Nope, another 5570. Don't believe us, even after you check the model number 1bitc0inplz already pointed out? Here, from Gigabyte's own site: http://www.gigabyte.com/products/product-page.aspx?pid=3344#ov (google model # GV-R557OC-1GI)
And here, from Diamond's site, the card you ordered: http://www.diamondmm.com/5570PE31G.php (google model # 5570PE31GSB)

Seriously, the card you ordered is not a 5770, that's 100% certain. This other one is not a 5770 either, also 100% certain. Save yourself the $ and cancel your order.

Agreed, no way that is a 5770:
- they don't come with DDR3 memory
- the pathetic little heatsink on the picture could never handle the 100W of heat a 5770 puts out under load
- the pcb on the picture is just wrong for a 5770 (too simple, too empty, too short ...)
- finally, the model number leaves no doubt

6770 cards are just re-labeled 5770s, apart for some minor bios differences. So yes, you can Crossfire them, as the chart linked by deslok also confirms.

Current break-even point, including NMC-BTC exchange fees, would be ~0.0407 BTC/NMC. The exchange rate is currently ~0.038, so that's only a 7% difference.
BTC difficulty increases tomorrow (about 15%), so if BTC/NMC price holds, NMC mining will be slightly more profitable.

The network hashrate just stagnated for a few days, it's picking back up now. It's not like that hasn't happened like every couple difficulty jumps. Current projection may be for just a 12% difficulty jump, but that's just an 'instant' projection based on the recent average block solving speed. As the network speed increases over the next couple of days, by the time block 135072 comes around, the difficulty jump will probably reach 25-30%.
I'd guess difficulty will keep going up by 30-50% at least until  ~5 mil.

Pretty much spot-on. At worksize=128, between 180 and 250 the hashrate varies by a couple Mh/s, below 180 it drops fast and above 250 it starts decreasing slowly. By mem clock 300, worksize=256 is ~10 Mh/s faster and peaks at mem clock 354. But the worksize=256 peak @ mem clock 354 is still about 5 Mh/s lower than the worksize=128 peak at mem clock 186.

If this is not a dedicated rig though, then there is one issue with using mem clocks as low as ~200. Any load on the video card, such as playing video or flash, will lower the hashrate a lot more than at higher mem clocks.

I don't have any 5830s, but all 58xx cards should behave similarly. On my 5870s I noticed that below ~250 mem clock worksize=128 works better. In fact, the optimal settings I've found we're at mem=186, worksize=128. That gives a few Mhash/s more than worksize=256 and mem anywhere between 250 and 400.

Maybe you should learn how to browse newegg before you tell other people to stfu. There's multiple 6870 models available for $165-$175 after rebate and a Sapphire one for $175 w/o rebate. At those prices the 6870 is certainly a better deal than the 5830 for $160.

If you can get a 5830 for ~$40 less, or even lower, than 6870, than it's a good deal. If the price difference is below ~$40, you're better off with the 6870:
- if you plan to keep the cards after you finish mining, the 6870 is far better at gaming and also scales much better in crossfire
- if you decide to sell the card(s) when you finish mining, the 6870 will have better resale value. The 5830 was always considered a very poor gaming card, with 5770-level performance at above 5850-level power consumption. And with all the ex-mining 5830s on the market in a few months, don't expect to be getting a decent price for one.
- if you expect to still be mining for bitcents a year from now, the 6870 will have made up the price difference with lower power consumption (~30W)

A lot of factors can affect a card's hashing speed:
- gpu clock speed: this one is a no-brainer - hashrate scales almost perfectly linearly with gpu frequency
- video memory speed: this one is a bit trickier - most cards actually improve performance if you lower the memory speed. The sweet-spot can be anywhere between 160 and 400 MHz for 5xxx and 6xxx series cards. The impact on hashing speed isn't nearly as big as with gpu frequency, but gains up to 5% or so can be achieved
- APP SDK version: 2.1 is supposed to be the best for 5xxx series cards, giving ~10% higher performance than 2.4. Versions 2.2 and 2.3 are not recommended for several reasons (low performance and high cpu utilization). 6xxx series is not supported in 2.1 though.
- Catalyst version: from what I've read results seem to be all over the place, so I can't really recommend which version to use. I'm sticking with 11.5 for my combo of 5870 and 6870 cards.
- OS: again, different people report differently about whether Windows or Linux is faster, same for 64-bit vs. 32-bit
- miner used: phoenix's phatk kernel seems to be the fastest for most users, at least under Windows. Under linux, again, different users have different results with various miners. In my case it's pretty much a wash between phoenix (phatk) and poclbm under linux.

As for card choice... a 5970 is more power efficient and has better performance at a given gpu frequency. You can also use 2.1 APP SDK with it. It should also be cheaper to buy. It's probably just about impossible to get either new or used. It will also be harder to sell when you stop mining, or, if you keep it for gaming, will be a good deal slower than the 6990.

The 5800 series are already last generation technology, almost two years old by now. The GPUs themselves have been out of production for months now as well as most cards. A few AIBs still built some cards from old stocks of GPUs, but by now I think all cards have been out of production for some time.

