Just tried SDK 2.4 on my 399.4 MH/s gpu.  It went down to 393.7 MH/s (actually 394.3 MH/s but occasionally dropping off to 390-391, I took an average).

Edit: I compared the new kernel with the old one using SDK 2.4 and the improvement was 3.1 MH/s (+0.79%).  This is a nice improvement but not enough to make me move away from SDK 2.1.  Also, SDK 2.4 causes the MH/s to drop suddenly by 3 or 4 MH/s every so often (variance is within 0.5 MH/s with SDK 2.1 for me) and even the peak values I achieve with SDK 2.4, new kernel or not, are below my SDK 2.1 average.

Ah well, good work though.  I sent you some BTC anyway simply because you tried to help me fix my problem.

I'm using SDK 2.1 because when I compared it with 2.4 I found a slight speed improvement with both phatk and poclbm (much larger with poclbm).

I haven't tried SDK 2.4 since applying the MA patch so that is maybe worth a try.

BFI_INT is definitely being used.  If I restart the command without BFI_INT on my 399.4 MH/s gpu I get 354.5 MH/s instead.

No change for me at all unfortunately, not even 0.1 MH/s.  I copied the file in place of phoenix-1.50/kernels/phatk/kernel.cl and restarted my instances of phoenix.

Settings:
Linux
Catalyst 11.6
SDK 2.1
phatk (bundled with phoenix 1.50 with MA tweak)
VECTORS BFI_INT FASTLOOP=false AGGRESSION=12 WORKSIZE=256
Solo mining

5850 Xtreme @ 1.0875V - 970MHz (399.4 MH/s) (75*C)
5850 Xtreme @ 1.1625V - 1050MHz (432.8 MH/s) (60*C)

Thanks for the tip, but after coming this far I think I'm pretty comfortable with writing my own scripts.  I might look at the code for ideas.

A nice idea.

I took what you did and extended it for my own purposes.  I'm not at all proficient with shell scripting and have just hacked this together - any criticism is welcome.

Caution: The constants at the beginning are appropriate for my miner and may not be appropriate for yours.  This script should not be run without being read and understood.

{
#!/bin/bash
# Scale the gpu clocks to keep temperature within preset bounds.

clock_step=10
sleep_time=60
card_0_min_temp=76
card_0_max_temp=79
card_0_min_clock=820
card_0_max_clock=970
card_1_min_temp=61
card_1_max_temp=64
card_1_min_clock=900
card_1_max_clock=1050

while :
do
   echo "------"
   echo
   card_0_clock=`aticonfig --adapter=0 --od-getclocks | grep Clocks | awk '{print $4}'`
   card_0_temp=`aticonfig --adapter=0 --od-gettemperature | grep -o [0-9][0-9].[0-9][0-9] | sed 's/\..*$//'`
   card_0_load=`aticonfig --adapter=0 --od-getclocks | grep -o [0-9]*% | sed 's/%//'`
   card_1_clock=`aticonfig --adapter=1 --od-getclocks | grep Clocks | awk '{print $4}'`
   card_1_temp=`aticonfig --adapter=1 --od-gettemperature | grep -o [0-9][0-9].[0-9][0-9] | sed 's/\..*$//'`
   card_1_load=`aticonfig --adapter=1 --od-getclocks | grep -o [0-9]*% | sed 's/%//'`
   echo "card_0_clock =" $card_0_clock
   echo "card_0_temp =" $card_0_temp
   echo "card_0_load =" $card_0_load
   echo "card_1_clock =" $card_1_clock
   echo "card_1_temp = " $card_1_temp
   echo "card_1_load =" $card_1_load
   echo

   if [ $card_0_temp -gt $card_0_max_temp ] && [ $(expr $card_0_clock - $clock_step) -ge $card_0_min_clock ]
      then echo "Card 0 is too hot; attempting to lower the clock speed."
      aticonfig --adapter=0 --od-setclocks=$(expr $card_0_clock - $clock_step),0
      echo
   fi
   if [ $card_0_load -gt 50 ] && [ $card_0_temp -lt $card_0_min_temp ] && [ $(expr $card_0_clock + $clock_step) -le $card_0_max_clock ]
      then echo "Card 0 is too cold; attempting to increase the clock speed by" $clock_step"MHz."
      new_clock=$(expr $card_0_clock + $clock_step)
      if [ $new_clock -lt $card_0_min_clock ]
         then new_clock=$card_0_min_clock
      fi
      aticonfig --adapter=0 --od-setclocks=$new_clock,0
      echo
   fi
   if [ $card_1_temp -gt $card_1_max_temp ] && [ $(expr $card_1_clock - $clock_step) -ge $card_1_min_clock ]
      then echo "Card 1 is too hot; attempting to lower the clock speed by" $clock_step"MHz."
      aticonfig --adapter=1 --od-setclocks=$(expr $card_1_clock - $clock_step),0
      echo
   fi
   if [ $card_1_load -gt 50 ] && [ $card_1_temp -lt $card_1_min_temp ] && [ $(expr $card_1_clock + $clock_step) -le $card_1_max_clock ]
      then echo "Card 1 is too cold; attempting to increase the clock speed by" $clock_step"MHz."
      new_clock=$(expr $card_1_clock + $clock_step)
      if [ $new_clock -lt $card_1_min_clock ]
         then new_clock=$card_1_min_clock
      fi
      aticonfig --adapter=1 --od-setclocks=$new_clock,0
      echo
   fi

   sleep $sleep_time
done
}

Yes, I'm worried about my VRM's.  I don't know how to find their temperature so they could be frying as we speak.

