RPC Miners (CPU/4way/CUDA/OpenCL)

[walidzohair]

Yep, -threads is the CPU only option. Read readme.txt file.

So it is normal that (rpcminer-cuda) makes heavy use of CPU just as the (rpcminer-cpu) ?

Also about the Khashes .. I am not getting any increase in performace (always at 1100-2200) is that normal or do I have a weak card ?

[LZ]

You may try to experiment with the -aggression option.

Code:

rpcminer-cuda -aggression=5 -url=http://mining.bitcoin.cz:8332 -user=login.suffix -password=password

[dishwara]

If u have to use CUDA then better use poclbm_py2exe_20110104

Coz when ever i use rpcminer to run cuda then it takes 95% of GPU & also slows down pc,
While poclbm.exe -d0 uses cuda & 95% GPU, but no slow down of pc
Windows 7, 9400GT with 260.99 & also now the new 266.58 driver.

[puddinpop]

I've updated the first post with a new release.  The CUDA miner is now using the driver API instead of the runtime API, and there's no more excess CPU usage when CUDA is running.  Also the CUDA and OpenCL kernels were optimized a bit to improve hashing speed.

[abitcoinuser]

The most recent build from 1/29/2011 gives me 0 khash/s on both the cpu and cuda miners with options that were successful for the last build.  Any ideas?

[puddinpop]

The most recent build from 1/29/2011 gives me 0 khash/s on both the cpu and cuda miners with options that were successful for the last build.  Any ideas?

I just double checked the cpu and CUDA miners in the archive, and they are working fine here.  If you are indeed passing the same parameters, your issue is most probably caused by not being able to connect to the RPC server.  It could also be a corrupt download, or incomplete unzip.

[LZ]

Maybe a firewall?

[BitCoinPurse]

Same here... New version just reports 0 khash...  Firewall off.

C:\Users\Ghandi Gupta\Desktop\Miner>rpcminer-opencl.exe -url=http://mining.bitcoin.cz:8332 -user=xxxxx -password=xxxxxx
Client will start 1 miner threads
Work will be refreshed every 4000 ms
1 OpenCL platforms found
1 OpenCL GPU devices found on platform 0
Setting OpenCL device to first device found
Available extensions : cl_khr_global_int32_base_atomics cl_khr_global_int32_exte
nded_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics
 cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_amd_d
evice_attribute_query cl_amd_printf cl_amd_media_ops cl_amd_popcnt cl_khr_d3d10_
sharing
Defining AMDMEDIAOPS
Create context rval=0
Create command queue rval=0
Creating program with source
Building program with options -D AMDMEDIAOPS
0 khash/s
0 khash/s

[puddinpop]

Same here... New version just reports 0 khash...  Firewall off.

C:\Users\Ghandi Gupta\Desktop\Miner>rpcminer-opencl.exe -url=http://mining.bitcoin.cz:8332 -user=xxxxx -password=xxxxxx
Client will start 1 miner threads
Work will be refreshed every 4000 ms
1 OpenCL platforms found
1 OpenCL GPU devices found on platform 0
Setting OpenCL device to first device found
Available extensions : cl_khr_global_int32_base_atomics cl_khr_global_int32_exte
nded_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics
 cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_amd_d
evice_attribute_query cl_amd_printf cl_amd_media_ops cl_amd_popcnt cl_khr_d3d10_
sharing
Defining AMDMEDIAOPS
Create context rval=0
Create command queue rval=0
Creating program with source
Building program with options -D AMDMEDIAOPS
0 khash/s
0 khash/s

I see at least 2 issues there.  The first is that you don't have a "Target = " message, which means you have not successfully retrieved work from the RPC server.  The second is that you don't have a "Build program rval=" message, which means the kernel is still being built.

Did the previous release of the OpenCL miner work on the same machine, and does the current CPU miner work on it?  Did you extract all the files in the archive to a new directory, not just over the old ones?  What graphics card do you have?

[BitCoinPurse]

I see at least 2 issues there.  The first is that you don't have a "Target = " message, which means you have not successfully retrieved work from the RPC server.  The second is that you don't have a "Build program rval=" message, which means the kernel is still being built.

Did the previous release of the OpenCL miner work on the same machine, and does the current CPU miner work on it?  Did you extract all the files in the archive to a new directory, not just over the old ones?  What graphics card do you have?

I did get the "Target =" message, it just didn't get in the paste.

Yes, the old OpenCL miner worked flawless, and the current CPU miner works too.  And yes, new directory for the new miner clients...

I have an ATI Radeon HD 5500 Series (Redwood) with Catalyst 10.11 drivers and ATI Stream 2.3.

My system is a W7 box with a AMD Phenom(tm) II X2 550 Processor


I doesn't seem to be "building the kernel"...

[puddinpop]

You can use the .cl file from the last release.  If you are so inclined, you can edit the current .cl file and see which changes are keeping the kernel from building.  There weren't too many changes, so it should be obvious what the culprit is.

[LZ]

And if you already removed the previous version, you may get it here: link.

[lior]

hi,

is it possible to use Tor with the miner

and if so can you give me hint?

[slush]

is it possible to use Tor with the miner

It is possible, but the network latency is big problem then. It will slow down your miner a lot.

[suntiger]

Is there any particular reason that the CUDA miner would only perform at maximum khash/s if its CLI status window were the foreground window? I'm seeing consistent drops from 16000khash/s to 6000-9000 khash/s when the window is not in the foreground, even if absolutely nothing is happening in whatever window is the foreground (even away from the computer).

System: Windows 7 Enterprise 32 bit, 2.9ghz Core2Duo, Geforce 9600GT, 3gb RAM.

Is there any way around this behavior? It poses problems for the time when I'm not actively using the computer, as if anything at all pops up in, say, the first hour of a six hour run, I lose quite a lot of potential khashes.

Also, is 16mhash about right for a 9600gt? I tried playing with the aggression parameter, but it only lagged the rest of the system more, it didn't increase max khash/s while foregrounded. Anything else I can do to bump performance up?

[swaits]

Is there any particular reason that the CUDA miner would only perform at maximum khash/s if its CLI status window were the foreground window? I'm seeing consistent drops from 16000khash/s to 6000-9000 khash/s when the window is not in the foreground, even if absolutely nothing is happening in whatever window is the foreground (even away from the computer).

This seems to be a new bug in the 1/28 binaries. I'm seeing it too. The previous version did not exhibit this behavior.

Currently, if my RPC Miner is in the foreground, it does its usual ~39M. As soon as I minimize, or even just click away, it drops straight down to ~4M.

I'm on Windows 7 (64 bit), GeForce GTX 260, 12GB of system memory, i7-920 CPU.

I'm running the CUDA miner, with no special parameters. But, just in case it matters, this is how I launch the software (via batch file):

Code:

@echo off

echo Launching Bitcoin. Please wait ...

rem start bitcoin
cd "C:\Program Files (x86)\Bitcoin"
start /min /low bitcoin.exe -gen -server

rem wait ten seconds
ping 127.0.0.1 -n 10 -w 1000 > nul

rem start rpc miner
cd "C:\Program Files (x86)\Bitcoin\rpcminer"
start "RPC Miner" /min /low rpcminer-cuda.exe -gpu -url=http://127.0.0.1:8332/ -user=XXX -password=XXX

I suspect it has to do with the lower CPU utilization fix. Which, by the way, is working very well. But perhaps this is why???

Thanks,
Steve

[LZ]

I can not reproduce this bug.

[swaits]

I can not reproduce this bug.

Bummer. Ok, I'll go back to the prior version for now.

Perhaps we could take a look at the diffs between the versions? Is the source in a public repo?

[LZ]

you may get it here: link.

[puddinpop]

I doesn't seem to be "building the kernel"...

Did you get it to work with the older kernel?

Is there any particular reason that the CUDA miner would only perform at maximum khash/s if its CLI status window were the foreground window?

No.  The hash rate is constant here whether the window is in the foreground or background.  You can try different graphics drivers (newer AND older) to see if that helps.  It could also be something with Windows 7 interfering with the priority of background apps.