You and I are clearly of one mind.

As you probably know from the latest patch thread
http://forum.bitcoin.org/index.php?topic=25860.40
I've already tested SDK 2.4 a little and found that at high core clock speeds SDK 2.1 is still slightly superior for the Sapphire HD5850 Xtreme.

If you are done with trying then let me know and I'll fill gellimac in on the details of my configuration.

My card RAM is already at 360 MHz and I've tested but I can't find a better frequency for the RAM at my core speeds if I'm only interested in MH/s.

As for CPU usage  I've not touched my CPU settings at all and the miners only use about 0.4% each.  I even removed the fan from the CPU and placed it to cool the back of my hot card (the heatsink on the CPU is not even warm).  I'm assuming significant CPU loads is a Windows thing.

What interests me is how SDK 2.1 seems to be better at higher clock speeds whereas SDK 2.4 with your kernel is better at moderate speeds (940 MHz or below).  I admit I have little data on this but if anyone else gets the same results it would be interesting to know why.

Can you get it higher with the kernel improvements?  No improvements have helped my rate since the MA tweak.  Feel free to use the recent improvements (from earlier today) to beat 413.3 MH/s.  If you have 412.0 MH/s then I'm hopeful you'll beat it and I'll find out how to improve my rates.

No.  I simply want to give whopper some time to try and improve on my rate.

I will but I'll give whopper a day or two to try and beat me and claim some BTC.  It's not the same challenge if I give him all of the information I have.

I will say that I am using Linux headless and driving everything with ssh so whopper and I are on a level playing field.  I don't know about adjusting base-clocks at bios level so I hope to learn some more myself with this little challenge.

Possibly you're using an old version of SDK.  I get a fatal error when trying this with SDK 2.1 but am fine with SDK 2.4.

Now that's what I'm talking about!  Very nice!

Can you increase the RAM speed a little without losing stability?  If so you might score another MH/s.  I'm thinking around 350-360 MHz.

With lots of tweaking and testing the best I've gotten from a 1000MHz Sapphire 5850 Xtreme @ stock voltage is 413.3 MH/s (fluctuating between 413.1 and 413.5).  If you can beat that then please let me know for some BTC.

Can you push the core clock any higher than 1000 MHz?  Are you bound by Trixx or by stability?

Interesting.  When I change to 128 my rate drops by a full 10 MH/s per 5850.  Perhaps there are other settings which need to be changed in tandem to yield an improved rate.

Could you let me know your other settings?  RAM speed, SDK version, Catalyst version, kernel type/version/patches, kernel parameters.

If your comments can improve my rate I'll tip in BTC.

Another improvement but still not enough to beat SDK 2.1 for me.

The two phatk kernels of interest to me are:
- Kernel A = standard phatk kernel with the MA tweak applied.
- Kernel B = the latest kernel from this thread.

I'm using: Linux, Catalyst 11.6, a Sapphire HD5850 Xtreme:

At 900 MHz things look promising...

[900 MHz - 360 MHz RAM]
SDK 2.1, kernel A: 364.4 MH/s
SDK 2.1, kernel B: Fatal error
SDK 2.4, kernel A: 360.8 MH/s
SDK 2.4, kernel B: 365.6 MH/s

...but at higher core clock rates SDK 2.1 takes the lead once more.

[980 MHz - 360 MHz RAM]
SDK 2.1, kernel A: 404.7 MH/s
SDK 2.4, kernel B: 404.3 MH/s

[1020 MHz - 360 MHz RAM]
SDK 2.1, kernel A: 421.5 MH/s
SDK 2.4, kernel B: 420.9 MH/s

I would give some higher clocks but I can't go much past 1020 MHz without overvolting my card and I don't want to do that.

I tried playing with the RAM frequency but everything dropped off slowly as I lowered it and quickly as I raised it no matter which kernel or version of SDK I choose.

It's a shame that SDK 2.1 cannot drive your latest kernel but then I guess you are specifically designing it for SDK 2.4 and are now using features which are not available in SDK 2.1.  It would be great to finally put SDK 2.1 to bed but another MH/s sounds like a tall order at this point.

I have no data on accepts and rejects (I mine solo).

The settings you are using are very good for these cards using phatk (I can only assume GUIMiner uses phatk).  Now that I see that these settings are good I ask whether or not you are using SDK 2.1.  For me SDK 2.1 is a fair bit faster than SDK 2.4 (with either Catalyst 11.5 or Catalyst 11.6).  If you are currently using SDK 2.4 then I suggest you give 2.1 a go.

