Topic: Phoenix - Efficient, fast, modular miner - page 18. (Read 760839 times)

No change

That fixed it for me. Thanks.

Are you sure that you were actually getting that speed?  Meaning, were you producing more stales?  Phoenix was rumored to be showing artificially high hash rates, although I cannot confirm this personally.

EDIT:  Add FASTLOOP=false

Fastloop is on by default and while it is not supposed to affect higher aggression levels like it used to, it is better safe than sorry.

+1 same here in Ubuntu 11.04.
It very strange that while running smartcoin with phoenix as miner gives 427+ Mhash/s , but when run standalone phoenix gives only 404 Mhash/s

WinXp x64
HD 6970

Was getting 420 Mh/s, but with the new version, I can't get over 405 Mh/s.

command line used:
WORKSIZE=128 BFI_INT AGGRESSION=11 VECTORS

What do I need to change to get back the missing speed?

finally the idle bug seems to be gone for good. not sure wha it was in the end.  =)

v1.50

changes:
windows update (xp);
installation of some modules and dlls to get phoenix running from source (though now it is not)
- pyopencl
- boost
- twisted
- zope
- to system 32
  - general dlls (from vc 9)
    - msvcp90.dll
    - msvcr90.dll
  - a bunch of 24 boost dlls

?

---------edit----------------------------
once again that was a false alarm. will not says anything about this any more. got one more idle stall. the last always happened on deepbit.

Have you considered branching your releases?  I, for one am always hesitant to run live with HEAD (or TIP, or whatever people refer to it as) code.

Mathematically, it can only be about 3% [although ever so slight variation due to quantum effects, meaning you submit single shares, not partial shares, so there is variation in any given analysis window and also a change in the amount of work dumped or lost as stale when a block changes (the faster the miner, the less stale in a perfect world I think)].

The reason that it has to be about 3% is it was a rearrangement such that fewer computations on the GPU needed to be performed for a given unit of work ... so the performance increase is actually fixed.

This has been on the SVN since Monday, and since kernels are not part of the compiled binary I'm not planning another release until I finish the RPC re-write.

Make sure you apply this fix if you haven't already:

http://forum.bitcoin.org/index.php?topic=23067.0

6% performance increase for me

no success here...

the error is "Could not locate the specified kernel!"

I've copied and pasted the kernels files in the same root folder, the kernels folders in the root folder, but no success...

Phoenix goes a lot faster when you use this fix.

http://forum.bitcoin.org/index.php?topic=23067.0

HD 6950 @ 890 || 360MH/s

Already been added to the latest poclbm.

And thanks OP works cherry in Win7 is PnP.

Change to the directory where phoenix.py is located and then execute.   It uses the current working directory as reference to find the kernels and should probably look in the directory where phoenix.py resides.  Since most people use a script to launch their console miners anyway, it is not really an issue to just add the directory change into it.  On Linux, many people launch it and use screen to detach it from the console allowing them to remotely access it at any time [via a shell, such as SSH or telnet into the machine]. 

FWIW, I have found Linux performed far worse than with Windows.  I used the latest Ubuntu at the time [3 weeks ago perhaps].  Finally put my old Vista Ultimate x64 on there [OEM, registered on a computer I built but died a terrible death ... some short between the board, case and power supply and lost a second power supply to it when I replaced it not thinking that I should have checked for a short first], and it activated fine.  I could have called Microsoft and I am sure they would have allowed me to activate it considering the original computer was recycled, but I didn't have too.  I think the difference for me was a solid 20%.  Modifying clock, fan, and other settings is mildly annoying as well since you have to restart X each time [well, at least for the clock settings].  I have read about a lot of posts that show people getting great hash rates on their Linux boxes though, roughly equal to Windows x64 (not sure about x86), so either there was more I could have done, or my hardware config was not optimal under Linux [auto-configure with Ubuntu perhaps did me an injustice that Windows did apparently correctly].

Linux -vs- Windows hash rate performance?  My experience says Windows, but by posts that I have read, I would say about equal and if what a few said are true, maybe a little better on Linux.  Tough to beat the cost of Linux through [I have been an avid Linux and FreeBSD user since Dec 1996 with the last 6 months being the first time I do not run a server based on one of these two in the home ... although my phone and my Tivo run Linux ].

