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

[cbuchner1]

I can't pass arguments to --algo=scrypt. Leads to show_usage_and_exit (Try `cudaminer --help' for more information.).

that was fixed shortly after :-)

Edit: Applecoin just entered the N 11 territory, so it's still young for us, but it will enter:
N 12 on 2014.02.22;

interesting!

[1715080083]

1715080083

[1715080083]

1715080083

[1715080083]

1715080083

[1715080083]

1715080083

[1715080083]

1715080083

[manofcolombia]

I've been running K13x1 ever since I got a the first unofficial compile someone posted but just today its been crashing on me with this config. first tells me I'm out of memory and then just flat out crashes cudaminer but not my drivers. The only thing I've been able to do to get it working again is rebooting but that takes time and effort. Is there a way to flush video memory?

Also, today my school started blocking my ports out to all my pools...which in their defense is smart but sucks to be them cuz I'm just gonna vpn back home anyway

[ktf]

May I recommend starting individual cards with -d 0 -l K5x32 -C 2 -H 1 and the other instance running -d 1 -l K5x32 -C 2 -H 1.

Ever since I migrated to CUDA 5.5 I have had inexplicable issues when trying to run multiple cards in a single miner instance.

Use the official 2013-12-18 version, that is definitely the fastest for scrypt.

Christian

I will install Cuda 5.5 when I get home and test again, but running individually didn't work ok for me so far.

Any tutorial on how to compile the latest source code from github ? I haven't done any C programming in ages, I am very rusty.

[icolsuineg]

I sent some YAC to cryptsy 2 days ago and they haven´t appeared yet.

They said they are doing some reversals and it should be up soon. It is not active trade right now. I also had few transfers and I opened a ticked when they didn't show. I never had problem with other currencies there.

[cbuchner1]

Here's a benchmark of a GT630 with GK208 chip and 2 GB GDDR5 (64 bit bus)

a single slot low profile card...

Code:

[2014-01-18 12:20:48] GPU #2: GeForce GT 630 with compute capability 3.5
[2014-01-18 12:20:48] GPU #2: interactive: 0, tex-cache: 0 , single-alloc: 0
[2014-01-18 12:20:48] GPU #2: 8 hashes / 32 MB per warp.
[2014-01-18 12:20:48] GPU #2: Performing auto-tuning (Patience...)
[2014-01-18 12:20:48] GPU #2: maximum warps: 59
[2014-01-18 12:24:05] GPU #2:    1.12 khash/s with configuration T8x5
[2014-01-18 12:24:05] GPU #2: using launch configuration T8x5
[2014-01-18 12:24:05] GPU #2: GeForce GT 630, 1.07 khash/s
[2014-01-18 12:24:28] GPU #2: GeForce GT 630, 1.04 khash/s
[2014-01-18 12:24:49] GPU #2: GeForce GT 630, 1.06 khash/s

a GT640 with 4GB does 1.80 khash/s. Hmm, not sure what to think of the GT630...

Christian

[ManIkWeet]

I will install Cuda 5.5 when I get home and test again, but running individually didn't work ok for me so far.

Any tutorial on how to compile the latest source code from github ? I haven't done any C programming in ages, I am very rusty.

Quite easy:
Get Visual Studio Pro 2010 (Trial if you must)
Get Cuda SDK 5.5
Get the source code from github (google cudaminer)
Exctract source code in it's own folder
Get libraries from first post by Christian (almost 50mb)
Extract libraries to the same folder that the source code folder is in, so not to the source code folder but one up.
Open the .sln file in the source code with Visual Studio.
Select which version you want to compile, and compile it.
If it errors:

Go to your Microsoft Visual Studio 10.0\VC\bin folder and rename or relocate cvtres.exe and do the same with another cvtres.exe under Microsoft Visual Studio 10.0\VC\bin\amd64, then run Visual Studio 2010 and try compiling.

If it doesn't work, just change everything back.

Page 130, you're welcome.

[dereinehalt]

 -C 0 -H 1 -i 0 -l T24x20 -o http://localhost:.......       mit gtx780  410khash/s

geht da mehr??

[cbuchner1]

-C 0 -H 1 -i 0 -l T24x20 -o http://localhost:.......       mit gtx780  410khash/s

geht da mehr??

try

-H 1 -i 0  -T12x32

also try overclocking

NOTE: on the very latest github based cudaminer this launch config is named -T48x32.
However for scrypt mining I still recommend the 2013-12-18 release as it's faster

[dereinehalt]

do crashes the cudaminer with T12x32. what am I doing wrong?
Overclocking is already
I have between 410-418 khash / s
the memory is belegft about 65%

[bathrobehero]

a GT640 with 4GB does 1.66 khash/s. Hmm, not sure what to think of this card.

Christian

I guess it starts to become more attractive at N 15.

My GTX 660 does around 2.7 kH/s at N 14 and only aroung 0.5 kH/s at N 15. Granted it has half the memory.

Problem is with the GT640 that even if it is decent at N 15, only YAC will be going it and let's be honest YAC is not in a great spot.
On the other hand, until YAC enters N 15 (31 May 2014) we have a long way, so yeah, I'm not sure how I feel about this card either.

