delete

[MaGNeT]

GTX470 is now at ~170Khash/s and it didn't crash for 3 hours
Stales are still 15-20% but he'll be working on that.
Will test on my 9800GTX but got to get some sleep now first.

I only received a few PM's from people with an GTX5xx series card. Seems logic, all ATI fanboys over here

1 person with 2 GTX580 cards received our Alpha and if he gets it up and running (he has to set up Ubuntu first) we hope to get some nice results

edit:

Please people, stop asking for the CUDA-miner on PM, we won't release it yet, and no, not for 10BTC also.
There will be a public release in about a week, maybe a bit longer (it's all done in spare-time of programmer) but now it's still "work in progress".
I suck in coding so all the kudo's go to him (well, I did a bit of Basic programming in the 80's and 90's )

Source-code will be released so it can be reviewed or compiled by the ones who don't like it precompiled.
I don't know how to compile it myself but I'm sure some smart guys figure that out.

We still need a name! PM your suggestions

Is the SC2 algorithm more GPU-friendly than Scrypt? Do you also plan to release a Scrypt version so I can play around with mining Litecoins on a GPU?

It's not profitable to create a scrypt miner, as far as my knowledge goes. But I'm sure other people are working on that too.

[MaGNeT]

Right, but that puts SC mining back into the same state as BTC mining.  I (genuinely) thought CH's goal was to be different in those regards.

If one could mine the same as or better on a CPU but have an option to mine GPU as well even if the efficiency in not ideal.... is that not different?

I suspect that is more for.... educational masturbation, just to see what could be done and open up the options that are out there.  Ideally, it is likely a proof that indeed CPU's are better at mining this than GPU's.... at least for the time being.

@Coinhunter

Since I know you are watching this forum. The only reason you are mentioning GPU mining now is that I said THIS about you having GPU mining and you know people are disassembling your code. Funny I post that and just a few hours later, you're "Looking" for an OpenCL miner. LOL

Anyway Magnet, Send me what you have and I will have a couple of Cuda/Open CL experts look at it. I jus PM'd you and yes as you can see it is from an apple.com email address.  

 I am especially interested in the CUDA as I think it will be poetic justice that a "Geoforce Army" take out SC by essentially turning it into a BTC type of difficulty rather quickly.

GPU mining will effectively end CPU mining in Solidcoin and you KNOW IT, even if only one person drops a decent rig on it. A single decent GPU rig will totally over power the entire SC network.

But I highly suspect there has been one person GPU mining all alone.

~BCX~

Thanks! Yeah, the first mail from Apple I ever got
I sent you all the relevant info I have.
Thanks already for the suggestions you made, this is very promising!
I wish I could code like that.  

No prob, but realize it's not my coding! A really cool guy that just seems to be able to just bang stuff out that works did it!

We're still getting crashes due to loops of some kind. Working on it, will keep all posted.

~BCX~

Sent you the latest version. Does Amazon have a CUDA farm?

[3phase]

Sent you the latest version. Does Amazon have a CUDA farm?

uh that would be no...

But don't sweat it, it's not that hard to put together a small pool of 15-20 for testing.

Aren't their GPU instances CUDA compatible?

Yes they are. Currently 0.68$/per hour spot instances for 2xM2050 GPUs and, I might add, 2 x X5570 Xeons. Two birds in one shot.

[saethan]

Cluster GPU Quadruple Extra Large Instance

22 GB of memory
33.5 EC2 Compute Units (2 x Intel Xeon X5570, quad-core “Nehalem” architecture)
2 x NVIDIA Tesla “Fermi” M2050 GPUs
1690 GB of instance storage
64-bit platform
I/O Performance: Very High (10 Gigabit Ethernet)
API name: cg1.4xlarge

- the M2050 is the same GPU as the GTX470, not sure about the differences in the memory layout of a Tesla though

[tacotime]

pm me the latest version, I have 4x gtx570, i'll give it a try if you want

edit: give me the source code, I'm on linux amd64

[MaGNeT]

Sent you the latest version. Does Amazon have a CUDA farm?

uh that would be no...

But don't sweat it, it's not that hard to put together a small pool of 15-20 for testing.

Aren't their GPU instances CUDA compatible?

I am not familiar at all with CUDA anything, never used it.

The Tesla Fermi has CUDA support. Would be big if we could rent a farm.
I don't have the money to do it though. Once miner is Beta, could you do an attempt?

[MaGNeT]

Sent you the latest version. Does Amazon have a CUDA farm?

uh that would be no...

But don't sweat it, it's not that hard to put together a small pool of 15-20 for testing.

Aren't their GPU instances CUDA compatible?

I am not familiar at all with CUDA anything, never used it.

The Tesla Fermi has CUDA support. Would be big if we could rent a farm.
I don't have the money to do it though. Once miner is Beta, could you do an attempt?

