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

[Treggar]

I don't mean for YaCoin, but other coins, for example VelocityCoin, which has an N factor of 7 and cudaMiner still says 14.
I compiled multiple instances with different fixed N factors, it's just if there's a coin without any pools whatsoever, I can't tell if I'm hashing in the right gear or not.

erm, really? I thought the N factor computation takes the currently mined block's timestamp and derives the N factor from that. But maybe the parameters for the coin are hardcoded. I have to check that.

Christian

Yeah start time and N factor can be different for each coin.  No blocks validate if you don't have the right values in your miner... the miner I that was released with QQcoin had these values configurable as arguments from the command line.  Ybcminer I think it was?

[bigjme]

Ahh, so it really needs an arguement for different coins.
Hopefully the compiled cudaminer im using has the right values for yacoin? Else it may be running for nothing :-)

As a side note to Christian
I have just compiled the latest github. Compared to my last compile I am seeing a performance drop. Not a lot but it is there. Using the same launch file I get

Old - 3.77khash/s
New - 3.66khash/s

[cdogster]

Does anyone have the latest windows x64 build that they could share?  I'd like to try mining YAC with my setup.  PM me please.

[aliens]

Seems running scrypt-jane on Windows performs worse than on Linux. I'm getting max ~1.8Kh/s in Windows as opposed to ~2.3Kh/s in Linux with same hardware. Is this to be expected?

[bigjme]

It seems that way. Linux has better control over the video memory. Using windows I could never get over 2.2gb memory usage and at that I got driver crashes. In linux I now have 2.86gb out of 3gb and bo crashes. Plus it gave me an extra 1khash/s for the extra memory. I was able to go from 16x1 to 20x1 all by switching. Even minerd for cpus has a performance increase in linux :-)

[coercion]

I don't mean for YaCoin, but other coins, for example VelocityCoin, which has an N factor of 7 and cudaMiner still says 14.
I compiled multiple instances with different fixed N factors, it's just if there's a coin without any pools whatsoever, I can't tell if I'm hashing in the right gear or not.

erm, really? I thought the N factor computation takes the currently mined block's timestamp and derives the N factor from that. But maybe the parameters for the coin are hardcoded. I have to check that.

Christian

See GetNfactor() in scrypt-jane.cpp. It calculates N factor by finding the difference between the coin's start time and the current time. start time, minN, and maxN are all hard coded right now, I imagine because the code came from a yacoin miner. YBCoin randomly has the same Nfactor as Yacoin right now, but that won't always be true. I provided parameters for a few coins a couple pages back, here they are again. They are all hardcoded in the mining programs for each of them, with the exception of the cgminer scrypt-jane patch which takes them as command line arguments.

YBCoin: 1372386273, minN: 4, maxN: 30
ZcCoin: 1375817223, minN: 12, maxN: 30
FreeCoin: 1375801200, minN: 6, maxN: 32
OneCoin: 1371119462, minN: 6, maxN: 30
QQCoin: 1387769316, minN: 4, maxN: 30

Also, the interactive changes to the Titan kernel confuse me. Even with -i 0 my system is completely interactive now. I'm not complaining though, my hash rate doesn't appear to have changed.

Just snagged a xeon workstation with 5 pcie slots for $120 (saw it on craigslist and made a beeline for it), filling it up with 640's tomorrow.

[Quora]

What you've done to this cudaminer is amazing, even the old Quadro NVS 320M can mine Litecoin a little bit, from 3 to sometimes even 50 kh/s without heating up much!

[cbuchner1]

Also, the interactive changes to the Titan kernel confuse me. Even with -i 0 my system is completely interactive now. I'm not complaining though, my hash rate doesn't appear to have changed.

Just snagged a xeon workstation with 5 pcie slots for $120 (saw it on craigslist and made a beeline for it), filling it up with 640's tomorrow.

by default the --batchsize parameter is 1024. To get previous non-interactive behavior, increase to 32768. ;-)  The difference when enabling -i 1 is that an extra synchronization + 0.1ms sleep is done between kernel calls.

Christian

[cbuchner1]

I have just compiled the latest github. Compared to my last compile I am seeing a performance drop. Not a lot but it is there. Using the same launch file I get

Old - 3.77khash/s
New - 3.66khash/s

pass -b 32768 to get same performance. Or maybe 8192, 16384 ... depending on how "interactive" you want the graphics to be when running a display with the card.

[manofcolombia]

Does anyone have the latest windows x64 build that they could share?  I'd like to try mining YAC with my setup.  PM me please.

Same, I would appreciate it allot

[patoberli]

This is up and until commit 82. It's built for Windows (32 bit executable, runs under x64). Please note that I didn't get any valid results with this version and my Fermi card though!
I don't know why, but output looked like this:
[2014-01-10 12:14:05] LONGPOLL detected new block
[2014-01-10 12:14:05] GPU #0: Quadro 4000, 0.48 khash/s
[2014-01-10 12:14:44] LONGPOLL detected new block
[2014-01-10 12:14:44] GPU #0: Quadro 4000, 0.48 khash/s
[2014-01-10 12:15:26] LONGPOLL detected new block
[2014-01-10 12:15:26] GPU #0: Quadro 4000, 0.48 khash/s
[2014-01-10 12:16:26] GPU #0: Quadro 4000, 0.48 khash/s
[2014-01-10 12:16:36] LONGPOLL detected new block
[2014-01-10 12:16:37] GPU #0: Quadro 4000, 0.45 khash/s
[2014-01-10 12:16:56] LONGPOLL detected new block
[2014-01-10 12:16:56] GPU #0: Quadro 4000, 0.47 khash/s
[2014-01-10 12:17:26] GPU #0: Quadro 4000 result does not validate on CPU (i=288, s=0)!
[2014-01-10 12:17:47] LONGPOLL detected new block
[2014-01-10 12:17:47] GPU #0: Quadro 4000, 0.48 khash/s
[2014-01-10 12:18:47] GPU #0: Quadro 4000, 0.48 khash/s
[2014-01-10 12:19:17] GPU #0: Quadro 4000 result does not validate on CPU (i=38, s=1)!
[2014-01-10 12:19:21] LONGPOLL detected new block
[2014-01-10 12:19:21] GPU #0: Quadro 4000, 0.48 khash/s
[2014-01-10 12:20:14] LONGPOLL detected new block
[2014-01-10 12:20:14] GPU #0: Quadro 4000, 0.48 khash/s