Strange, I can set the clocks for my two 5870s with aticonfig that are outside the 'Configurable Peak Range'. Might have something to do with the fact that I flashed their bios with already altered clocks (to 920/354). In my case aticonfig reports the range as 600-900 for gpu and 354-1300 for memory, but I'm running them at 950/200 (set with aticonfig). Maybe it's because both clocks set in the bios are already outside of aticonfig's range.

I would use AMDOverdriveCtrl, but my rig is in the basement, without a monitor, keyboard or mouse, so I only control it over ssh. Guess I should have set up a vnc server while I had a monitor still attached .

You need to do:

export DISPLAY=:0.0
aticonfig --pplib-cmd "set fanspeed 0 100"

export DISPLAY=:0.1
aticonfig --pplib-cmd "set fanspeed 0 100"

and so on, for all your adapters.

edit: missed the question about changing clocks. To over/underclock, use:
aticonfig --od-setclocks=950,300 --adapter=0

That would set your first video card (adapter=0) to gpu clock 950 and memory clock 300.

Yeah, just ignore hardly's comment about the psu being crap. Bequiet actually produces very good power supplies and you can be confident that if the spec sheet says 30A @ 12V, that's what the unit will deliver.
I say give it a try, but keep an eye on the 12V, 5V and 3.3V voltages (use a program like cpuid hardware monitor, http://www.cpuid.com/softwares/hwmonitor.html). If they deviate for more than 5% from their nominal values under load, the psu is being dangerously overstressed.

Well, the PSU can output 30A on the 12V rail. That will be enough if this is going to be a dedicated mining rig.
On the other hand, if you plan to use the system while mining, then no, it probably will not be enough. And certainly not, if you plan to keep the 8800GTS in the rig. The CPU can use over 100W under load, the 8800GTS close to 150W and the 5870 around 180W.

Well, the Athlon XP motherboards are AGP only, so no good mining video card will fit.
P4 on the other hand, comes in several varieties, including socket 775, so at least these will fit into pretty much any socket 775 motherboard. It will be powerful enough for mining, even on multiple video cards. Just make sure, since you're looking at used systems, that the motherboard actually has any x16 pci-express slots (some motherboards, particularly from oem manufacturers, use integrated graphics and don't have any pci-express slots for discrete video cards).

A few adjustments:
- 5870 will do close to 440 MH/s at 1 GHz gpu clock. Not many will do that without a slight voltage increase
- 5850 can do up to 400 Mh/s at 1 GHz gpu clock. That will usually require a voltage bump, which brings power consumption very near 5870 levels. At a more modest gpu overclock of 900 MHz, it will do about 350 Mh/s
- 5830 power consumption is usually nearer 5850 level, around 160-165 W
- 5770 and 6770 are the same gpu so all the numbers should be the same
- 6950 can hit 6970 MH/s numbers only with unlocked shaders. That also brings power consumption right up to 6970 level.

I'd add:
- RAM - only use one stick to save a few Watts, even if going for 2GB. One 2GB stick uses the same amount of power as a 1GB stick
- CPU - downclock and undervolt it to save power. Pretty much all CPUs should run at 1-1.5 GHz and =< 1V. Also, if using a multi-core CPU, disable all but one core. And if using a hyperthreading-enabled CPU, disable that too.
- Video cards - downclock the memory to save power and possibly improve performance (depending on model of card and operating system). On a 1GB video card you can save some 20-30W per card, depending on original and final memory clock. To hit optimal performance, you will have to experiment, as the sweet spot can be anywhere in the 180-400 MHz range
- PSU - ignore the total wattage and focus on just the 12V rail. Single 12V rail is the better design for mining, as opposed to multiple 12V rails

Naturally, the 70A is meant as total output of the 12V rail. You can't count on all those amps being available to the video cards; but in a dedicated mining rig, the cpu, motherboard and fans shouldn't use more than maybe 50W from the 12V rail. Even far less if you do your power saving homework .

4x5870 would be no problem, 5x5870 will be cutting it a bit close. Don't get me wrong, it's a very good PSU (it supplies 80A on a single 12V rail,  basically perfect for a mining rig), it's just that 5 5870s will push even that close to the limit. But if you optimize your power consumption (underclock your cpu, underclock the memory on all video cards, don't overvolt the GPUs), it should just about handle it. Still, for five 5870s, you're probably better off playing it safe and getting a 1.2kW unit.


How is estimating about 200W per 5870 in a 4-way rig 'overestimating' needed power? There's more to power supplies than just the number printed on the box. What really matters is the 12V rail, which will supply close to 100% of the power the video cards will draw. A 1kW PSU with a 36A 12V rail is just plain crap and it will actually be dangerous to use it in a 3x or 4x 5870 rig. There's a realistic danger of something in the PSU blowing up, melting, etc., possibly causing damage to attached components or even starting a fire.
That said, a good 850W PSU (70A @ single 12V rail) can handle 4 5870s, if you optimize power consumption and don't overdo it with GPU overclocking. A 1kW unit is still a safer bet though.

That's not how PSU efficiency is measured. The rated Wattage is what the PSU can output, not what it will draw from the socket. So, an 850W PSU will actually give you ~800-850W to play with, depending on the rail configuration. Again, the important one is the 12V rail, which in a typical quality 850W unit will deliver 70A or thereabout, so ~840W.