To give you an idea, the first FPGA miners were running on serial data connections, probably 115kbaud with room to spare

Use Catalyst 11.6 or 11.9 (maybe) drivers.  11.7 and 11.8 both exhibit this bug from what I've read.

I am not a Windows guy, however, so I can't vouch for 11.9's stability. 11.6 should work fine though.

Not all of the CPUs will have a viable second core.. generally if you try to unlock it and things fail, the CPU won't support it.

Unless you're making a general purpose machine, you're better off saving power and leaving it locked.

Do you have another board to try the CPU in?  I think you just narrowed it down to either that or the PSU.

Just the stock bios at the moment, I haven't played with it yet.   That one is my least concern really, it's nice and cool and doesn't draw a whole lot of power for it's 300mh/s, may just leave it

I've had this issue with every 6XXX series card I've used in Linux (Ubuntu 10.10).

Using aticonfig I've never been able to set my XFX 1GB 6950s memory clocks less than 125mHz lower than the GPU clock.  In order to set memory clocks outside of the default ranges at all, I have to use AMDOverdrivectrl to set the default clocks for each power state, but even those don't 'stick' - as soon as I 'push' the card the memory clock ramps up to max.

The 6870 I picked up recently won't drop below 100mHz less the GPU clock.

It seems to be the same issue as Windows boxen, but at least there the manufacturers have specific utilities to get around it.  It also appears that it may be vendor/card version specific, as some people have had success underclocking mem on a handful of these cards in Linux.

With the BTC prices so low and my power usage so high lately, I'm resorting to bios flashing a lower memory clock and voltages on all of these cards this weekend.  Totally kills their use in Windows or for gaming (unless you flash them back), but will be worth it for mining - at this point I need to eek out every bit of efficiency I can in order to stay above water.

Oh, that's ok, we don't mind looking at porn.

How about you get comfortable and show us a little water-pump

Please tell me you sued the city afterward

As for the insulation, the point is to keep the environments isolated - yes it keeps the hot air in, but it also keeps ambient hot air out.  I brought my attic insulation from R30 up to R60 last fall and drastically reduced my heating costs in the winter, and my cooling costs in the late spring/summer.

Then I started mining and threw those savings away

My experience with dual-core cards is nonexistent, but I do know they divide the available bandwidth between the GPUs..  even though mining is bandwidth-minimal, I'd be concerned that a 6990 wouldn't be too happy dividing up a 1x slot.

Take it with a grain of salt, though, again I know little of these things.  If someone else has run 5970/6990 or other dual-gpu cards on 1x PCIe, please chime in

My experience with the sapphire 5830 extremes is that the stock cooler is pretty good - even in a case with little breathing room between them, in a hot room, OCd as hard as they go on stock voltage, they're consistently < 75C.

Maybe your $$ is better spent on improving airflow?

Quick suggestion - it may be worthwhile to offer some boards with SATA power connectors.

I've been buying a lot of higher-end PSUs lately, and even the 'cheapo' CX600 I just picked up has 6 SATA power connectors but only 4 molex.  I've also got more molex > SATA converters than I can count, as I'm sure most system builders do.

I imagine the SATA power connector is slightly more expensive, but it'd take appx the same board real estate, and will help 'future proof' the design.

I could power 10 of these on this CX600 using SATA power cables, but only just 4 with molex.  My HX750 has twelve SATA connectors and only 8 molex - it could power 20 of these (on the 12V rail anyway, are these pulling 5V?).

SATA > molex adapters are sure to be difficult to find as well.

Well now that I know the CX600 is a group-regulated PSU, this makes more sense.

When the system tried to bring up the cards it was only loading the 12V rail, without 5V/3.3V load the PSU was shutting down as a protective measure.

I got it working fine by using a 100 Ohm resistor between +5V and a neighboring ground on the 24pin connector.  The same could be achieved by connecting a spare HDD or optical drive, or anything else that completes the 5V circuit.

When I was powering it on with nothing connected, the lack of load on any rail meant there was still a balance, and it didn't trip the current/voltage protection circuit.

I retract my CX600 recommendation - it's rated capacity is 480W on the +12V rail, which will not be nearly enough.

Go with a 750W+ single rail and pay attention to the +12V rating, or go with dual PSUs.

Thanks for pointing this out BkkCoins

I recently picked up a new mobo and a pair of Corsair CX600's so I could pull the cards out of my file server, where they were baking the crap out of my drives.  I've got the new system running with one 6870 on the first CX600, but I'm having some odd behavior with the second, which will be powering a pair of 5830s.

When I jumper the PSU (green PWR_ON to black ground) and turn everything on, it spins up and all seems well, but after about 30 seconds or so the PSU just shuts off.  

My initial thought was that it required some 3.3V or 5V load in order to stay up, however this is where it gets weird - if I jumper the PSU and power it up with nothing connected, it stays on indefinitely without any problem..

I've got several other Corsair PSUs - HX650, HX750s and CX430s, and all of them work just fine jumpered.  Architecturally, this CX600 should be near identical to the CX430, so I'm a little stumped why it'd behave differently.  I haven't yet tried swapping it with the working CX600, will be doing so in the morning, I'm just wondering if anyone has seen this kind of behavior with a jumpered PSU before?

[edit] These are the CX600 V2's, 80+ and 480W rated on the 12V rail.

To my knowledge, electricity and how we use it hasn't fundamentally changed in, well, since we started using it.  I didn't watch either, but I have no doubt the concepts may still hold up.  Perhaps you might not have ideas 20 years ahead of their time, but that doesn't mean someone else didn't.

Ok, next time actually read the post you're responding to.  I said:


I was clearly referring to other rigs as having better cooling, and using fewer and cheaper materials.


Funny thing about the English language, sometimes a word can have more than one meaning.  Also, many different words can mean the same thing.  I was giving praise, not writing a technical manual.  In the future I will just say, "I'm going to praise you now by saying this thing is not terrible."   Better?


Ugh..  name one other component that uses as much power or produces as much heat as a modern high-end GPU.  Next, show me a computer with 4 or more of that thing.

Outside of really high-end gamer builds (which don't run the GPU at 95%+ 24/7), or folding@home builds, nothing else really compares.

The fact stands that my opinion is that loglow's build is nice - it's a hell of a lot nicer than my open-air $5 wooden frame, and incidentally appears to stay cooler as well.  You can keep shitting on my opinion if you'd like, but I'm done responding to it.

Lets try making our points without being condescending twits, eh?  Building mining rigs with power-hungry hot-as-hell GPUs isn't exactly old hat for anyone.  By 'smartly built', I meant this is clean, concise, and functional. 

Certainly there are rigs with better cooling, fewer and cheaper materials, that doesn't make this one any less nice.

You might've missed the part where i calculated at 150W, these cards don't really pull much more than 100W.  There's quite a bit of room to breathe, and the Corsairs are quite happy working at their maxium stated pull.

It comes down to price really, if you want absolutely safe and highly efficient, go with a 750W gold.

Lmao..  I'm wager my server closet is ~80-90dB, there's no such thing as 'sleep' anywhere near my rigs  I couldn't hear my phone ring, at full volume, while I was in there yesterday.


Also, that rig is kickass loglow, very smartly built.  You've got me thinking about redesigning my setup yet again

5770 has a TDP around 110W, lets call that 150W to give you some overhead and room for overclocking.

x4, that's 600W.

That's a pretty good number to be at - there are bazillions of power supplies around the 600-750W range, and even at 600W I think you'll be just fine, base system included.

My recommendation if you want to go cheap is this guy: http://www.newegg.com/Product/Product.aspx?Item=N82E16817139028

[edit] Don't use a single CX600 - see my post below.

The Corsairs are strong PSUs, I'm using 6 of them currently and just picked up two of the above as spares, though I kind of doubt I'll ever need them.

That PSU will give you 2x PCI-e 6pin connectors, and you'll just need a couple adapters to go from molex > PCI-e 6pin.