[ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX]

[djm34]

The dev for JackpotCoin posted that they will hardfork the coin as a "fix" to the faster algorithm code... hmmm

https://bitcointalk.org/index.php?topic=584481.msg6413849#msg6413849

the question is did they know about it previously ?

In the way the jackpotcoin.cl is written, it would be difficult for them to say they didn't know...
Ultrafast algo first then 2nd round (or not) of slow crap.
(they shouldn't have released their own miner   that's what killed that coin... I mean their profit).

Waiting for their hard-fork with impatience: keccak+somethingelse (for sure this time) just need to select the fastest second algo and
do the same trick.
This coin is definitely a joke... (unless they put one slow crap first (but still, you can always bypass 3/4 of the logic), this coin will still be a joke after the hardfork)

[cayars]

Chris,

1) I know you are busy adding new features and algorithms to the programs.  But I was wondering if there is any plans to merge the features of cudaminer and ccminer into just one program? Kind of gets confusing trying to remember which version to run for what coin etc.  Not a big deal but was just wondering if this is in the pipeline or if you plan on keeping them separate programs.

2) Also since you had originally said you were going to try and release the new killer groestl by EOM.  Have you figured out a way to do that yet being able to protect your code?

3) Have you taken a look at MNR https://bitcointalk.org/index.php?topic=577437.0 Hefty1 implementation minus the voting?  I'd imagine the HVC code with only slight modifications could work on MNR too.  Have you given this any consideration?

4) THANK YOU so much for all your hard work!

Carlo

[bigjme]

yeh i can see a lot of people are going to ask, here is 32bit and 64bit ccminer 0.6 compiled
https://mega.co.nz/#!Aw13XIYD!ETSB1GRSQdmqGt1jnlMLV4ZOhJ__qtfOSc0vOMKyUQY

Thanks bigjme. Is this Compute 3.5? I tried using your compile with my same old groestl bat file. The performance was worse. From 3100+ kh/s down to ~2450kh/s. You experiencing any difference?

on Windows this should now be built in four flavours, when done correctly

Compute 2.0  with #define MAXWELL_OR_FERMI 1   <-- Fermi
Compute 3.0  with #define MAXWELL_OR_FERMI 0   <-- old Kepler
Compute 3.5  with #define MAXWELL_OR_FERMI 0   <-- highend Kepler and GT630/640 with compute 3.5
Compute 3.5  with #define MAXWELL_OR_FERMI 1   <-- for Maxwell

the flag must be set in both .cu modules with "groestl" in its name. It affects
the speed optimization of the Groestl code.

at some point I will code a runtime selection of the MAXWELL_OR_FERMI flag.

Christian

I'm guessing the build bigjme made was for maxwell and that's why my 670 in the system is reporting crazy high hashrate, but is there a chance it's actually hashing between 1-2Mhash? lol  (I normally run separate bat files but I just left it since it looks like I've missed the train on this one anyways)

I just compiled the build straight from github sorry guys.

[Dotcommie]

I'm guessing the build bigjme made was for maxwell and that's why my 670 in the system is reporting crazy high hashrate, but is there a chance it's actually hashing between 1-2Mhash? lol  (I normally run separate bat files but I just left it since it looks like I've missed the train on this one anyways)

I just compiled the build straight from github sorry guys.

Hah, no worries.  That's just on my desktop which has 2x750s and 1x670.  750s are at about 9800khs.  My other rigs are all maxwell and work fine. That coin looks to be dead already so i'm gonna skip it anyways.

[xfelinho]

can somebody compile it for maxwell cards pls?

[bigjme]

yeh i will compile it now
christian if you put nvml into it, can't you get it to query the card for compute version then just get it to change from that?

EDIT:

soo i compiled it with #define MAXWELL_OR_FERMI 1, and compute 3.5
on groestl i am still seeing a 600khash/s drop

[cbuchner1]

yeh i will compile it now
christian if you put nvml into it, can't you get it to query the card for compute version then just get it to change from that?

