Phoenix - Efficient, fast, modular miner - page 55.

xlcus

legendary

Activity: 966

Merit: 1009

Getting quite a lot of these errors...

Code:

[25/04/2011 15:57:05] Disconnected from server
[RPC (+LP)]Unhandled error in Deferred:
Traceback (most recent call last):
  File "/usr/lib/python2.6/dist-packages/twisted/internet/defer.py", line 354, in _startRunCallbacks
   self._runCallbacks()
  File "/usr/lib/python2.6/dist-packages/twisted/internet/defer.py", line 371, in _runCallbacks
   self.result = callback(self.result, *args, **kw)
  File "/usr/lib/python2.6/dist-packages/twisted/internet/defer.py", line 330, in _continue
   self.unpause()
  File "/usr/lib/python2.6/dist-packages/twisted/internet/defer.py", line 325, in unpause
   self._runCallbacks()
--- ---
  File "/usr/lib/python2.6/dist-packages/twisted/internet/defer.py", line 371, in _runCallbacks
   self.result = callback(self.result, *args, **kw)
  File "/home/xlcus/bitcoin/phoenix-1.1/minerutil/RPCProtocol.py", line 244, in
   d.addErrback(lambda x: self._failure())
  File "/home/xlcus/bitcoin/phoenix-1.1/minerutil/RPCProtocol.py", line 307, in _failure
   self._setLongPollingPath(None)
  File "/home/xlcus/bitcoin/phoenix-1.1/minerutil/RPCProtocol.py", line 174, in _setLongPollingPath
   self.activeLongPoll.cancel()
exceptions.AttributeE

I'm running the linux version on Ubuntu, with a 5750, connected to deepbit.

bolapara

member

Activity: 78

Merit: 10

Updating with new results...

*985/900*
poclbm with f=1, ws=128: 385MH/s
phoenix 1.0 with agg=12: 377MH/s
phoenix 1.01 with agg=13, ws=128: 380MH/s

*985/300*
poclbm with f=1, ws=128: 327MH/s
phoenix 1.01 with agg=13 ws=128: 386MH/s

*1010/300*
phoenix 1.01 with agg=13, ws=128, BFI_INT: 437MH/s (!!!)

ubuntu 10.10, 5870, 2.1 sdk, vectors

Since I was able to bring down temps by going to 300MHz memory clock, I was able to bump up the GPU clock to 1010MHz and with BFI_INT I am getting 437MH/s. It's run for about 7-8 hours so far without issues.

allinvain

legendary

Activity: 3080

Merit: 1080

Some numbers for all to meditate upon:

Radeon 5970
Core clock 850 Mhz
Memory Clock 300 MHz

Aggression level 7 with the BFI_INT turned on

Getting 346 Mhash/sec for each GPU core!! Pretty nice improvement over the 317 I used to get with poclbm!

So overall I'm happy

Maybe it is just me but I get the impression that the GPUs aren't "working as hard" with this miner as the temps are a bit lower than with poclbm yet I get higher performance.

So far it has been 5 min running this miner and the hashing rate is averaging around 347 Mhash/sec.

grndzero

sr. member

Activity: 392

Merit: 250

Quote from: jkminkov on April 25, 2011, 08:36:54 AM

Quote from: grndzero on April 25, 2011, 08:26:58 AM

Any idea whether this would lighten the work loads on pool servers too?

I see increased number of rejected units, probably because current work is not being dropped but reported to server and rejected, happens right after long pool message

I'm not seeing any increase since I've been running it. Perhaps it is work that is submitted right before LP is checked so it's rejected by the server because of new data?

jkminkov

hero member

Activity: 698

Merit: 500

Quote from: grndzero on April 25, 2011, 08:26:58 AM

Any idea whether this would lighten the work loads on pool servers too?

I see increased number of rejected units, probably because current work is not being dropped but reported to server and rejected, happens right after long pool message

Cheeseman

newbie

Activity: 23

Merit: 0

I'm seeing around a 10% increase in speeds on my dual 5870 rig.

965/300 5870 on poclbm - 360 mhash
965/300 5870 on Phoenix - 390 mhash

910/300 5870 on poclbm - 330 mhash
910/300 5870 on Phoenix - 365 mhash

Working on Deepbit with long polling. Much lower temps and power consumption as well.

