Author

Topic: [ANN] cudaMiner & ccMiner CUDA based mining applications [Windows/Linux/MacOSX] - page 745. (Read 3426976 times)

hero member
Activity: 756
Merit: 502
EDIT: and the linux oc soft? I want precision x for linux :p -will donate if these things come out-

happy with a basic command line tool, too? Wink
hero member
Activity: 812
Merit: 1000
Hey guys,

A friend has 2x 690s

The 690s have dual-cores so when running benchmarking the following GPUs are listed:
GPU #0
GPU #1
GPU #2
GPU #3

There are also 4 GPU's listed in the device manager in windows.

We can only get GPU #0 and GPU #1 to mine in a pool however.
Can anyone advise if they are using the hashing power of the other cores on the circuit board, or whether we have missed something in the config.

Currently using:
Quote
cudaminer.exe -H 2 -d 0,1 -l K16x16,K16x16 -i 1,0

I know the above is specified for 2 cards but not sure if this is relating to cores? Should we setup another bat file and define -d 2,3 ?
Didn't want to experiment with this incase it forced too much load onto the cards or something.

Thanks for any help in advance.



Same bat, -d 0,1,2,3 ?
sr. member
Activity: 588
Merit: 251
Hey guys,

A friend has 2x 690s

The 690s have dual-cores so when running benchmarking the following GPUs are listed:
GPU #0
GPU #1
GPU #2
GPU #3

There are also 4 GPU's listed in the device manager in windows.

We can only get GPU #0 and GPU #1 to mine in a pool however.
Can anyone advise if they are using the hashing power of the other cores on the circuit board, or whether we have missed something in the config.

Currently using:
Quote
cudaminer.exe -H 2 -d 0,1 -l K16x16,K16x16 -i 1,0

I know the above is specified for 2 cards but not sure if this is relating to cores? Should we setup another bat file and define -d 2,3 ?
Didn't want to experiment with this incase it forced too much load onto the cards or something.

Thanks for any help in advance.

hero member
Activity: 812
Merit: 1000

Here's my roadmap for continued development

1) fix the issue with PCIe bandwidth, so people with 1x risers are more happy. CUDA can do all the
    hash evaluation so we save a LOT of bus bandwidth.

2) Add blake256 for blakecoin and related coins.

3) look at implementing some failover options

4) see if I can get a monitoring API included


Nvidia will give you some big $$...would turn those 40w/350kh/s 750ti's into proper mining cards if you did.

EDIT: and the linux oc soft? I want precision x for linux :p -will donate if these things come out-
hero member
Activity: 756
Merit: 502
Here's my tentative roadmap for continued development. Note the lack of a time table Wink

1) fix the issue with PCIe bandwidth, so people with 1x risers are more happy. CUDA can do all the
    hash evaluation so we save a LOT of bus bandwidth.

2) Add blake256 for blakecoin and related coins.

3) look at implementing some failover options

4) see if I can get a monitoring API included
legendary
Activity: 1400
Merit: 1050
Nice increase in the hashrate, the 780ti runs now around 315Mhash/s (before it was around 250~260Mhash/sec)

Do you overclock? Everytime I see your hashrates they're always a bit higher than mine. My 780ti is getting ~280Mhash/sec with:
--algo=keccak -d gtx780 -m 1 -L 1024 -l T2880x24
Card is stack EVGA SC clocks.
Yes It is overclocked (on a mod bios)
TDP 120% (that's why I modded the bios)
Temp Limit: 75°C
Core Clock: 1240MHz    (in MSI AB: +230)
Mem Clock: 3500Mhz (stock)

here the command line I use:
cudaminer.exe --algo=keccak  -i0 -q -m1 -d gtx780 -b16384 -C2 -lT2880x24 -L1024 -s3 -R2 -H 2

I am not entirely sure the the -b16384 play a role with i0 though (the -C2 is in principle useless)

newbie
Activity: 53
Merit: 0

Also have it scripted and all I have to do is enter the build date  Wink
newbie
Activity: 43
Merit: 0
Nice increase in the hashrate, the 780ti runs now around 315Mhash/s (before it was around 250~260Mhash/sec)

Do you overclock? Everytime I see your hashrates they're always a bit higher than mine. My 780ti is getting ~280Mhash/sec with:
--algo=keccak -d gtx780 -m 1 -L 1024 -l T2880x24
Card is stack EVGA SC clocks.
newbie
Activity: 53
Merit: 0
CudaMiner for OS X has been updated to run without requiring Xcode, MacPorts, and other setup steps. 