Another thing to consider is whether or not the read-out from GUIMiner is taking a moving average before displaying the hashing speed.  I know phoenix does this by default and I have to pass it '-a 1' to make it stop.  If you manage to remove the moving average filter from the readout then you may see your hash rate jumping up and down, perhaps by a few MH/s.  Try to reduce this jitter as much as possible by closing other running programs and/or reducing desktop effects/resolution if your card is driving the GUI.

Finally, because your clock is as high as 1000MHz you may find you gain another megahash by increasing the RAM frequency to 350 or so.

You can probably reach 400MH/s by playing with the software rather than pushing the card further.  I've found setting which will give me 413 MH/s for a 1000MHz 5850 but you might have to pull that back a bit if your rig is not a dedicated miner.  Let me know if you want details.

Do you have the latest miner kernel code?  There have been some improvements in the last few weeks.

Well, you may get better results.  For me:

SDK 2.4:  393.7 -> 396.8
SDK 2.1:  399.4 -> 399.4

That's for a Sapphire HD5850 Xtreme 970/300@1.0875V using catalyst 11.6 on Linux and phatk
VECTORS BFI_INT FASTLOOP=false AGGRESSION=14 WORKSIZE=256

Definitely try out both SDK 2.1 and SDK 2.4 though.  If you are using Windows there's also an early version of SDK 2.5 with catalyst 11.7 which may be worth a look.

When you overclocked the card did you also increase the voltage at all?  If so, by how much?

How long has the card been mining?

The reported temperature will probably be for the GPU and it seems unlikely that 75*C has killed it, particularly if you've only been mining for a few months.  However, if you've increased the voltage much then the VRMs could easily have gotten too hot.

I used AMDOverdriveCtrl and Catalyst 11.6 (I'm on Linux).

Edit: You can also use aticonfig directly, all that's needed is Catalyst 11.6.  To set the clock rate to 1020 say on your first card use:
aticonfig --adapter=0 --odsc=1020,0

I have a dedicated Linux miner with 2x Sapphire HD5850 Xtreme and they clock to 980 MHz and 1020 MHz respectively (so far I'm only 16 hours into the test but it looks promising).  This is without flashing the BIOS or mucking around with voltages (stock voltage is 1.0875V) or touching Windows or DOS.  This nets me 826 MH/s so I'm sure you can get a little more than 300 MH/s each out of these cards.  Don't be confused by the many threads concerning the reference 5850's, these cards are far superior when it comes to mining.

Be warned that there are some problems with the fans on these particular cards so it would be wise not to push them to 100% for too long but at stock voltage you should be fine at 70% or so.

Let me know if you'd like some details.

Oh, and you can overvolt your cards on Linux without flashing card BIOSes or using Windows/DOS.

This is what I don't understand.  If you overclock your card but you don't touch the voltage and you keep the card cool (say 50*C), what causes it to "toast"?

I know that one of my 5850's goes over 1000MHz on stock voltage (1.0875V) and is able to sustain it for hours (I haven't done a 24 hour test yet).  Everyone else on the forum likes to raise the voltage before bringing the clocks that high and I wonder what I'm missing.

I don't want to kill my card, indeed I'd like it to last a long time, but I feel like it would be wasteful to keep the card at stock clock (725MHz) or even just at 850MHz or so.

With Catalyst 11.6, yes.  I'm able to at least.

I only use AMDOverdriveCtrl for changing the voltage.

Yes, after 45 minutes things look highly suspect to me.  A good 3 hour test is useful for comparing different very-good setups but this tweak seems to have really hurt your accept/reject ratio.  Let us know your best when you're done testing.

I found when pool mining that most of my rejects came shortly after new work was pushed.  As a result I had to try much longer test runs (3 hours or so) before coming to a conclusion about the miner's efficiency.

Just to be clear, these positives are all well and good but few are actually "advantages" over Linux:

- Linux can handle process/core affinity too but there is little need for this here.  CPU mining is very inefficient and GPU mining requires practically no work from the CPU.

- Linux has easy remote management with ssh (many thanks to the BSD guys).

- I'll give you that Windows has a big advantage over Linux when it comes to drivers.  The ability to overclock comes under this heading too.  That said I'm able to overclock and overvolt my non-reference 5850's in Linux without any BIOS flashing so things aren't so bad, but certainly Windows has the advantage.

- You can't list power management as a positive and a negative.

- Being easier and faster to setup is a matter of opinion.  Indeed, I find Windows much harder to use and more confusing.

Also, I doubt there is any "uberness" associated with using Linux these days.  Both Mac OS X and Ubuntu Linux are seen as viable alternatives to MS Windows.  I think mining bitcoins says more about computer nerdiness than using Linux.