Thanks jedi, great work.

grndzero

sr. member

Activity: 392

Merit: 250

Quote from: jedi95 on April 25, 2011, 06:44:01 AM

Quote from: grndzero on April 25, 2011, 05:26:58 AM

I had to escape the ;askrate on linux or else OpenCL would say it couldn't find the device.
phoenix -u http://USERNAME.WORKERNAME:[email protected]:8332/\;askrate=12 -k poclbm DEVICE=2 VECTORS AGGRESSION=12

Standard poclbm/poclbm-mod were giving me 317Mh/s.
Phoenix was giving me 310Mh/s.
That went up to 343.5Mh/s with BFI_INT!

If you're using askrate make sure to set it properly. Calculate (type into Google) 2^32 divided by your hashrate. So mine would be 2^32/343500000=12.5035438 (so 12)

Nice to see that BFI_INT is working for you!

As for askrate, it's not necessary with Phoenix because it maintains a work queue and only requests work when needed. Phoenix also ignores the askrate setting automatically for RPC servers with long polling support.

Any numbers on efficiency vs poclbm and -mod? Any idea whether this would lighten the work loads on pool servers too?

Also, I know BFI_INT most likey won't work on CPU's, but have you explored using any other of these ideas for CPU mining since there are probably still quite a few CPU miners out there (no matter how futile).

keybaud

full member

Activity: 120

Merit: 100

My HD5870 at 950/300 was originally hashing on my Win7 machine at 355 MH/s, but with Phoenix I'm now getting a reported 394 MH/s.

kindle

member

Activity: 84

Merit: 10

I am running 9 now..seems to max out the core at 99% and at the same time the fastest hash rate

xlcus

legendary

Activity: 966

Merit: 1009

In fact Aggression=7 seems to be the sweet spot for me...

144.98 Mhash/sec

xlcus

legendary

Activity: 966

Merit: 1009

Quote from: kindle on April 25, 2011, 07:18:46 AM

So far my 5850 and 5970 works well with aggression=8 when its set at 12 for my pure mining rig, the hash rate reports very very low hash rates. Is it cuz the miner is unable to report the true hashrates?

Seing a similar thing here on my 5750...

Aggression=12 : 141.75 Mhash/sec
Aggression=8 : 143.96 Mhash/sec

jedi95

full member

Activity: 219

Merit: 120

Quote from: marcus_of_augustus on April 25, 2011, 07:20:42 AM

Is it really dependent upon the Twisted package?

It is not installed on my system but phoenix appears to be running fine.

I don't see how that's possible since we make extensive use of Twisted in Phoenix. The entire miner is based around a Twisted reactor. Without it you won't be able to send or receive work, much less actually mine.

marcus_of_augustus

legendary

Activity: 3920

Merit: 2349

Eadem mutata resurgo

Is it really dependent upon the Twisted package?

~~It is not installed on my system but phoenix appears to be running fine.~~

Edit: sorry it is there, listed as package "python-twisted" on ubuntu 10.10 (also there are other "python-twisted-*" packages in default installation twisted stack)

kindle

member

Activity: 84

Merit: 10

Quote from: jedi95 on April 25, 2011, 07:04:24 AM

Quote from: kindle on April 25, 2011, 06:57:50 AM

jedi95: Thanks for releasing this miner freely, so far the hash rates are too good to believe but I am running none the less to see if it works. I have a question, will the (H==0) mod for poclbm work here?

Already applied to the kernel Wink

Thanks for the clarification. So far my 5850 and 5970 works well with aggression=8 when its set at 12 for my pure mining rig, the hash rate reports very very low hash rates. Is it cuz the miner is unable to report the true hashrates? Also on a 6990 there is no speed gain, with BFI_INT the hash rate drops.

CFSworks

member

Activity: 63

Merit: 10

Quote from: yomi on April 25, 2011, 06:53:48 AM

Anything I might have missed?

Thanks!

I must have had twice your headache setting up the ATI Stream SDK on Linux for the first time. I'm guessing you downloaded it as a single archive from ATI's site?
Try setting the ATISTREAMSDKROOT environment variable to point to the directory where you extracted the SDK, since those compilation errors are most likely the result of the Stream compiler not being able to find the Stream libraries to link against.

Also, this document was extremely useful in getting the SDK set up. I wish I had found that before I started my install, it would have saved me a lot of trouble. Wink

jedi95

full member

Activity: 219

Merit: 120

Quote from: kindle on April 25, 2011, 06:57:50 AM

jedi95: Thanks for releasing this miner freely, so far the hash rates are too good to believe but I am running none the less to see if it works. I have a question, will the (H==0) mod for poclbm work here?

Already applied to the kernel Wink

Quote from: marcus_of_augustus on April 25, 2011, 06:57:58 AM

Can confirm following hashrates;

linux ubuntu 10.10, SDK 2.1, fglrx 11.2, HD 5970, phoenix.py kernel=poclbm vectors aggression=12, worksize=128, gpu-clock 725, mem. clock 300 : 285 Mhash/s (same as poclbm but uses less power ~20 watts less)

and then this bolter ...
BFI_INT enabled 312 mhash/s ! (what exactly does this flag do ? suspect it is something to do with optimising integer arithmetic ALU)

Looking good, this is just great if it works long term after some testing, can now run my rigs cooler and quieter with more speed, where's that donation address i want to throw some bitcoins at you guys!

Down clocking memory on linux is big kudos and BFI_INT is off the charts ... unfortunately once this gets widespread then difficulty is going to take another hike ~10-15% I estimate.

BFI_INT is a hardware instruction that implements the SHA256 Ch function in a single instruction instead of 3. This is why you get a performance gain.

Donation address is:
1PHoenix9j9J3M6v3VQYWeXrHPPjf7y3rU

kindle

member

Activity: 84

Merit: 10

Ops i realised its already using the (H==0) mod

marcus_of_augustus

legendary

Activity: 3920

Merit: 2349

Eadem mutata resurgo

Can confirm following hashrates;

linux ubuntu 10.10, SDK 2.1, fglrx 11.2, HD 5970, phoenix.py kernel=poclbm vectors aggression=12, worksize=128, gpu-clock 725, mem. clock 300 : 285 Mhash/s (same as poclbm but uses less power ~20 watts less)

and then this bolter ...
BFI_INT enabled 312 mhash/s ! (what exactly does this flag do ? suspect it is something to do with optimising integer arithmetic ALU)

Looking good, this is just great if it works long term after some testing, can now run my rigs cooler and quieter with more speed, where's that donation address i want to throw some bitcoins at you guys!

Down clocking memory on linux is big kudos and BFI_INT is off the charts ... unfortunately once this gets widespread then difficulty is going to take another hike ~10-15% I estimate.

kindle

member

Activity: 84

Merit: 10

jedi95: Thanks for releasing this miner freely, so far the hash rates are too good to believe but I am running none the less to see if it works. I have a question, will the (H==0) mod for poclbm work here?

yomi

newbie

Activity: 39

Merit: 0

I am getting a 'Compilation Failed' message when I try to run it:

Code:

root@apretao:~/src/phoenix-1.1# ./phoenix.py -u http://u:[email protected]:8332/
No device specified or device not found, use DEVICE=ID to specify one of the following

[0] AMD Sempron(tm) 140 Processor
[1] Cypress
root@apretao:~/src/phoenix-1.1# ./phoenix.py -u http://u:[email protected]:8332/ DEVICE=1
Traceback (most recent call last):
File "./phoenix.py", line 122, in
miner.start(options)
File "/home/edu/src/phoenix-1.1/Miner.py", line 77, in start
self.kernel = self.options.makeKernel(KernelInterface(self))
File "./phoenix.py", line 111, in makeKernel
self.kernel = kernelModule.MiningKernel(requester)
File "kernels/poclbm/__init__.py", line 203, in __init__
self.loadKernel(self.device)
File "kernels/poclbm/__init__.py", line 279, in loadKernel
self.context, kernel).build(self.defines)
File "/usr/local/lib/python2.6/dist-packages/pyopencl-0.92-py2.6-linux-x86_64.egg/pyopencl/__init__.py", line 138, in program_build
"Build on %s:\n\n%s" % (dev, log) for dev, log in build_logs))
pyopencl.RuntimeError: clBuildProgram failed: build program failure

Build on :

Internal error: Compilation failed.

Before that, I also had to solve this:

Code:

Build on :

sh: /bin/x86_64/clc: not found

I did it by symlinking /usr/local/bin/x86_64/clc to /bin/x86_64/clc

Anything I might have missed?

Thanks!

Topic: Phoenix - Efficient, fast, modular miner - page 55. (Read 760771 times)