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

[djm34]

Microcoin just switched Nfactor 7 to Nfactor 8 (N=512).

WARNING: cudaminer does not handle this gracefully during mining. The miner must be restarted.

yep... the fun is over...
though when it happened it was kind of scary... just right after it was still working (at a lower rate) with some "does not validate ...". Then I restart it again and got only "does not validate...".
I then restarted the computer still there hahhh. I started to think that I might have broken the card...
when back to yesterday discussion on the forum and yes it was 7 and not 8....
So needed to retune the whole stuff...  
ok I lost 1000khash/s in the process and i wonder if it is still profitable right now.... (knowing that users with large computing power won't be less affected....

May-be if this happen again, it could be better if the program is crashed (with some explanations) or to initiate when this happen a new autotune ?

Nonetheless, I made a few sell today to the place where you can trade MRC:

The first one was on BTC and I able to get 5e-3BTC fro 100k MRC
went on an other place changing to LTC and I was able to sold 100k  for 0.2LTC (a somewhat better deal)
so now I am kind of stuck with LTC...

[jots]

Yes so it could perhaps shut off the miner and launch itself again with the new nfactor

cudaminer terminating is enough- get your batch/bash script to restart it:

Code:

#!/usr/bin/env bash

while true; do
    ./cudaminer --algo=whatever --url=whatever --userpass=username:password -l Y7x32 -C 1 -i 0 -m 0 -H 1
    echo "Restarting in 3..."; sleep 3
done

[cbuchner1]

Autotune working better now on Windows? any comments?

for low n-Factor coins accuracy should be much improved...

[ghur]

Pulled the last version, autotune still seems wonky.
Also, the K kernel performance degraded since the latest stable release.
Used 32-bit binary mining scrypt for all the tests below:

On my 670 it landed on K215x8 at 270-ish khash. (-H 1 -i 0)
Using a setting I used with the latest stable, K14x16, on this release gives me 260-ish (-H 1 -i 0 -l K14x16)
The latest stable with auto tune gives me the K14x16 setting at 314-ish (-H 1 -i 0)
Autotune may have found the optimal setting for this release, but the K kernel performance has considerably dropped.

edit:

So autotune for Y, Y43x10 which runs at 270-ish (-H 1 -i 0 -l Y)
Next on this same build using a setting I generally use I get: 347-ish (-H 1 -i 0 -l Y7x32)
So for the Y kernel autotune didn't find the most optimal setting by a large margin.

I can get it to run a little faster by playing with -C and -m, but I left those out as it's a minor difference.

[bitico]

Hi,how can I get the most out of my gtx 780TI? I'm getting really frustrated trying to oc as much as possible but topping at 605kh/s  mining deogecoin :/

Can't get more out of OC due to display driver crash  

Strange my I have never seen my gpu go above 65C, and at idles at 23-25c

My .bat file was given by my pool (shibepool) and haven't made any modifications
I'm using the 18th Dec build

Code:

@ECHO off
set GPU_MAX_ALLOC_PERCENT=100
cudaminer.exe -O user.worker:password -o stratum+tcp://stratum1.shibepool.com:3333 -i 0 -H 1 -lT15x32  
PAUSE

I've heard about the z kernel, where do I get the newest build?

[bigjme]

You have to compile it from the github yourself. And you probably have the most out of it. My 780 only hits 550khash/s

[incorrect]

I've heard about the z kernel, where do I get the newest build?

Wait for the next official build. If the dev could post the latest build that he's compiled I would appreciate that.

[bitico]

You have to compile it from the github yourself. And you probably have the most out of it. My 780 only hits 550khash/s

I have been searching,(don't remember where) that some people are getting 700kh/s + on 780 ti's

Maybe in the newest build? I'll wait

[bigjme]

Most likely under linux. It does give a bit performance difference as it uses the memory better

[djm34]

Hi,how can I get the most out of my gtx 780TI? I'm getting really frustrated trying to oc as much as possible but topping at 605kh/s  mining deogecoin :/

Can't get more out of OC due to display driver crash  

Strange my I have never seen my gpu go above 65C, and at idles at 23-25c

My .bat file was given by my pool (shibepool) and haven't made any modifications
I'm using the 18th Dec build

Code:

@ECHO off
set GPU_MAX_ALLOC_PERCENT=100
cudaminer.exe -O user.worker:password -o stratum+tcp://stratum1.shibepool.com:3333 -i 0 -H 1 -lT15x32  
PAUSE

I've heard about the z kernel, where do I get the newest build?