For binaries and instructions see http://www.johnchapman.net/cudaminer/

Prerequisites

CudaMiner for OS X previously required installing Xcode and MacPorts, in addition to a number of configuration steps.  These are no longer required.  Grin

The 2014-02-28 release for OS X is available here: http://www.johnchapman.net/crypto-currency/cudaminer-2-28-2014-release-now-available-for-os-x-10-7-10-8-and-10-9/
legendary
Activity: 1400
Merit: 1050
I am almost at 5 helix lol
For the new version what kind of config should we used ? same as the previous version except T instead of K ?

and lower the x32 to x24. yes.

Nice increase in the hashrate, the 780ti runs now around 315Mhash/s (before it was around 250~260Mhash/sec)
hero member
Activity: 756
Merit: 502

Thanks , I was lucking brain features , cause i was using the 18.02 release. now i get 167mhs each.  Grin Grin Grin Grin Grin
You rock man!!!

I had the lucky idea to check the 64 bit rotate performance, and Alexey @nVidia did it this afternoon - and this is the result. Turns out the compiler was originally creating a complex instruction sequence, which Alexey was able to replace with two funnel shifts only.

Christian

hero member
Activity: 676
Merit: 500

What setting do you use for 570 ti . i can`t get more than 76000?

you're lacking a hardware feature that enables this speed-up: the funnel shifter.
The keccak improvements target Compute 3.5 devices or better.

unless you meant to say 750 Ti

Code:
--algo=keccak -d gtx780,gt640,gtx750ti,gtx750ti#2 -i 0 -m 1 -l T2304x24,T768x24,T1024x24,T1024x24 -L 1024 -o stratum+tcp://maxpool.1gh.com:17333 -u mKbA9HoN13R6MYzhfuRCAeAyVAsqb7TiDy -p x

use this line to mine right into my wallet ;-)

Christian

Thanks , I was lacking brain features , cause i was using the 18.02 release. now i get 167mhs each.  Grin Grin Grin Grin Grin
You rock man!!!
hero member
Activity: 756
Merit: 502
I am almost at 5 helix lol
For the new version what kind of config should we used ? same as the previous version except T instead of K ?

and lower the x32 to x24. yes.
legendary
Activity: 1400
Merit: 1050
I am almost at 5 helix lol
For the new version what kind of config should we used ? same as the previous version except T instead of K ?
hero member
Activity: 756
Merit: 502

What setting do you use for 570 ti . i can`t get more than 76000?

you're lacking a hardware feature that enables this speed-up: the funnel shifter.
The keccak improvements target Compute 3.5 devices or better.

unless you meant to say 750 Ti

Code:
cudaminer --algo=keccak -d gtx780,gt640,gtx750ti,gtx750ti#2 -i 0 -m 1 -l T2304x24,T768x24,T1024x24,T1024x24 -L 1024 -o stratum+tcp://maxpool.1gh.com:17333 -u mKbA9HoN13R6MYzhfuRCAeAyVAsqb7TiDy -p x

[2014-02-28 23:43:01] accepted: 55/69 (79.71%), 555165 khash/s (yay!!!)
[2014-02-28 23:43:11] GPU #3: GeForce GTX 750 Ti, 147494 khash/s
[2014-02-28 23:43:38] GPU #2: GeForce GTX 750 Ti, 161607 khash/s
[2014-02-28 23:43:38] GPU #0: GeForce GTX 780, 208707 khash/s
[2014-02-28 23:43:39] GPU #1: GeForce GT 640, 38774 khash/s

use this line to mine MAX right into my wallet ;-)

Christian
hero member
Activity: 676
Merit: 500

MINE MAX, HELIX WITH THE 2014-02-28 release. Also use this for improved Yacoin performance.

For keccak, use the T kernel now for Compute 3.5 devices. Careful, its warp limit is 24 not 32.

My GTX 750 Ti makes 164 MHash/s   << this is the new keccak ACE CARD.
My GTX 780 now breaks 200 MHash/s at 80% TDP.  YEAH BABY!

Thanks to Alexey for the submission, even though he bungled up the ROL macro at the first
attempt, which is why we didn't make the Helixcoin launch Wink

Christian

What setting do you use for 750  ti . i can`t get more than 76000?
sr. member
Activity: 350
Merit: 250
The coin was a giant flop
full member
Activity: 182
Merit: 100

So who became rich today, lol? Wink

Christian

Not yet, I have lowered it alot and MIGHT get almost 2BTC out of this, I am fine with that.
Jump to: