Is your miner dedicated to mining or are you using it for other tasks?

My experience is with Linux and my cards go to 975 MHz and 1020 MHz stably at stock voltage (Sapphire HD5850 Xtremes)  I thought my slow card could handle 980 MHz but it crashed after a few days.

The most important factors for me in achieving stability were:

• Not running a GUI,
• Keeping the temperatures low.

On the first point, people find their cards are more stable after disabling Flash hardware acceleration for example.  Removing desktop effects can help too.  I've removed everything: no mouse, no console mode, not even a black screen (when a monitor is connected to a card it reports no signal).  I don't know how to do this in Windows but thought it worth mentioning as it made a big difference to me.

I don't know anything about "speed vs. voltage sweet spots".  I just fix a voltage and manually search for the max stable clock (which can take days).  Currently my cards are both undervolted to 0.9875V and are clocked to 847/308 and 899/327 but with only 48 hours continuous mining so far I can't be too sure of the stability of this.

I don't think that Linux is any better at handling crashes.  When one of my cards crashes I've not found a way of restarting it without rebooting the system!

I'm a nervous kind of guy

From these forums (I've been away for 2 weeks and don't have links for the threads).  I've just re-read my statement and it's not really fair.  I don't mean to imply that all 5850 Xtreme coolers will fail within a few weeks if left at 100%, just that more than a few have failed.  I certainly didn't struggle to sell my 5850 Xtreme coolers after I'd installed my Zalmans.

I agree with this general sentiment.  With the prices of Bitcoin as they are right now you could easily find that you make more by lowering the voltage and you won't have to sacrifice many Mh/s for large power savings.

As I said, I'm running both cards at 0.9875V and I'm still getting 733 Mh/s in total but the miner is only pulling 300-302 watts from the wall (and that's using an ordinary HDD and a Silver certified PSU).

It doesn't seem so bad to me, even for 24/7 usage.  90*C makes me nervous though and I would consider lowering the voltage just to get a handle on the temperature.  If you can find a way of cooling the card to 75*C or below and keep the 1.1625V then more power to you!

(I believe the VRMs on these cards work in increments of 0.0125V so if you tell the cards to run at 1.16V they'll probably be at 1.1625V.  Again, I could easily be wrong; I have no source for this).

Careful with this, the Zalman VF3000A is designed for reference 58xx cards and the Sapphire HD5850 Xtreme is far from reference.  I've attached these coolers to my Sapphire Xtreme 5850s but:

• I had to leave the existing heatsink in place for the VRMs,
• the coolers make the cards about 3.5cm longer,
• the coolers use 3-pin fan connectors but the Xtreme cards use 4-pin fan connectors so I use manual fan control.

I have some pictures here.

With these coolers in place and a fan-heater set to cold focused on the VRMs of one card I was able to take the card to 1.3V (Catalyst 11.6 + AMDOverdriveCtrl wouldn't let me go any higher) and the card ran fine for 3 hours at 1110MHz, 1.25V (461.1 Mh/s) and maintained a temperature of about 50*C.

For 24-7 use I would draw the line at 80*C and would try hard to keep the GPU temps below 75*C.  I'd also limit myself to 1.1625V.  At 90*C I would expect the life of the card to be shortened significantly but I'd guess it would run fine for many months.  I'm certainly no authority on the subject though.

Personally I keep my two cards at 0.9875V (the cores are at 847 MHz and 899 MHz).  For this the Zalman coolers can keep the cards at 58*C and 44*C respectively when turned down to the minimum I can with the manual controllers.

One final thing, the stock coolers on the Sapphire HD5850 Xtreme cards are known to fail if they are set to 100% speed for a few weeks.  From what I've read on the forums I'd be uncomfortable having them above 80% continuously.

I'm also curious.  What was decided in the end?  Anyone?

Yeah - seriously!  I've come up against this before when trying to find the maximum hash-rate for a 1GHz 5850 and ended up being well and truly trumped by a Windows user with Catalyst 11.4.  I may have to try playing with this version of Catalyst again.

Have you tried using phatk 2.1 or phatk 2.2?  It has an option "VECTORS4" which you could give instead of "VECTORS" which might improve your hash-rate.  "VECTORS4" is basically for people with high RAM clocks.  Don't forget to lower your WORKSIZE too.  Here's the main thread.

Right now my temps are 57*C and 43*C respectively (one card is sucking in air being heated by the back of the other).  The miner is in a pretty small room and if I cut off all ventilation for 12 hours or so the temperatures can reach as high as 62*C and 48*C (peak from the last two weeks).  This is with all fans turned down to minimum to reduce noise as much as possible.  If I turn up all the fans and keep the room reasonably well ventilated the temps sit at about 44*C and 36*C.

I've been through a phase of maintaining high clock rates but I much prefer having a cooler and very quiet miner achieving more than 2.4 Mh/J and I certainly don't mind sacrificing around 100 Mh/s (from 828 Mh/s to 733 Mh/s) for this.

Good question, I've wondered this myself.  I'm guessing it was the GPU rather than the VRM that was suffering simply because everything was fine for 3 hours at 1110 MHz and consistently died in seconds at 1140 MHz (same voltage).  When I replaced the stock cooling I was not able to improve the VRM cooling and the sticky pads and stock heatsink remain in place.  For the testing I had a fan-heater turned to cold and focusing a blast of cool air on the VRMs.  I have no way of measuring the temperature of the VRMs and can only assume they were roasting.

Thanks for the tip but I'd rather take good care of my cards by running at lower settings.

I found my old post about my personal best MH/s for a 5850.  461.1 MH/s at 1110 MHz core clock, 3 hours continuous with very few stales.

Higher clocks such as 1140 MHz were fine for a short time but after 30 seconds or so the cards cut out due to high temperatures.  I couldn't find a way to circumvent this so I guess (I have no water-cooling).

As I said before, I don't run at these speeds normally - this is just a bit of fun.  The same card is now running at 377.5 MH/s (899MHz/327MHz@0.9875V) and I'm trying to get my miner up to 2.5 MH/J.

Just had a quick look to confirm that the two offered reward structures are proportional and pay-per-share.

I personally like to think that some bitcoin miners have a grasp of probability sufficient to understand the fundamental flaw with the proportional reward structure.  No amount of obfuscation can prevent pool-hopping.

For now I'll continue to mine solo.  I'd consider joining this pool for PPLNS.

I don't believe it's possible to fold@home with 5850s on Linux and GPU-based Go playing software is still in early development so if mining becomes unprofitable I'll only use my miner for the occasional bit of research and/or fun.  60 fps full-HD Mandelbrot fractal explorer anyone?

My mining rig costs about 0.8 GBP (about 1.3 USD) a day to run and makes about 0.4 BTC a day.  I can 'hang on' by selling my BTC on the market and pocketing the difference.

I guess the trick is to avoid mining with CPUs and stick with GPUs or FPGAs.

Your title has reminded me of a fact useful for quickly mentally converting kh/s into BTC/day:
At a difficulty of 1`000`000, 1GH/s is worth very close to 1 BTC/day.

Also, where I live electricity costs are such that wattage is almost exactly GBP/year so I have a very easy time with the figures.

Sapphire HD5850 Xtreme: 899MHz/327MHz@0.9875V
Linux x86_64, Catalyst 11.6, SDK 2.1
VECTORS BFI_INT AGGRESSION=14 WORKSIZE=256

phatk 2.1: 376.9 Mh/s (+/- 0.1 Mh/s)
phatk 2.2: 377.5 Mh/s (+/- 0.1 Mh/s)

I'm genuinely surprised that low RAM clocks cause your computer to freeze.  It doesn't seem logical to me and runs contrary to the collective experience of the forum members.  Are there any other programs using your cards?  Is one of your cards driving a window manager for example?  Have you tried lowering the RAM clock on a card which is not being used by any programs?

I'm not the best guy to ask about flashing BIOSes and just try to do the best I can with the given BIOS.  There was a time when I was limited to 900 MHz core clock which was very frustrating but flashing the BIOS was simply too much work to be worth doing.

Best of luck.

My cards run fine with the RAM at just 50MHz with no change in voltage.  I doubt there are stability issues with low RAM clocks.

I'm assuming you're having difficulty setting a low RAM clock speed in the first place.  For this I use AMDOverdriveCtrl.  I create a simple overdrive profile such as


and then apply the profile to the card with the command


I imagine in Windows this would be even easier.  I found it was necessary to give the same RAM clock speeds for all three performance levels for the setting to apply.  I don't think AMDOverdriveCtrl can handle lower performance levels to have higher RAM clock speeds but I could be wrong.

My cards are also Sapphire Xtreme's but they are a few months old which might make a difference.  I've heard that the latest BIOSes Sapphire have put on their cards are more restrictive.

In testing I got my 5850 over 460 MH/s (can't remember the exact figure) at 1110 MHz core clock, 370 MHz RAM, 1.25V.  I can't vouch for the long term stability of my settings (I managed a continuous 3 hour run and then the card crashed immediately when I started mining on an adjacent card).

These days the same card is giving me 373.1 MH/s (890 MHz core, 300 MHz RAM, 0.9875V).  I haven't tried the Catalyst 11.7 or SDK 2.5 yet as I've been away for 2 weeks.  I personally prefer the lower voltage and my dual 5850 rig is currently running at 300W, giving me about 2.43 MH/J.

There are many factors which have a significant impact on a cards stability and even a simple thing such as disabling Flash hardware acceleration can allow one to reach a higher stable clock rate.  There would likely be a long list of differences between the card which crashes at 900 MHz and the one which runs fine at 1030 MHz.

What I find interesting is how two seemingly identical cards, run in the same slot of the same mining rig, kept at the same temperature, configured and maintained in the same way, can have such different maximum stable clock rates.