I think this is going to be one of those 'sounds better in theory' kind of things. Still give it a go and report back with pics of your crazy creation, of course! I suspect that the restriction of flow from the pipe head alone will cancel out most of the hot buoyant air effect you are trying to exploit. I would confidently say that it WILL have some airflow but the question becomes "is it sufficient?".

I am not sure about the H-chimney design, I think the two 90 degree bends the flow would have to make would limit the momentum and flow rate of the rising air. I would almost bet BTC that a single straight tube of constant diameter would be better.

EDIT: I re-read your idea about putting the fan on it too. I was thinking you meant it would work passively. I suspect you will still find that having this contraption with the fan(s) on top will still be worse than just having the fan next to the heatsink as normal.

Where are all the process engineers / industrial chemists to model this scenario for us??? Do we have to bribe them with BTC lol?

Was this your motivation behind your idea by any chance? http://en.wikipedia.org/wiki/Solar_updraft_tower

There is a usually unknown but still inherent limit to how much current a motherboard can - in total - pass to its PCI-e connectors.

A full-power-rated PCI-e full length slot by international standard can supply at least 75 watts to the card. This is not enough though of course for most good cards, so we have 6 and 8 pin molex to help supply more power.

The problem becomes, that if you are running a lot of cards then you don't know how much your motherboard can safely supply via PCI-e to all the cards. Maybe its just 75 watts? Maybe its 400? In any case it is a unknown and it is a risk.

The solution is to bypass some or all of the power connections via the PCI-e connector and instead use risers that have power connectors spliced on, so you can power it 'directly' from the PSU just like with the 6 and 8 pin molex connectors for the card.

This decreases risk of: random power transistors and capacitors and other parts on the motherboard failing due to overheating or worse, just exploding, and it arguably increases stability in the case that your motherboard can physically take the load but still supply insufficient current.

There are threads here about users who have charred components on their motherboards or melted/charred the main motherboard ATX connectors or cables.

https://bitcointalk.org/index.php?topic=161242.0

https://bitcointalk.org/index.php?topic=102890.0

It depends? There are people here with ASIC farms that have hundreds or maybe even thousands of gigahash/s (0.1 - 1.0 or more terahash/s) so I suppose they make look most of us look small. But then you could argue that mining is about improving the network security and the more distributed devices under control of different people the better.

User gigavps has the following pages showing stats of his farm(s):


He has at least a few hundred gigahash/s by the looks of it.

Sure that's degrees F, right?

By 'Delta' do you mean ambient<-->load temps? Just checking.

No. ASIC are good for any SHA-256 algorithm based coin. For example... PPC, Freicoin, and others.

Thats a big difference in temperatures between those cards, if they are the same cooler design I think there must be a thermal paste issue or a severe airflow problem somewhere. Perhaps the heatsink is clogged with dust more than the other one.

I don't run a string for cgminer I use the .exe and the config file

The cards I have run with the default settings that cgminer uses for the GPUs when it finds them the first time. Tweaking them usually makes it worse.

I have never had a fan fail, and I have run dozens of cards under heavy load even before bitcoins popularity boom running a 4870x2 farm for distributed computing. Usually when it is time to clean the fan of dust I apply some light machine oil ($2 at hardware store) to the bearings. On some cards you have to drill/dremel a small hole (0.5mm or less) into the front of the fan to access the bearing.

Hard drives: Benchmarking tests with HD Tune, Hard Disk Sentinel, or any of the countless different disk benchmarking/testing utilities. Alternatively you could get away with doing chkdsk C: /R which will take 9999 years to complete (and on any other drives other than C:)
GPU: Stress test with miners, FurMark, or OCCT GPU tests (fullscreen)
CPUs: OCCT or any Prime number computing software

Keep in mind that when you are running a server or blade or similar custom engineered hardware with the covers off you have changed the air flow configuration from the design specifications and parts may be overheating and you might not even realize and/or they might not have any thermal monitoring. Especially when testing under load conditions like those above...

You have 3 options usually with solar, depending on what is available in your location, in order from Best to Worst case in terms of profitability and ROI:

Best: You feed all of your solar output directly back into the grid without consuming any of it yourself on your consumption meter. Your government or municipality pays you for the kWh at a rate higher than your consumption rate. Eg for a while some friends in Australia had $0.56/kWh out while their consumption rate was $0.20 at the time.

Meh: You have some kind of smart metering system or load sharing inverters installed that has fully replaced your old meter or works in conjunction with it such that the solar energy you produce is 'subtracted' from your regular consumption. You do not really sell your energy to the grid unless you are consuming very little power, in which case you ARE paid for the energy you return to the grid.

Worst: As above in 'meh' except your excess solar energy production is NOT sold back to the grid, usually because your local power grid is already saturated with solar installations (a huge problem in Australia at the moment) and you could not get approval to connect your installation to the grid. *This could actually be a win for some bitcoin miners though, because you could time your setup to mine during sunlit hours consuming the excess solar power. Essentially 'free excess power for mining'. You would DEFINITELY want some very nice batteries on the DC side or some UPS units on the AC side to pull this off though.

Worst, with bonus points: You think you can run your GPU mining rigs directly with DC (!!!) output from the panels, OR, from the inverter but with no fallover from the grid or energy storage, so if you have a droop in voltage from the panels your inverter shuts down and your mining stops. I suppose that with something like raspberry pi and some very clever programming/fiddling or even a windows laptop that had some kind of triggering/monitoring from the real world (eg. arduino board) you could pull this off with ASICs or FPGA, set to turn on and off quickly as power comes as goes from the panels but it would most likely be a huge pain in the ass to do it on a large scale. The investment and ongoing replacement costs of batteries for DC side storage or for in UPS (for the worse scenarios) is substantial and should be taken into account when doing any calculations.

This is a complex argument that depends on so many factors. For example, the quality of the components that the card vendor used when the card was assembled following the reference designs, and again is different if they've changed the reference design. To make it really complex, the quality of the wafer the die is made from, and then the quality of the fabrication itself of the GPU core and some of the other components will have various impacts on how many of the shipped products fail. There are always going to be unavoidable elemental contaminants in the silicon and there will be some physical flaws with the die after fabrication, in every chip. The QA testing of silicon and the final fabricated processor is its own little expensive industry that you could study for 10 years at university and get a PhD in and yet still not grasp all of what is going on.

This is continued onto all of the hundreds of subcomponents of the card, even the tiny SMD resistors and capacitors that are all over the cards.

Of course though the parts that fail the most will be the ones under the most load and operating as close to, at, or over (eg. when overclocking) their rated design limits.

Critically, for our purposes, we should follow that having substantially lower temperatures has a great increase in the lifespan of a card. Of course, the card you are mining with now could die 10 seconds from when you read this, but statistically, if you had n number of cards, less will fail when running at lower temps.

It's important to realise that with things like GPU cores, MOSFETs and other transistors, and capacitors is that their mean lifespan does NOT vary linearly with increasing high temperatures. If you drew a graph it would look almost exponential, going exponential as you reach temperatures in the 80-120 deg C range for most parts you'll find on a graphics card. I suppose it is useful that AMD now has hard limits to the temps that cores/VRMs can reach, which are monitored by the hardware itself on the card and will force a driver crash, a hard lock or reboot the computer if they are exceeded. This stops noobs from buying a 7990 for $1000+ RRP and running the cores at 150 deg C and wondering why they now have a paperweight (but it's also useful if say, the thermal compound loses contact).

tl;dr - cooler temps = long life, but failures always can happen due to unrealistic user expectations, poor user handling of the card, design limitations of the cards and its part, cost versus product quality considerations, and the inherent material/manufacturing flaws in all microprocessor and integrated circuit devices.

For what it's worth I've abused countless 4870x2 cards back in the day at 90-100 deg C temps for YEARS of continuous operation (BOINC projects eg mostly MilkyWay@Home GPU), then quite a few 5970's, including one I've kept which is now watercooled and going strong for many months of 24/7 mining now. I have a 6870 that has also done at least 8 months of 24/7. I have a 5750 that is more lower-end of the spectrum that I've made run so hot it smelt like burning PCB (scrypt mining, probably memory controller or VRM overheat) but it still works flawlessly on SHA-256 crypto. I've also had (and later sold / given away to family or friends) several other 4xxx, 5xxx and 6xxx cards some of which ran distributed GPU computing tasks for months or years of 24/7... in fact I don't think I've ever had a card fail permanently.

It will vary from country to country, but I know for a fact that if you do your own mains wiring of a fixed structure where I live (Australia) you are opening up a huge can of worms. For example, if you have a fire and your $20k of mining equipment burns to the ground, or worse, your whole house/unit etc. when your insurance company conducts a fire investigation (or worse if there was a death and the authorities conduct their own investigation) then there is a very substantial chance they will realise you've done your own non-qualified electrical work and you will get zero payout from your insurer. An event like that could easily land you in debt for life, or in jail for some years.

Run the .exe directly (or make a shortcut to it)

Enter your particulars as it guides you to setup the miner.

Press "S" --> "W" and then "ENTER" to save the default named config file.

Then when you want to mine just run the exe and it will load the config file.

You can either modify the config file while the miner is closed, or you can use the GUI prompts change settings (you have to do the save step though).

