I doubt based on keeping costs low for us that the internals of miners are designed with highly technical thermal flows using wind tunnel and scientific measurements, if at the very least the thermal imaging camera can pin point some hot spots, it maybe a useful tool to help increase flow around these and keep internals cooler and avoid temperature build up.
The real point is to keep temperatures right near the chips from building up causing shutdown or premature failures. So lets hope that the heat sinks and fans do a great job, I suspect after a few iterations we'll get there but that doesn't stop anyone here doing some improvement mods and sharing these personal experiences with there fellow miners to help one another and to pass this along to AMT to improve the product for us..
The real point is to keep temperatures right near the chips from building up causing shutdown or premature failures. So lets hope that the heat sinks and fans do a great job, I suspect after a few iterations we'll get there but that doesn't stop anyone here doing some improvement mods and sharing these personal experiences with there fellow miners to help one another and to pass this along to AMT to improve the product for us..
My goal with the thermal imager and kill-a-watt is to ensure "100%+ Constant-duty", can be achieved with minimal consumption. Using only the kill-a-watt, you can gague the efficiency of an efficient or inefficient system, but with the addition of the thermal-imager, to isolate thermal-runoff, you can tune that efficiency or inefficiency into pure efficiency.
I plan to run my machines at about 80% "suggested limits", on a constant-duty. However, I plan to do a full comparison of tuned results, with the thermals as only part of the comparison. If a voltage regulator or capacitor or transformer is being "stressed" I will know it before it pops. (In the end, 600-Watts = 600-Watts, no matter where it is generated. However, inability to remove 600-Watts will impact production. Having the ability to remove 800-Watts gives you more head-room for expansion or "production variance reduction". EG, constant hash-rates not being thwarted by heat build-up throttling. More heat = more resistance = slower speeds or burn-out.)
Not that I believe the design or layout will not be sufficient for "normal operation". However, there is always something that can be done to make things better. Always...
Even if it is something as lame as flipping the fan upside-down to suck air evenly through the fins, as opposed to attempting to push it through unevenly with dead-spots, using an open-air fan improperly, trying to use it to deliver static-pressure or volume through a setup that creates static-volume-resistance. (That is the standard CPU cooling trick for better instant cooling. That is what happens when you let programmers design physical things. They just don't understand real-world physics. But they may have a great understanding of CG-Physics. lol.) Hell, they still design all aluminum cooling-fins parallel to one another... Aluminum reflects heat... If you reflect it at another reflective surface, that just creates a cascade of heat reabsorption. Eg, face one heater to another heater, both get twice as hot. It is almost as redundant as liquid cooling with water, on a constant duty cycle operation. It only stores heat and makes it take longer to get rid-of, consuming more power (generating more heat), to exhaust it through an even more horrible radiator design. But that is beyond this topic. lol. (Water cooling is only good if you get intermediate bursts of heat, acting like a buffer, keeping a constant warmth and "in time" allowing the heat to be cooled until the next burst of heat comes. In a constant-duty cycle, it is just an added source of heat generation to exhaust and more power to consume/waste.)
In the end, I am sure this will only provide an interesting read for the many, and be of great importance to those wanting to "push the system to near limits", as safe as possible. Or, for the frugal... This will give them a nice baseline to reduce costs and pull a few extra months from the end-of-life of the machine, when the operating costs begin to dip into operating costs.
BTW, for the "calculators"... Best way to calculate is by the "expected value at the time you intend to cash-in", not at the "prices now". If you believe it will get up to $6000/BTC by next year, calculate with that as the "price". Since you will be holding all coins until they reach that price. That is the "value" you will get when you cash-in. That is like making $6000/BTC the whole time. You only calculate with "now prices" if you plan to cash-out immediately after making every BTC. (Only idiots do that.) That is why diff matters more than value, for "now".
Yes, $6000 is not only realistic for next year, but it is also an honestly low realistic estimate. If you go by the trend, estimates are in the high $16,000/BTC and linear averages are around $12,000/BTC respectively. (Just wait until the "income-tax returns" start getting used to buy BTC again this year. Another 2-months, we could see $6000 hit, but chances are, it will only go up to about $2800 for a final season peak. With lows in the $600's to $800's for the summer-time drought, before the next seasons investment spikes.)
It will also be a bonus if AMT can use any of this information to "make the units better", before they are delivered.
Don't forget, I can also see the heat generated and dissipated through all the circuit-boards internal hidden PCB traces too. One bad trace being beefed-up in a next generation design could aid everyone. For the electronically inclined, it would just require adding an extra wire. (Again, that is not for normal operators, that would be for those wanting to push these near limits.)
Good thoughts, at least reading through this things like bang for the buck, ROI, cost vs benefit come to mind, engineering electrical/mechanical excellence without careful thought of the ROI short and longer term, is it really worth the effort when battling cost and devaluation, I believe you allude to the balance of both, still it needs some careful thought and maybe some trial error to figure it out. Quick wins like switching fan direction/placement even type of fans/blades or switching out the power regulars for more costly but more efficient ones for instance, may have an impact but it's always going to get into cost/benefit and diminishing returns on any mods.
I was hoping that with help of growing community, that modders get to play and figure out what works or doesn't without too much cost associated with doing the mods and sharing those results, passing on that knowledge so we don't waste time over again on things that don't work.
I believe liquid cooling unless on par with or close to air based methods, won't be viable but that at the very least pushes liquid cooling solutions to get cheaper and to find practical application when used with miners, note: I avoid heat dispassion and the medium being used, etc. Net efficiency and cost of that are really interesting things, well unless your hard core engineer first and miner second?
Interesting to see where this heads with some speculative volatility in the markets still. The meeting yesterday by the US DFS was nonetheless encouraging for the future of BTC, US markets and exchanges though (http://www.totalwebcasting.com/view/?id=nysdfs). I recommend a listen for anyone interested in US Dept. of Finance Services current stance on crypto currencies.
