That's a man's work!... (no offence if you are a woman)

Perfect, now all you need is a gate-T so you can manually route heat into the house in the winter!

Don't forget to varnish that wood! (Or wax it)

You may want another intake, if that doesn't have a booster-fan in there too. Always have a larger intake than exhaust, because the exhaust is compressed and the intake is negative. You want to reduce negative pressure so it does not pull air from other doorways or windows or wall-outlets.

Wait, what??? These cards can be used to play games too! I thought these were only for mining! (Talking about 7950's )

temp sensors are NOT accurate. They are cheap ball-park values, at best.

Chances are... you have reached a limit of a voltage regulator, or the heat has destroyed a seemingly useless "sensor" from being displayed correctly, or you just have an actual "hot" chip.

Measure the actual temps with an IR temp device. (You can return it when you are done.) They are like $20-30 at tiger-direct or walmart or home depot.

Best solution for reduced temps is to lower your voltage (which will also require a slight clock lowering if you are at an upper-limit for clocks.) That will give you greater cooling than clock adjustments, thus keeping most of your speed and lowering power consumption and heat. (If you know a safe-range of clocks, you can put that range in cgminer, and it will auto-adjust to keep a target temp.)

MSI Z77A-G45

It just works. lol, sort-of... has one trick-slot Middle slot 16x needs a presence jumper when used as 1x with all 6 slots full. Only 6 of the 7 are able to be populated. One 1x slot disables, when a 16x slot is filled. But they are worth every penny.

The "ground" is all the same, for everything. You are getting ground from the PCIe pins, and also from the 6-8pins above. (Thus the two extra ground wires in the 8-pin, which are just the two from the 6-pin, shared.)

That is the "common" = "-12,-5,-3,-1.5" all in one. Also the "Frame" and anything-else metal in the computer.

You can also undervolt them, and they will run more efficiently, within that load.

I had 4x 7970's running at 980mv and 980core at 75memclock... was under 700 for the whole system. You can also power the fans with an external PSU, or wall-wart. You don't HAVE to have them controlled by the computer and GPU controllers. Save the clean and precious power for the GPU.

You are forgetting a few major components...

1: Alt coins = FAD = SO was/is bitcoins. That point is moot. If it is "cheap" to produce bitcoins, the price will drop. Since it is harder (more overhead) to produce alt-coins, that alone is what gave coins the actual value, and will give them value when we all go there and start using them. You have to remember, there are millions of US and only a few hundred or thousand with large quantities of ASIC's that will be still in the game here. Thus, they can have all the coins. They can trade them with themselves. We will go where the actual gains are. We are the majority they stole it from. We will just up and leave.

2: You are forgetting "transaction fees", which are increasing, and will eventually be larger than any "rewards". As long as the pool-operators are not squandering them, and just feeding us "treasure-hunt coins". (That is where solo mining pays-off, and rewards for coins are just a bonus.)

3: If the hardware lasts only a few months, it has been paid off. Provided the coin value does not fall, and you didn't cash-out... or provided you did cash out now, while the value is still on the high end.

4: OMG THE SKY IS FALLING... Didn't they say all this when GPU mining became "new"... OMG the value will plummet! OMG you are going to loose everything! OMG Its going to crash! Um... No, we just buy ASIC's instead of GPU's and we retire the GPU's, or move them to a profitable alt-coin. The sky is not falling.

lol.. or... we just "change the code" to add security that asic's can't handle. It wouldn't break anything behind the scenes, but it would surely keep all the people holding coins from loosing value to the ASIC corporate-manufactures... Who will NOT be us.

Plus the 7990's will be $800 soon, and consume less power, again. For those still doing the GPU thing.

Bitcoins are the new "Linden", the new "paypal", the new "myspace", the new "facebook". People won't put-up with all the massive synchronization and oversized wallets and account theft, unless someone with real structure manages all this correctly. (With less loss, and cutting out all the childish bickering.)

The issue is because each card is assigned an ID. When the card fails, and is reset, it has a new ID, and cgminer is trying to talk to the old ID. (Thus, the old ID appears dead.)

If it is failing, you don't want it to keep running. You want to make adjustments so it does not fail. (Otherwise you will kill your hardware for good, if it keeps rebooting and using the "crashing" settings.)

Turn down your clock a few notches, or bump-up your volts a bit, or try adding more memclocks if you have those down real low. (Unless it is a temperature issue, or PSU supply issue... then you need to reduce voltage and clocks, so it draws less power and runs cooler.)

The "hash" is actually a "double-hash"... They do it twice, per problem.
SHA256(SHA256(x)) = (This hash)

The RATE = The number of those... per (second/minute/hour/day)...

EG, 600,000,000 hashes/s = ~0.00055THs, ~0.559GHs, ~572MHs, ~585,937KHs

http://egret.net/kb__mb.htm

Hashes / time = hash-rate

600,000,000 hashes done... took 3 seconds = 200,000,000Hs = ~191MHs

It is a rough estimation of the "ability" to process "hashes", which are the majority of the workload of the processing of blocks. Other things happen too, but that is the most dramatic, for processing.

It shows two things... The speed of the code used to process, and the speed of the cards ability to use that code to process. (Faster code processes faster, as does a faster card. "Diablo" is slightly faster code, in some instances, over "poclbm".)

Stealth cases...

1: Large oven-top microwave. Gut it, it can hold 2 Mobo's and up to 14 (single-gpu) cards, 4 PSU's... Has kick-ass venting on most models.

2: Medium/large in-window AC unit. Gut it, it can hold same as above...

3: Fireplace insert. Can hold same as above... Heat can be ducted in to the house in winter, exhausted out the chimney in the summer. (Might want to provide supplemental "fresh-outside air" for the supply.)

4: Mini-Fridge. Gut it, add a vent and add a silent bathroom-blower-fan. Can fit 1-3 mobo's, depending on the size of the unit.

5: Dryer machine. Gut it, you can fit up to 6x mobo's inside. Needs a good green-house or oven blower to vent it. Set it up next to the real dryer, and blow exhaust into the dryer to save on drying bills. (Use that 240v line for more efficiency.) Add your water-in for your hot-water tank into the exhaust, and save money on hot water too.

6: File cabinet (tall). Can hold 4x to 5x individual setups, in a convenient pull-out access setup.

Um, just take a switch off a tower, or connect two wires to the pins. Touch them to turn it on, touch them again to turn it off.

Use a screwdriver... a coin... a paperclip... a jumper...

Are you willing to pay all that extra money for a simple switch to be on the mobo?

If you set it up to "turn on after power failure"... you just use your PSU switch, or power-bar switch to turn it on... (after shutting down properly). Just turn it off then back on... poof.. instant power button!

No need to flash anything... That will just brick most cards.

AB and catalyst is all you need. (But you have to unlock them by agreeing to the disclaimer, and do an initial "sorta-hack" to get the extended ranges to function. Involves just deleting the profiles and restarting AB without "shutting down" the PC.)

Voltages are NOT "fixed", they are "locked". Locked = can be changed if unlocked, but have a limited range and steps that are "stable". (This is per-card, as it is a hardware limitation thing.)

You want to run at the fastest core-clocks, for the lowest voltage. UNLESS you have water-cooling and unlimited power. Then you can go into the higher/stable voltages and OC clocks.

EG, my standard rigs are roughly this...
CoreVolts: 1100mv
MemVolts: 1500mv
PowerLimit: +0 (-20 on special occasions)
CoreClock: 1100MHz
MemClock: 150Mhz

That results in about 215watts per card at the wall, at 640MHs. (About 2.98 Hashes per watt) All cards air-cooled and under 75c, as low as 65c. (FYI the temp sensors are crap on these cards. Uncalibrated and not true to actual temps. Actual temps are about 69c on all cards.) Fans inappropriately running from 75%-50%, due to false temp readings from the cards. 80% for all, temps stay as low as 72c-58c.

My limiting factor is wattage. With more watts, I would push the clocks up to 1200, with a 1150 core volt and a 300 memclock, taking them easily to 690MHs, where cooling becomes a concern. Water-cooling just isn't worth the added cost. It is cheaper to get more cards and build another rig, with standard power and cooling.

Best hashes per watt that I have ever had, was only stable for a day of operation, at 3.33 hashes per watt. Running 640MHs @ 192watts. (1080corevolt, 1120coreclock, 1500memvolts, 75memclock, -10% power, fans running 70%.) One temperamental card destabilized the bunch.

There is no such thing as separate power plains on cards. They all share the same power supply, the same ground, and gate regulated 12v sources...

Again... as I said... if your 6-8pin does NOT supply enough POWER, it will DRAW from the PCIe rails.

Obviously, you do NOT supply enough power through your other connections.

If it was isolated, it would NEVER be able to power the card. It would have to power something unrelated to the card, like a fan or data-controller, which is still connected to the other sources.

It may have horrible designed gate regulators, but that is NOT "isolation". If that were true, then "everyone" should have that issue, and those who use correct power, do not, thus, it is your assumption that it is isolated.

I just tested one, and the traces are NOT isolated, and all connections test as connected.

Cards haven't been built with "isolation" since the voodoo cards came-out.

Check your power. Your PSU is obviously "short" on its values. Possibly you are seeing the MAX/peak values, and treating them as if they are constant values. (Read the AMPS off your 6-8pins, your card regulator is just trying to distribute the load across all three supplies.)

If you are using a 5870 still... Power is the least of your worries.

If you spent $2500 on a 7970 rig, and mined from now, until release-day...

You would make more than you would-have, if you gave them $2500, and let them mine with your hardware for the expected 9-months delivery. (After they mine and setup 50% of the market, drive up the difficulty, and make your 5GHs toy that will only mine pennies a day, by the time it ships to you.)

They are already made, and running... Earning them bitcoins. Why would they actually deliver. They Want you to get a refund. They have no intention to ship. They are taking money to build themselves more and more, and using you as a free loan.

Sure, they throw out a few 5GHs machines, after creating an equal number of 500GHs machines for themselves. Thus, mining our dollar, and giving away pennies.

Suckers!

Funny thing is, this is a free market, and we can easily lock-out them with code that they just can't process.

We will do it to survive, and because we can. And because they can't.

Funny thing is.. if they hold all the bitcoins... who is going to want/trade/buy/sell them? They are on a path to destroy bitcoins, and will end-up with them holding all the bitcoins that no-one wants to use or sell or buy. lol.

It has nothing to do with 1x or 16x... all have the same power. The "extra" x's are just data-connections, not power.

It has to do with "your available power on your 12v rails" which connects to the 6-8pin connections. If you do not supply power there, it will be drawn from the PCIe rails.

If the total of your cards exceeds 75watts (the standard "must supply" for PCIe), using the (standard 25w draw max for cards on a PCIe slot), then you NEED supplemental power.

Or, if you use cheap china risers which can barely handle 15watts of power on those thin wires... then you need powered risers.

Or, if you simply have more than 3 cards, which would possibly pull more than the PCIe supply of 75w, from the multiple 25w-pull from each card. 3 = 75w (ok)... 4 = 100w (bad)... If your 6-8pin connections are not supplying the required "load".

Although the "standard" is 75watts, many OC boards are designed for more than that. Crossfire-ready and NVidia-SLi boards and PSU's are usually designed for the "extra PCIe loads". But the china-wires (old hard drive data-cables used to make cheap risers), are not.

The best way to determine if you need "powered or not", is to simply do the math.

If your 12v rail supplies 80a (960w), and your PCIe can supply (75w), and you have four cards that pull a maximum of 250w (1000w total draw)...{always calculate for max}...

Then you can put 4 cards (1000w) without powered connections, because...
(960w + 75w = 1035w) Thus, you should have a surplus of 35 watts... or rather, only 40watts will be "pulled" from the PCIe, at max-load. (40w / 4 = 10w per PCIe slot, which the thin wires can handle.)

NOTE: I run 6x 7970's which pull only 230w max/constant. (Total cards: 1380watts + 45watts for the other stuff. {Mobo, SSD, Ram, CPU}) = 1425watts max, at the wall. (Thus, it is actually about 20% less than that, at the mobo and PSU, sue to inefficiency at 90% load on 120v power.)

My 12v rails supply 115a (1380w max/constant) + (75watts from PCIe potential) = 1455watts. Thus, I have a surplus of 30w, or I draw only 45watts from PCIe at max/constant, which is 7.5watts per slot. (Less, because 45 watts is going to the other stuff, but that ALL still goes through the 14-pin connector on the Mobo.)

The PCIe bus can only "safely" supply about 75watts total.

Each card can pull up to 25watts... Through the PCIe rail. (Poorly designed ones will pull more)

They "pull" whatever you can not supply through the 6-8pin connections, through the PCIe-rail.

EG, if your 6-8pin connections can supply the full 250watts(20a@12v) needed for each running card, almost no power will be drawn from PCIe rails. (almost)

However, some 12v rails supply up to 25a, some only PEAK at 20a, thus, in constant operation, they will TRY to pull power missing, from the PCIe rail.

This is where it is an "issue"...

The pins have a better contact rating which can handle the 25w feeding to the card. The "riser cables" are cheap thin wire, and not sufficient for 25w to be delivered through the PCIe connection. Thus, the wires melt.

If you only have 3 cards, and a good 20a(250w) 6-8pin connection on them, you do NOT need to worry about PCIe bus supplemental power.

If you have only 18a(216watts) on your 12v rails, and the remaining 34watts of power has to come from the PCIe... (in a 250watt draw card)... Then you NEED powered risers.

If you are going to use more than 3 cards, you should have powered risers for more than two of them. (eg, 4 cards, 2 should have powered risers, or a powered-jumper in one slot.) {Again, if you are NOT using cheap china risers with thin wires, and you ARE able to supply the majority of power through the 6-8pin connections.}

Remember, we are running in "constant operation", which these cards and MoBo's were NOT designed to do... (Not the consumer stuff).

Cheapest way is to buy something you will use later...

Drywall is cheap. Doors can be too, free on the side of the road.

Box it in at the window. You only have to exhaust the top-hot air, shoot it out far enough that it does not get sucked back in by the intake. (The intake being cold, can be a simple cloth-sheet sock. (Cheap bedspread which you duct-tape to form a long-ass filter-sock. Throw a rock inside and let it hang down.)

Even cardboard would work, if you want to go ghetto-diy.

Exhaust top, cool at bottom, fans directing it all up, or just mixing it. The heat will rise, the cool will fall. (Less if you are just mixing it.)

When you are done, just disassemble it. Someone will buy the unused sheet-rock dry-wall. I am sure you will find a new purpose for it. (Life-time supply of driveway and sidewalk chalk for the kids!)

I have a request...

A few actually...

1: Can you have the title change to... (The part within the quotes only)
- "BitMinter 1.45 GHs (@3)" - {So this reflects the rate and running cores when minimized, in the taskbar.}

2: A true "regulation" tool...
- Not the interval, as that is real buggy when they are set without similar values, and has no real effect.
- Setting to 1 makes cards bounce all over, 5 causes strange card round-robin like effects, 10-20 seems normal, 30-50 is like performance-mode, above 100 it just stalls the computer with no real gains.
- Suggesting an "overall throttle", all or none. Just adding a delay between "completed works", delaying longer per interval. (Before getting next job, not after getting the job, which would just cause stales.)

3: A "cool-down" interval... Options for "round-robin", or "all cooling". With the round-robin option asking how many to cool at once.
- EG, cool-down for 5 min, after 3 hours of work, 2-cards at a time, round-robin. Would slowly reduce load per card, until two are resting, staying rested for x-minutes, then slowly starting again. (Slow = less card fatigue and cold-weld shearing and allows voltage/wattage to be less of a shock.) {pun intended}

4: Correct the display when set to "performance mode".. it shows the last card in the list, not the "total", as it once did.

5: Include an "advanced mode" for those of us who know how to use the additional settings. (Also, for the hopeful addition of litecoins, which will need those settings, per card.)

6: Keep up the good work. I love this program. Saved me a lot of headaches. (When I didn't know how to use cgminer correctly.)

I want to see the picture after the fire department visits him...

Why not just gut a window-ac unit, and put the PC in there? Or, um... move the PC to the window, exhausting directly out there?