I think I'm going to end up having one of each of the most promising cards - including a GT640 - in my rig.... and I thought the hard part would be to decide between ATI/AMD and NVIDIA, but ASIC made that decision easier.

[cbuchner1]

a GT640 with 4GB does 1.66 khash/s. Hmm, not sure what to think of this card.

I guess it starts to become more attractive at N 15

I just found out I can get 1.80 kHash/s now with Nfactor=14 (Linux, no overclock). I will benchmark for factor 15...

Code:

[2014-01-18 14:40:45] 1 miner threads started, using 'scrypt-jane' algorithm.
[2014-01-18 14:40:45] Given scrypt-jane parameters: 15
[2014-01-18 14:40:45] Nfactor is 15 (N=65536)!
[2014-01-18 14:40:45] GPU #1: GeForce GT 640 with compute capability 3.0
[2014-01-18 14:40:45] GPU #1: interactive: 0, tex-cache: 2D, single-alloc: 1
[2014-01-18 14:40:45] GPU #1: 8 hashes / 64 MB per warp.
[2014-01-18 14:40:45] GPU #1: Performing auto-tuning (Patience...)
[2014-01-18 14:40:45] GPU #1: maximum warps: 63
[2014-01-18 14:48:19] GPU #1:    0.65 khash/s with configuration K4x12
[2014-01-18 14:48:19] GPU #1: using launch configuration K4x12
[2014-01-18 14:48:26] GPU #1: GeForce GT 640, 0.65 khash/s
[2014-01-18 14:48:26] Total: 0.65 khash/s
[2014-01-18 14:48:32] GPU #1: GeForce GT 640, 0.59 khash/s
[2014-01-18 14:48:32] Total: 0.59 khash/s

beats your GTX 660

[ozie]

GT 630/640 (GK208), GT 630 (GK107) and GT 640 (GK107) Specs:


Unified Shaders : Texture mapping units : Render output units

GK107 (630)  -> 192:16:16
GK107 (640)  -> 384:32:16
GK208          -> 384:16:8

Maybe the 8 ROPs are hurting performance of the card?

I dont think its related to the 64bit RAM Bus because the GT 640 4 GB has GDDR3 which even at 128bit has less troughput than GDDR5 of the 630.

Power consumption of the 630 GK208 would have also been too nice for getting 1,8 kh/s out of it 

[bathrobehero]

I can't pass arguments to --algo=scrypt. Leads to show_usage_and_exit (Try `cudaminer --help' for more information.).

that was fixed shortly after :-)

It still won't let me (cudaminer.exe --algo=scrypt:2048 -> show_usage_and_exit).

[cbuchner1]

It still won't let me (cudaminer.exe --algo=scrypt:2048 -> show_usage_and_exit).

ah, right. there was a missing else condition in the code...

[Mikk36]

Does anyone here provide daily builds of the Cudaminer?

[cbuchner1]

some promising first results for lookup_gap 2

Code:

cudaminer --algo=scrypt-jane -i 0 -b 32768 -L 2 -l K70x2 

[2014-01-18 17:56:35] Nfactor is 14 (N=32768)!
[2014-01-18 17:56:35] GPU #0: GeForce GTX 760 with compute capability 3.0
[2014-01-18 17:56:35] GPU #0: interactive: 0, tex-cache: 0 , single-alloc: 0
[2014-01-18 17:56:35] GPU #0: 8 hashes / 16 MB per warp.
[2014-01-18 17:56:35] GPU #0: using launch configuration K70x2
[2014-01-18 17:56:37] GPU #0: GeForce GTX 760, 3.66 khash/s
[2014-01-18 17:57:35] GPU #0: GeForce GTX 760, 3.78 khash/s
[2014-01-18 17:57:54] GPU #0: GeForce GTX 760, 3.73 khash/s
[2014-01-18 17:57:54] accepted: 1/1 (100.00%), 3.73 khash/s (yay!!!)

makes me think we can get 5 kHash/s or more on 780, 780Ti etc.

this code is not yet committed because it hurts performance when lookup gap is not in use.

[bathrobehero]

this code is not yet committed because it hurts performance when lookup gap is not in use.

b... b... but I want to try it 
All I managed to do trying to figure this out on my own is crashing the driver about 20 times.

[cbuchner1]

this code is not yet committed because it hurts performance when lookup gap is not in use.

b... b... but I want to try it  
All I managed to do trying to figure this out on my own is crashing the driver about 20 times.

I might get 4.7 kH with a gap of 3 on the GTX 760 but I still get some validation errors.Still... Very exciting!

[bathrobehero]

I might get 4.7 kH with a gap of 3 on the GTX 760 but I still get some validation errors.Still... Very exciting!

Bah.. I give up, waiting for the driver to recover over and over again is not that exciting
GL to you though!

Edit: This is as far as I could get, just jamming the numbers in without knowing what I'm doing 

[Scribbles646]

(Running Scrypt on Christians 12-10-2013 x64 binary)
Wanted to note odd behavior on my GT 750m's when my Lenovo laptop was set to Optimized Battery Health, while still plugged in and set to Maximum Performance, it cut performance and I ended up with some really random autoconfigs, until I set it back Maximum Battery Life mode. Figured I'd note it in case someone else is having such an issue.