Main Menu

AMD OpenCL Momentum miner

Started by NaN, January 26, 2014, 08:41:40 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

NaN

Indeed, aticonfig does not work in multi GPU setups. The easiest thing for me was to write a cgminer config-file so that after booting up, cgminer is started, then closed and after that I execute my miner.

mbender71

I've switched over from Windows 8.1x64 to Ubuntu 13.04x64 and INDEED there is a significant
increase in speed. Wow.. thank you NaN! Now I'm delving into tweaking the GPU clock, memory,
fan, etc. I've read extensively about aticonfig and I can seem to get it to work on the first (default)
adapter, but every time I try to use the --adapter parameter it seems to only be setting/getting the
first (default) GPU. For example, aticonfig --adapter="all" --odgt  returns four values but they are
all the same. I'm suing kernel 3.8.0-35 and catalyst 13.12. Is this a good combo? Perhaps there
are other (GUI) utilities out there? Much appreciated!

Running 4 x MSI R9 290 Gaming

NaN

#297
Quote from: tazbox on February 12, 2014, 08:03:08 PM
Quote from: NaN on February 12, 2014, 04:12:28 PM
Quote from: iamcount on February 12, 2014, 03:56:58 PM
three his 280x ,win7 13.1dirver.just get 57xx, it's too bad. use -t 0,0,1,1,2,2 .
if add -a 1, the miner can not run. 
so, help me!
Which version do you use? I guess that you're using an old one. Use v0.1.4 or v0.2.2. If your monitor is attached to GPU 0, then I would use -t 0,1,1,2,2 and I would try -t 0,1,2 too. -a 1 should be fastest.

Hello NaN, I thought 0,1,2.... is the number of card ? because for 4 card i use 0,1,2,3.
What exactly is his number? eg for a single card I put t 0? it would change something to put 0,1,2 for a single card?
The numbers after -t are the DeviceIDs, which are shown during startup (DeviceID x). For each number in the list the miner starts one thread. Normally, the first GPU has DeviceID 0 so that -t 0 mines with one thread on the first card and -t 1 means that the miner uses one thread on the second card, etc. Thus, -t 2,2 means that the miner should launch two threads on the third GPU. It might increase your performance but at least the Windows driver of R9 290(X) seems to be broken and only the Linux driver functions with full speed. Just try whether  -t 0,1,1,2,2,3,3 or -t 0,1,2,3 give you better performance if you are using four GPUs.

tazbox

Quote from: NaN on February 12, 2014, 04:12:28 PM
Quote from: iamcount on February 12, 2014, 03:56:58 PM
three his 280x ,win7 13.1dirver.just get 57xx, it's too bad. use -t 0,0,1,1,2,2 .
if add -a 1, the miner can not run. 
so, help me!
Which version do you use? I guess that you're using an old one. Use v0.1.4 or v0.2.2. If your monitor is attached to GPU 0, then I would use -t 0,1,1,2,2 and I would try -t 0,1,2 too. -a 1 should be fastest.

Hello NaN, I thought 0,1,2.... is the number of card ? because for 4 card i use 0,1,2,3.
What exactly is his number? eg for a single card I put t 0? it would change something to put 0,1,2 for a single card?

lty616

Hi NaN, any update for CUDA miner so far?

NaN

Quote from: iamcount on February 12, 2014, 03:56:58 PM
three his 280x ,win7 13.1dirver.just get 57xx, it's too bad. use -t 0,0,1,1,2,2 .
if add -a 1, the miner can not run. 
so, help me!
Which version do you use? I guess that you're using an old one. Use v0.1.4 or v0.2.2. If your monitor is attached to GPU 0, then I would use -t 0,1,1,2,2 and I would try -t 0,1,2 too. -a 1 should be fastest.

Hukkel

Quote from: iamcount on February 12, 2014, 03:56:58 PM
three his 280x ,win7 13.1dirver.just get 57xx, it's too bad. use -t 0,0,1,1,2,2 .
if add -a 1, the miner can not run. 
so, help me!

Make in the pool 3 different workers and make 3 different .bat files. Each for one gfx card and one worker.

so -t 0,0 worker1 in the first, -t 1,1 in the second etc etc.

Make sure you turn crossfire off.

Go and have a look what each card produces.

Oh and use the latest clpts2.2 miner software.

iamcount

three his 280x ,win7 13.1dirver.just get 57xx, it's too bad. use -t 0,0,1,1,2,2 .
if add -a 1, the miner can not run. 
so, help me!

Coindgr

Anyone tried using a virtual machine to run linux in windows ? Any gain? Or the primary system needs to be Linux?

crazycoin

I am using stock Ubuntu 3.10 and Josh Patten Bootstrap mining.

dongboy

Quote from: NaN on February 11, 2014, 05:20:34 PM
Quote from: Aselert on February 11, 2014, 03:31:00 PM
So finally... no CUDA miner realized?  ::)
I'm working on Nvidia support but I do not have a fully functional version yet.
gtx580 not deviceid 0

plane501

Quote from: wsnnwa on February 12, 2014, 12:53:15 AM
Got Ubuntu installed on my second rig, this time I decided to try out some OC. I used one 290X for this test, clocked at 1150/1400. Only Running -t 0,0, no gains noticed from -a.

[STATS] 2014-Feb-11 19:49:25 | 4582.29 c/m | 17.71 sh/m | VL: 62 (100.00%), RJ: 0 (0.00%), ST: 0 (0.00%)

would you please tell me how to OC in Ubuntu? I'm new to linux :)

wsnnwa

#287
Quote from: mbender71 on February 12, 2014, 02:18:29 AM
Seriously??? I know the 290X is a wee bit faster than the 290 (more shaders) but I've over clocked
my 4 x 290's to 1125/1350 and running -t 0,1,2,3,4 -a 1 and only getting 12300 c/m and 46 sh/m.
Is the 290x really THAT much faster or are we seeing a big difference between using clpts on linux
versus windows? If it weren't for the over clocking ease of windows I would definitely be on linux!!

Quote from: wsnnwa on February 12, 2014, 12:53:15 AM
Got Ubuntu installed on my second rig, this time I decided to try out some OC. I used one 290X for this test, clocked at 1150/1400. Only Running -t 0,0, no gains noticed from -a.

[STATS] 2014-Feb-11 19:49:25 | 4582.29 c/m | 17.71 sh/m | VL: 62 (100.00%), RJ: 0 (0.00%), ST: 0 (0.00%)

Yes seriously. clpts runs better on Linux. I was only getting about 3000cpm per 290X on Windows, around 3200 if I overclocked.

I just created some shell scripts, one sets both GPUs to 1150/1400 and the other sets my fans to 100%. I have one that also prints the GPU temp, but I don't know how accurate that is lol.

mbender71

#286
Seriously??? I know the 290X is a wee bit faster than the 290 (more shaders) but I've over clocked
my 4 x 290's to 1125/1350 and running -t 0,1,2,3,4 -a 1 and only getting 12300 c/m and 46 sh/m.
Is the 290x really THAT much faster or are we seeing a big difference between using clpts on linux
versus windows? If it weren't for the over clocking ease of windows I would definitely be on linux!!

Quote from: wsnnwa on February 12, 2014, 12:53:15 AM
Got Ubuntu installed on my second rig, this time I decided to try out some OC. I used one 290X for this test, clocked at 1150/1400. Only Running -t 0,0, no gains noticed from -a.

[STATS] 2014-Feb-11 19:49:25 | 4582.29 c/m | 17.71 sh/m | VL: 62 (100.00%), RJ: 0 (0.00%), ST: 0 (0.00%)

shadowK

#285
Hi NaN. I have 4*R9 290x running on Xubuntu 64 and they are round 16000 cpm for two days. Today I replaced my old motherboard and restart the miner. It hits 16000 cpm than decreases gradually about 100 cpm per sec. After it hit around 8000 cpm, the computer freezes. I restart the computer and try to run the miner. The miner isn't be able to start anymore. Do you have any idea about it? My catalyst is the latest version from amd.

I also tried the old version and it didn't work too.

EDIT: The error message is "error in xptMiner/clInit.cpp, line 96: Method clGetDeviceIDs(...) failed (errorcode -1)"