We availability to use as much EC2 as we want that pertains to us, I will see what I can to. I was just informed that it maybe possible, If I get it cost will be ZERO.

That's cheap  

Alpha 0.1.4 cuda_miner is now stable but he had to do some coding again. Gained a bit on efficienty, ~140Khash/s per thread now for the GTX470 (running 2 threads @ 1 GPU).
You got the update?

Funny part is: the GTX470 is only consuming ~120Watts with one thread and ~165Watts running 2 threads. That's a lot lower than I got when running SETI@home so I guess we are not on the limits of the card yet.

[tacotime]

Funny part is: the GTX470 is only consuming ~120Watts with one thread and ~165Watts running 2 threads. That's a lot lower than I got when running SETI@home so I guess we are not on the limits of the card yet.

Proper coding for CUDA requires knowing both CUDA and the nVidia architecture well.  It's very easy to code for it inefficiently if you are not very familiar with it.

[MaGNeT]

Funny part is: the GTX470 is only consuming ~120Watts with one thread and ~165Watts running 2 threads. That's a lot lower than I got when running SETI@home so I guess we are not on the limits of the card yet.

Proper coding for CUDA requires knowing both CUDA and the nVidia architecture well.  It's very easy to code for it inefficiently if you are not very familiar with it.

He did some Boinc / Seti@home improvements for one of the forked crunchers so I think he knows what he's doing. But he says this is something else.
Reversengineering the SC2 client was lots of work and only thursday he could start on the miner. He had some friends who helped him, he had to admit.

I should have asked BCX earlier, Apple knows a lot of steeling and reversengineering software

I'm a total coding noob, so can't give you much of the details.

[sd]

Can anyone tell me anything about the SC2 algorithm? I'm guessing it's something recursive?

I know people who work on transferring C functions to CUDA devices using Kahn process networks. It's works very nicely for some applications. I can't afford their rates but I might get some free advice.

[tacotime]

Also, you should post the algorithm here.  If it does rely on recursion, there is likely an iterative form of the algorithm which would not be hard to make and while would performs orders of magnitude faster than the recursive algorithm (honestly, if this is what coinhunter did it's incredibly stupid).

[DeathAndTaxes]

Sent you the latest version. Does Amazon have a CUDA farm?

Yes.  They have instances w/ a pair of Tesla M2050s.

Cluster GPU Instances

http://aws.amazon.com/ec2/instance-types/

Instances of this family provide general-purpose graphics processing units (GPUs) with proportionally high CPU and increased network performance for applications benefitting from highly parallelized processing, including HPC, rendering and media processing applications. While Cluster Compute Instances provide the ability to create clusters of instances connected by a low latency, high throughput network, Cluster GPU Instances provide an additional option for applications that can benefit from the efficiency gains of the parallel computing power of GPUs over what can be achieved with traditional processors. Learn more about use of this instance type for HPC applications.

Cluster GPU Quadruple Extra Large Instance

22 GB of memory
33.5 EC2 Compute Units (2 x Intel Xeon X5570, quad-core “Nehalem” architecture)
2 x NVIDIA Tesla “Fermi” M2050 GPUs
1690 GB of instance storage
64-bit platform
I/O Performance: Very High (10 Gigabit Ethernet)
API name: cg1.4xlarge

[Bobnova]

Just shy of 2kh/w, that's insane compared to the CPU version.

I bet you could run a few more threads, too!

[MaGNeT]

We got a nice offer in SC if we stopped this project and never release source, so we decided to stop.

[DeathAndTaxes]

Oh awesome.  Another win for ScamCoin.  Got to love closed source code (ScamCoin not CUDA Miner).

[simonk83]

We got a nice offer in SC if we stopped this project and never release source, so we decided to stop.

Dammit Magnet.

[johnj]

Take the SC, cash out in BTC, 'accidentally' leak the code

[MaGNeT]

We got a nice offer in SC if we stopped this project and never release source, so we decided to stop.

Dammit Magnet.

Hmm, I understand how you and some others might feel but I hope you can all forgive me for this.
I have bills to pay, just like you...

[Bobnova]

Nope, sorry.
Can't forgive it.

Last time I heard you were dedicated to SC2's downfall, and now you're supporting it.
I'm disappointed in you.

[Ten98]

We got a nice offer in SC if we stopped this project and never release source, so we decided to stop.

Lol, from who? Show the transaction in blockexplorer?

I think what happened is Reaper was released which beat you to the bounty.

Reaper is already in it's 7th version, mtrlt is working hard on it and adding cool features every day:

http://zerosignalgame.com/reaper0.07src.zip
http://zerosignalgame.com/reaper0.07.zip (Win32 Binary)

We still need a Win64 version as well as binaries for Linux, so any compiling whizzkids out there please have a go...

Bring your geforce2mx's to the party