Startparameter:
D:\bitcoin\yacoin\test>cudaminer.exe -a scrypt-jane -i 1 -l F12x1 -o http://yac.coinmine.pl:8882 -O username.2:password -C 2


https://www.dropbox.com/s/0yfhgko6iodecxa/commit82_32bit.rar

[cbuchner1]

This is up and until commit 82. It's built for Windows (32 bit executable, runs under x64). Please note that I didn't get any valid results with this version and my Fermi card though!
I don't know why, but output looked like this:

Startparameter:
D:\bitcoin\yacoin\test>cudaminer.exe -a scrypt-jane -i 1 -l F12x1 -o http://yac.coinmine.pl:8882 -O username.2:password -C 2


https://www.dropbox.com/s/0yfhgko6iodecxa/commit82_32bit.rar

-C 2 combined with scrypt-jane in the Fermi kernel is not yet working. Without it should validate.

Those looking for an (untrusted) cudaminer binary for testing scrypt-jane mining should look in the quoted posting!!!
patoberli was kind enough to provide it.

[patoberli]

Thanks for the info cbuchner1!
Please note that it's compiled with VS2012 Prof. I think that's the reason why I had to add the few additional files to the archive.
It somehow didn't work when I just installed the "Visual C++ Redistributable for Visual Studio 2012 Update 4" from Microsoft (installed 32 and 64 bit). Maybe a reboot of my machine would be needed, but didn't have time for that.

[Lussekatten]

We're using the "Temp Target" trick I figured out to undervolt the cards even further than MSI Afterburner/PrecisionX will allow u.

You could probably downclock your memory to get more headroom for GPU overclock.


Please spread the word about our spreadsheet/survey combo:

Have a look at the Google Docs Spreadsheet for configuration and performance data. Please enter new data using this form.

Looking at some of the entries for my card (GTX 760), i find a couple results very suspicious. There's no way a 760 can get up to ~500 khash/s. 2x760 in SLI, maybe, but that entry doesn't have any sort of information. The most I've been able to get mine to get up to is ~300 khash/s, but it crashes after a while. Stable, about ~290-292 khash/s

I also don't see how you can underclock the GPU Memory Clock -442. As soon as I set those settings, it instantly froze and crashed (Although I don't know TOO much about OC, so I could be wrong on this part).

It needs to have some sort of moderation, or require screenshot proof of achieved hash rates.

500kh/s seems a bit odd. only got my  760 to 284kh/s so...

[bathrobehero]

If I were to fire up a BAMT or SMOS linux, could I use a precompiled cudaMiner version (if someone were to share theirs with me), or would I need to use a full linux and compile it myself?

Sorry for the newb question, but I don't know much about linux. I'm willing to learn though!

[ManIkWeet]

This is up and until commit 82. It's built for Windows (32 bit executable, runs under x64). Please note that I didn't get any valid results with this version and my Fermi card though!
I don't know why, but output looked like this:

Startparameter:
D:\bitcoin\yacoin\test>cudaminer.exe -a scrypt-jane -i 1 -l F12x1 -o http://yac.coinmine.pl:8882 -O username.2:password -C 2


https://www.dropbox.com/s/0yfhgko6iodecxa/commit82_32bit.rar

This version has terrible Scrypt rates for me using: -i 0 -H 1 -l K7x32 -C 1 -m 1
The 12-18 would get 300khash/s with that while this version gets me 163khash/s; GPU usage is at 55% but CPU usage went up to 90% while it was 10% before.

[patoberli]

This version has terrible Scrypt rates for me using: -i 0 -H 1 -l K7x32 -C 1 -m 1
The 12-18 would get 300khash/s with that while this version gets me 163khash/s; GPU usage is at 55% but CPU usage went up to 90% while it was 10% before.

Use the 18.12 version for normal scrypt operations. This version is mostly for scrypt-jane.
You could also try -b 32768 for the old scrypt behavior (if I understood cbuchner1 post earlier today correct). Check the older messages from the last 24 hours.

[CaptainBeck]

What do you think is a profitable scrypt-jane coin at the moment, we are all talking YAC, but there are others.

[cbuchner1]

What do you think is a profitable scrypt-jane coin at the moment, we are all talking YAC, but there are others.

I have not researched this yet, and the support for different coins is still missing. will get to this tonight....

[ilostcoins]

Just noticed the news about scrypt-jane. I have a 650ti with 1GB DDR5 RAM. Will performance be substantially better in Linux? I only have Windows now and isn't compiling the test version to run, but want to plan ahead a little.

BTW, is the cuda miner equally advantageous compared with ATI cards in scrypt-jane whether the N is high or low?

Thanks.