Honestly, I'm just seeing how far my card will go (and enjoying trying to keep ahead of the reference 5850 guys) while I await the power meter that I ordered.

Cool, I particularly like the network booting.  This gives Windows an 'easy and quick to configure' bonus over Linux which I won't deny.  If you are setting up even a moderate sized farm and/or are just looking for maximum profit (and your time is valuable) then this could be a good choice.

However, it may well be that one can achieve better MH/s or MH/J in Linux and I think there are a few people on these forums who value the extra MH/s more than the 2-3 hours time configuring the miner.  Indeed I doubt I'm alone in actually enjoying the configuration and tweaking.

Care to elaborate?  What is MinerPE?  Do you have any stats which will convince us that MinerPE is superior to other approaches?

I've just had a play and my cards are currently at:

970/330@1.0875V (400 MH/s - 83C - 50% fan)
1030/330@1.1625V (425 MH/s - 65C - 50% fan)
(The first card is sucking in hot air from the back of the second card).

I haven't truly verified the stability of the system yet but the cards have been running well for 2 hours now.  From what I've read on these forums it is VERY hard to get a refenence 5850 up to 425 MH/s so I would suggest that the Sapphire 5850 Xtreme cards are quite a bit better than the reference 5850s.

1.2375V!  That's pretty serious.

How are you cooling your voltage regulators?

Is this a reference 5850?

Yep.  I discovered this just a few hours ago.  Thanks anyway though.

True, you could do this.  But it does mean cannibalising the stock coolers.  Also, I don't think the card will be able to control the speed of the fans after doing this but I could be wrong.

Whacking them into the motherboard was an easy solution for me because I had two 3-pin fan connectors on the board not in use anyway and the cards only need minimum fan speeds with these coolers (temps are 64C and 49C respectively, the hotter card is taking in air from just behind the cooler card).

I'm sure you could get a 4-pin molex to 2x3-pin fan connectors adapter and use that too and use the fan mate 2 which comes with the Zalman VF3000A to force minimum fan speed at all times.

I like the Sapphire Xtreme's in Linux too.  I have 2 and they are both running at 900/300 at 1.01V (735 MH/s total).  I would try higher voltage and core speed but I've not yet found a way to set anything higher than 900 MHz core speed in Linux.

I found the fans on my 2 Sapphire 5850 xtreme's too loud so I replaced them with Zalman VF3000A coolers.  The stock coolers work perfectly well but are of no use to me at all and just sitting in a box .

Note:  Although the Zalman VF3000A coolers are excellent they are not designed for the Xtreme cards.  Each card will block two pci slots and the cooler will extend 4cm beyond the end of the card so make sure you have room before ordering one of these.  Also, the coolers largely obstruct the fan connector on the cards and the fan connectors are 3-pin (the board supplies 4-pin) so you'll need to connect the fans to the motherboard or something.  On the plus side, if your gpus are still the loudest element of your computer then I have to say you must have a VERY quiet PSU.

359 -> 367 (2.2%) on my two 5850's (900/300 1.01V).

Thank you kindly.

I disagree.  Technically kilo as a prefix means 1000 in all cases.  There is an informal understanding that in some situations to do with bit and byte, the prefixes represent 1024, but this is officially incorrect (according to SI) and kibi, mebi, gibi, et cetera should be used instead.

There is certainly no issue with the unit 'hash'.  Like 'watt' and 'hertz', the prefixes work in base 1000.

Silk road is one thing but software piracy?  Are we really that dastardly?  I am shocked!

I would suggest Windows would allow you to get better hash rates because you will have much more control over the card (card creators consider Windows before Linux in general and there are many tools out there to tweak your cards which only run on Windows).

I'm in the middle of configuring a 2x5850 (Sapphire Xtreme) mining rig and would love to clock the cores past 900 MHz but I've found no way of doing that on Linux (Windows users can either flash the bios or use Sapphire's TriXX).

Of course, this all assumes you already have a valid Windows OS License.  If you do not have a Windows OS License then the cost of this will probably outweigh any benefit from greater power over your cards.

Of course.

1LyVjLv5aBwUw8864jYRZEGibLRRuqdnjT

You're welcome.

I believe 'pool hopping' refers to switching pools at certain times to take advantage of the proportional payout system.

With the proportional payout system, the pool mining effort is divided into rounds, each round ending when the pool finds a block.  The pool asks users to submit shares (solutions for an artifical, low difficulty).  At the end of the round the 50 BTC reward is divided between the users, weighted by the shares they have submitted.  (transaction fees are almost always kept by the pool and there may also be a fee).

One issue with this is that as a round goes on, the shares become less and less valuable.  A miner can make quite a lot more (I'd guess a bit more than 120% of solo mining in real-world conditions) by connecting to a small number of large proportional-payout pools and always channeling his shares to the pool with the youngest round.  Proportional pools try to combat this by not immediately publishing when a new round has begun but this is not 100% effective.

To be clear, the pools do not lose money because of this (or other tricks such as griefing), the other pool users do.  Also, some people view this tactic as a form of cheating so you may find pool operators trying to block or ban you for such behaviour.

Switching pools when one goes down so that your miner is always busy is typically not seen as a form of cheating.