If you want to put one of the command line arguments into the config file... just insert it as a new line following the formatting of the others in there.

Software: Please try cgminer with default settings, it should run those cards quite well as well as give you some monitoring of temps. Use GPU-Z simultaneously to monitor the other sensor data for the cards. I recommend on a 6870 no more than 80 degrees C on the core temps. In my experience the VRM temps of the cards that have reference coolers are very low, which is nice, but still worth checking.

Hardware: What you are better off doing is running them at their stock settings (revert back to original BIOS if you've changed it) and then doing some error checking stress testing with a program like FurMark or OCCT GPU tests.

Unfortunately neither of these programs will make your life easy for you because the 2nd card will be hard to test under load even if you have a monitor connected to it as well as your primary display device. My recommendation would be to test the primary card and then swap the positions of the two cards and then retest the other card. It might be possible to test both cards simultaneously with these programs (eg multiple instances) but in my experience it was a headache.

Because it runs for 25+ minutes and then crashes it to me indicates a thermal instability issue ie. it's getting so hot it crashes. Also it's possible you have a problem with your powersupply though 2x 6870's won't even put 400 watts load on it... should be able to handle that.

Lastly... what fan speed are your running??? You'll need to force them to run at fan speeds that are higher than normal, either by using a program like Sapphire Trixx or one of the other vendors alternatives (they work on other vendors cards though). You could also try to setup cgminer or other mining software to control the fanspeed either statically or dynamically. Another option is to edit the BIOS of the card so it has a custom fan profile, although it's safer to use the custom fan profile in a program like Sapphire Trixx than edit BIOS if you are not up to the challenge.

An example of thermal instability on my 6870:

(900 Core, 300 Ram, 1.174 v) with air cooler (reference design): 70+ deg C core temps at 100% load --> Immediate crashes with any overclock more than just +15 MHz

versus

(900 Core, 300 Ram, 1.174 v) with watercooling: 43 to 45 deg C core temps at 100% load --> Easily overclocks to 990 Core and stable for days, not sure how much higher it can get (yet!)

Unless you have a source of air <20 deg C you will want to run those cards at 90 to 100 % fan speed at all times to keep the VRM temps as low as possible (monitor with GPU-Z)

My suggestion would be to:

-Disable fan changes with mining software
and
-Use Sapphire Trixx to set fan speeds OR modify each GPU's BIOS with Radeon Bios Editor (RBE) to give a fan profile that is much more vacuum-cleaner like.

No pics of your setup in your previous posts?! Outrage! 

Because the width of the USB connector is standardised, I can easily work it out by counting pixels. I measured the width of some standard USB connectors around my place with calipers and they were all 12.00 +/- 0.02 mm

From your supplied image, 90 pixels = 12.00 mm, thus 7.5 pixels per mm

Also from your pic:

X = 449 px = 60.0 mm
Y = 188 px = 25.1 mm
Z = 299 px = 39.9 mm

I have measured leniently the dimensions, they may be about +/- 0.1 mm, favoring the larger for your convenience so that it fits wherever you plan it to.

I would try the card in another computer (friends computer? spare computer?) just make sure it has a suitable powersupply and you remove the old card's drivers first.

Make sure your using original/stable bios and not overclocking past manufacturers stock speeds for the test.

If it still fails the tests then I would suggest doing an RMA or if it is no longer under warranty try underclocking it until it is stable and/or remove the heatsink fan assembly and check for any signs of damage or poor thermal contact (dried or cracked thermal transfer material over cores, or more likely the VRM MOSFETS.) Also you can get a magnifying glass and carefully inspect all the tiny SMD capacitors and resistors that are all over both sides of the cards for damage (from mechanically breaking them off) or in the odd case of them exploding (which is a problem on some cards, eg. overclocked 67xx and 57xx cards in particular). If you find a damaged component it is not impossible to repair it yourself. If you do not have the skills/knowledge to do these kinds of things then I'd suggest pawning the card off as a 'faulty / parts' auction on eBay, you should get at least a few $$$ for it.

Edit: Also inspect the PCI-e (motherboard) connector pins for damage. If they appear dirty, clean them (google). Check the PCI-e power connectors and PSU cable plugs for signs of burning, if some of the pins/cables have charred then current will most likely not pass through it sufficiently and the card will be unstable, this kind of thing can happen - not surprisingly - on high-power cards like 5970, 6990, 7990 etc.

PCB begins to thermally decompose (it stinks, too) at about 250-300 degrees Celsius. Of course, other components and the chips will fail by then.

Sent Canary a PM regarding additional 1 BTC payment for international post cost. Sorry for the confusion, I will learn to read one day.

Trillium; 2; 2.2418; 18qKgev5cdaBF2ni5BjBV464XkhfjB16u3