I believe that I saw you were using AGGRESSION=11, correct?  Phoenix 1.50 (maybe a little bit earlier even) modified things enough that FASTLOOP essentially has no impact beyond a certain aggression; I believe > 9.  I do know that FASTLOOP is now on by default, so passing FASTLOOP or FASTLOOP=yes is redundant (I still have FASTLOOP in my options for my workstation).  So, that would explain why you didn't notice a difference.  If you are using the machine as dual use, I would recommend a maximum of AGGRESSION=9 and leave FASTLOOP on.  You can always restart the miner to the more aggressive settings when you aren't using it.  I use this for my 5850s and probably should have this as an option for my workstation while running overnight.


Sorry for quoting all that, but I think it is needed in context (the initial question with options for instance) and since the model here is "forum in a single thread", it isn't easy to view the topic in a threaded manner.

I had a computer shop build me a cheap $600 build around a ATI 5830 and just installed Ubuntu via USB, this computer does not have internet so everything is going on through USB.

If someone can be there as a sounding board for my issues I'd really appreciate it and I''ll give everything I have generated in Bitcoinplus.com over the last week. BTC0.01.

I put the file bitcoin-0.3.23-linux.tar.gz onto the USB, and see it in ubuntu when I plug it in. What are the next steps to take? Double clicking and Extracting are not getting it running.  Is it because I'm using the Natty version of Ubuntu and they aren't compatible? I may miss your response so if you try to help please PM me with a link to your post and if it gets me there I'll send it over.

just installed phoenix on ubuntu 11, but when I try to run, I get something like "cant find kernel" or something like that... what is wrong?


same thing with poclbm kernel...

and the files seems to be correct...

pyrhon.py
kernels/phatk/...
kernels/poclbm/...

so do you find your hashing rate differences negligible between linux and win7?

thanks everyone for your help my cards are now performing the same as my win 7 box woot

ill test fastloop now; i couldn't see any real difference with it on or off

anyway thanks again

You shouldn't at low aggression levels (9 or below).  It will drop while you are using the machine however [which is what it is for, to help reduce lag].  FASTLOOP is neutral at best and detrimental in practice at aggression levels higher than 9 (older versions of Phoenix it was patently obvious).

In fact, I just shut my miner off, set FASTLOOP=false and restarted it and my hash rate dropped about 5MH/s (it varies a little obviously).  I turned it back on and the rate is back where it should be.  Here is what I use with my 6970 930MHz core clock with all else at stock.  This machine is my Windows 7 x64 Pro with Catalyst 11.6 drivers installed.  I use this machine all day for work (VPN with Remote Desktop) and run many services for development purposes as well as iTunes going and Aero on.

any ideas to reach 300+Mhash/sec on a 6870 using Phoenix 1.5?

running stock voltages and clock speeds I'm only getting 265Mhash/sec
any way to get that up to 300 w/o overclocking? (don't want to shorten the life of the card)
if not, then what should i overclock/underclock to hit 300Mhash/sec?

rig specs:
Ubuntu 11.04
2x HIS PCI-Express ATI Radeon HD6870 Video Card (Engine Clock: 900 MHz; Video Memory: 1GB DDR5; Memory Clock: 4.2 GHz; RAMDAC: 400 MHz)
AMD Sempron 140, Socket AM3, 2.7 GHz, 1 MB Cache, 45 Watts
ASRock AM3 processors AMD 770-140W 4DDR3/ATI CrossFireX motherboard
Seagate Barracuda 7200.12 500 GB 7200RPM SATA 6Gb/s with NCQ 16MB Cache
Kingston ValueRAM 2 GB 1333MHz PC3-10600 DDR3 DIMM Desktop Memory
Thermaltake V3 Black Edition SECC / Plastic ATX Mid Tower Computer Case
Antec EA-650 Green ATX Energy Star Certified Power Supply


running:
python phoenix.py -u http://user:pass@server:8832 -k phatk VECTORS BFI_INT AGGRESSION=12 DEVICE=0

thanks in advanced!