If you want to get more hash/sec, you need to overclock a bit the card (and use a mod bios to modify the 106% TDP limit at 100% I get 650khash/sec, at 120% 700khash/sec on the latest release... however it is a bit noisy with the fan running at 90% ).

btw what does the "set GPU_MAX_ALLOC_PERCENT=100" ?

Regarding the latest release and autotune, it doesn't work very well on Vertcoin or MRC.
It gives very low hasrate (something around 50).
Yesterday I had an other problem, with -m1, the autotune didn't start at all.

[patoberli]

btw what does the "set GPU_MAX_ALLOC_PERCENT=100" ?

This is required for AMD cards so that it will run on 100% of the cards resources. I think this is needed because for cgminer. At least on my AMD builds I need to set this to even be able to run it.

[Silverwolf_Ru]

I'm still not having any luck mining anything, but Scrypt coins with the latest compiled build. Christian, please tell me what on Earth I am doing wrong with my Fermi card? Danke schon! I really need my 560Ti to be able to bring in some profit as well.

[cbuchner1]

I'm still not having any luck mining anything, but Scrypt coins with the latest compiled build. Christian, please tell me what on Earth I am doing wrong with my Fermi card? Danke schon! I really need my 560Ti to be able to bring in some profit as well.

Currently the F kernel doesn't support a lookup_gap, so it's a bad choice for scrypt-jane, unless you are aiming for low N-factor coins.

So using the X-kernel...

Code:

./cudaminer -H 2 --algo=scrypt-jane -C 2 -L 2 -l X68x2 -b 2048 -i 0 -m 1 -o stratum+tcp://yac.coinmine.pl:9088 -O ...

[2014-01-28 18:01:58] 1 miner threads started, using 'scrypt-jane' algorithm.
[2014-01-28 18:01:59] Nfactor is 14 (N=32768)!
[2014-01-28 18:01:59] GPU #0: GeForce GTX 660 Ti with compute capability 3.0
[2014-01-28 18:01:59] GPU #0: interactive: 0, tex-cache: 2D, single-alloc: 1
[2014-01-28 18:01:59] GPU #0: 8 hashes / 16.0 MB per warp.
[2014-01-28 18:01:59] GPU #0: using launch configuration X68x2
[2014-01-28 18:02:01] GPU #0: GeForce GTX 660 Ti, 2.24 khash/s
...
[2014-01-28 18:04:25] GPU #0: GeForce GTX 660 Ti, 2.79 khash/s
[2014-01-28 18:04:25] accepted: 1/1 (100.00%), 2.79 khash/s (yay!!!)
[2014-01-28 18:04:28] GPU #0: GeForce GTX 660 Ti, 2.43 khash/s
[2014-01-28 18:04:28] accepted: 2/2 (100.00%), 2.43 khash/s (yay!!!)

Now using the F-kernel... (note that lookup_gap support is not present the, hence -L 1)

Code:

./cudaminer -H 2 --algo=scrypt-jane -C 2 -L 1 -l F21x1 -b 2048 -i 0 -m 1 -o stratum+tcp://yac.coinmine.pl:9088 -O ...

[2014-01-28 18:08:18] 1 miner threads started, using 'scrypt-jane' algorithm.
[2014-01-28 18:08:18] Starting Stratum on stratum+tcp://yac.coinmine.pl:9088
[2014-01-28 18:08:18] Stratum detected new block
[2014-01-28 18:08:19] Nfactor is 14 (N=32768)!
[2014-01-28 18:08:19] GPU #0: GeForce GTX 660 Ti with compute capability 3.0
[2014-01-28 18:08:19] GPU #0: interactive: 0, tex-cache: 2D, single-alloc: 1
[2014-01-28 18:08:19] GPU #0: 32 hashes / 128.0 MB per warp.
[2014-01-28 18:08:19] GPU #0: using launch configuration F21x1
[2014-01-28 18:08:23] GPU #0: GeForce GTX 660 Ti, 1.37 khash/s
[2014-01-28 18:08:27] GPU #0: GeForce GTX 660 Ti, 1.31 khash/s
[2014-01-28 18:09:04] GPU #0: GeForce GTX 660 Ti, 1.43 khash/s
[2014-01-28 18:09:04] accepted: 1/1 (100.00%), 1.43 khash/s (yay!!!)

This test was done on Linux using a Kepler device, but running Fermi kernels. My code is OK.
Your build is broken.

The Fermi and X kernels are now compiled for compute_10 (sm_10) and should basically
run on every CUDA capable device, including legacy stuff like nVidia 8800 GTX.

Christian

[Silverwolf_Ru]

Alright, thank you for the tips and hints Christian! You are still suggesting high N-value Jane coins to be mined then? Could you please explain the difference between scrypt-jane:2048 and scrypt-jane:MRC algorithms. I might have compiled a binary, but I have no idea what they do PS! are you compiling for win64 platforms with /Ox optimisations? Could that have any effect on the CPU part of the program? A manual would surely not hurt either.

[cbuchner1]

Alright, thank you for the tips and hints Christian! You are still suggesting high N-value Jane coins to be mined then? Could you please explain the difference between scrypt-jane:2048 and scrypt-jane:MRC algorithms. I might have compiled a binary, but I have no idea what they do PS! are you compiling for win64 platforms with /Ox optimisations? Could that have any effect on the CPU part of the program? A manual would surely not hurt either.

VertCoin: scrypt:2048    <--- Salsa20/8, SHA-2 (SHA512)
here N is specified directly, as there is no coin-specific logic yet to compute N as a function of time.

scrypt-jane:2048 does not make any sense.  You're supposed to give an N-factor, not the N value.

MicroCoin: scrypt-jane:MRC  <--- ChaCha, SHA-3 (Keccak) currently Nfactor=8, N=512
Yacoin: scrypt-jane:YAC       <--ChaCha, SHA-3,  currently Nfactor=14, N=32768

alternatively scrypt-jane:14 works for Yacoin, scrypt-jane:8 works for MicroCoin.... but only at the moment.

Christian

[Silverwolf_Ru]

Alright, thank you for the tips and hints Christian! You are still suggesting high N-value Jane coins to be mined then? Could you please explain the difference between scrypt-jane:2048 and scrypt-jane:MRC algorithms. I might have compiled a binary, but I have no idea what they do PS! are you compiling for win64 platforms with /Ox optimisations? Could that have any effect on the CPU part of the program? A manual would surely not hurt either.

VertCoin: scrypt:2048    <--- Salsa20/8, SHA-2 (SHA512)
here N is specified directly, as there is no coin-specific logic yet to compute N as a function of time.

scrypt-jane:2048 does not make any sense.  You're supposed to give an N-factor, not the N value.

MicroCoin: scrypt-jane:MRC  <--- ChaCha, SHA-3 (Keccak) currently Nfactor=8, N=512
Yacoin: scrypt-jane:YAC       <--ChaCha, SHA-3,  currently Nfactor=14, N=32768

alternatively scrypt-jane:14 works for Yacoin, scrypt-jane:8 works for MicroCoin.... but only at the moment.

Christian

Finally we got some good answers, thank you very much! Those should be outlined in the binary at "--help" And pinned to your first OP post here Christian. Would lead to much less confusion!

Addition №1 I got the YAC to work finally with those parameters cudaminer --algo=scrypt-jane:YAC -o stratum+tcp://yac.coinmine.pl:9088 -i 0 That's really barebones and giving me 0.16 khashes/s with the configuration F6x1. I'll fiddle some more with the command line switches. The speed increased to 0.44 if I overclock the core and Memory with 80 MHz each.

[ktf]

Getting ~3.4kh/s with a GB GTX 660 OC and ~2.9kh/s with an Asus  GTX 660 (ok, considering that it is holding 3 monitors meanwhile) 

[Prima Primat]

Sooo Christian, I don't want to be pushy, but any progress on that next version? Specifically I'm interested in a failover option.

[incorrect]

Sooo Christian, I don't want to be pushy, but any progress on that next version? Specifically I'm interested in a failover option.

This, lol. I wish I knew how to compile it myself. 

[bathrobehero]

Sooo Christian, I don't want to be pushy, but any progress on that next version? Specifically I'm interested in a failover option.

In my opinion failover at this stage is kind of a luxury, but the simplest way to do it was to set cudaMiner to exit if it crashes instead of getting stuck in a loop and spamming the screen (counting the number of total errors and if that number reaches X in the last Y seconds, exit) so you could use a batch file where you could just list all the failovers you want.