EDIT:

soo i compiled it with #define MAXWELL_OR_FERMI 1, and compute 3.5
on groestl i am still seeing a 600khash/s drop

what CUDA toolkit was used? I've tested with 5.5

[bigjme]

5.5  
i will leave it to run for a while as it may not have adjusted to diff properly? no idea

UPDATE:

ok so its been running a while now, it is exactly 600khash/s drop
2.8MH/s down from 3.4MH/s

so the drop is exactly the same even with overclocks

this might sound daft, but did you test it on x1 risers? if not i wonder if the table implementation is having a cut back on speeds due to the risers?

[cbuchner1]

5.5  
i will leave it to run for a while as it may not have adjusted to diff properly? no idea

UPDATE:

ok so its been running a while now, it is exactly 600khash/s drop
2.8MH/s down from 3.4MH/s

so the drop is exactly the same even with overclocks

this might sound daft, but did you test it on x1 risers? if not i wonder if the table implementation is having a cut back on speeds due to the risers?

hey it would appear I have a typo in the cuda_groestlcoin.cu (and possibly also in the cuda_myriadgroest.cu) file.

#if MAXWELL_OR_FEMRI

should say

#if MAXWELL_OR_FERMI

instead. Then the switch should work as intended. Sorry for the inconvenience.

Christian

[antonio8]

@cbuchner1

I know this was brought up a while ago but wanted to ask again.

Have you thought about adding an hard pay to your miner yet to help support it. I know some people might not like it but I like how Claymore did his XPM miner. Hard coded at 10% but you can turn it off and it slows down the speed of the miner. I have no problem paying him the 10% fee that I find. He updates regular so I think the method works.

I see it as a win/win situation. You get paid for your time and hard work and we get coins that we would not have gotten as fast without your miner. I don't know if your ccminer is based on another miner but I have seen another where part of the fee went to the original miner's creator also.

[bigjme]

Sorry for the inconvenience.

Christian

sloppy work! lmao i am compiling it with the changes now, will let you all know

update: soo i changed it, same performance as before, was that to be expected?

[Dotcommie]

yeh i will compile it now
christian if you put nvml into it, can't you get it to query the card for compute version then just get it to change from that?

EDIT:

soo i compiled it with #define MAXWELL_OR_FERMI 1, and compute 3.5
on groestl i am still seeing a 600khash/s drop

The version you compiled originally was already compiled for maxwell. I guess the other poster was confused cause i said my 670 was giving crazy numbers but my maxwell cards were fine.

[bigjme]

The version you compiled originally was already compiled for maxwell. I guess the other poster was confused cause i said my 670 was giving crazy numbers but my maxwell cards were fine.

actually the version you had was the one with these settings
Compute 3.5  with #define MAXWELL_OR_FERMI 0   <-- highend Kepler and GT630/640 with compute 3.5

but obviously christian had his spelling issue so we will see

[nomad1109]

but obviously christian had his spelling issue so we will see

Let us all know if there is improvements please, and thanks for testing these things, I know it takes some time.

[bigjme]

Let us all know if there is improvements please, and thanks for testing these things, I know it takes some time.

for my 750Ti's i didnt see an improvement in the new release, mining groestl that is. but i believe that christian uses the code to bypass the table lookups so the cards that can take advantage do, and those that cant, dont

[atc1]

So what improvements does v0.6 of ccminer have?

[bigjme]

So what improvements does v0.6 of ccminer have?

the ability to mine jackpotcoin

[nomad1109]

the ability to mine jackpotcoin

I would have to imagine that right now Christian is about public enemy #1 for new coin devs. If there is a flaw in your coin he seems pretty adept at finding it.

There aren't a lot of guys running around that have his level of skill doing this, because it isn't like I (or most of the others just mining) would every find flaws like this.

These are new algorithms though, so I think this is all part of the growing process that others have already gone though.

It does make it all stronger in the long run, because at least Christian lets everyone know there is a flaw.

Imagine the outrage that would happen if this went unnoticed for months and the coin was at a more successful point when this came out. People would freak out.

[atc1]

So what improvements does v0.6 of ccminer have?

the ability to mine jackpotcoin

Which is also gone now that have hard-forked it?

[nomad1109]

I also notice that Groestlcoin gets it daily jump up the profits list on CoinWarz, and that can't hurt for visibility purposes.

It just showed up as #2 in the daily profit ratings, and since you can't deactivate seeing the Groestl algo coins with the Scrypt coins, there are a lot of people mining lots of Scrypt laying eyes on this new algo for the first time.