy to get up and running, I'd give it a quick go and test it.  

If a bootable flash drive had everything you need ready to go would you be more willing to give mining a try ... i.e., if it were to exist now that would be something you would try?
  http://bitcointalk.org/index.php?topic=2696.0

[Edited]