Topic: Avalon 6 Derping at 30C ambiant? -> Evap Cooling - page 2. (Read 2907 times)

Actually there are 2 covers in the bathroom. One on the ceiling and one on the bottom near the water supply lines for the bath taps.

The one on the bottom seems to be blowing cold air in at the moment (its windy outside).

That would probably be a poor design? (If its "Cold air") I mean we're in Canada and it get cold, so you'd want those 1 way flap to allow air out but not winter in, because those gets pretty damn cold ^_^"

At some point i was able to pull a bit of air, from i assume surrounding appartements during summer, however. So it was 20C~ air, not Cold, but it wasn't too bad either. (Had both windows in the living room to output ASIC heat)

Mostly though it was a HUGE draft from my bed room if i ever dared to open that window.

I have the same cover, but it seems to push cold air inside, not suck warm air out.

Its because of the bathroom vents things, it seem to connect apartments together, for instance if the lady upstairs smoked in her bath, it would smell like cigarette in my bathroom. And those are connected to exhausts ports. It's probably easier to have common vent than each apartment having their own. We also have cloth washing rooms which are also connected to exhaust ports.

There's not supposed to be but there's a small square manhole they cut above my bath to access the bath above for some repair in the past and if i leave that open, there's a significant (heat) draft going from my living room to it. It always seem to suck a lot of air from my apartment, unless i go severely negative pressure. Since i generate a lot of heat i'm pretty much always positive pressure so its easier to plan that way.

I'm moving there next Monday and before i start running test its going to be hard to say. Most likely if possible i will want a setup with the room closed and seal on its own circuit, hopefully with evap cooling, so that the noise coming out of it, with acoustic panels will be nil.

heheh, it's all knowledge that I paid a much higher price to learn than I should have!    The reality is the math is pretty straight forward once you get it, and there's no way around it... Trying to get creative just ended up costing me money, and usually to little effect.


Yeah, depending on the age of the building is how sealed it's going to be - the older it is, the more leaky you'll find it to be.  In Arizona, where practically every house is only 10-15 years old, they're sealed up pretty well, but even then I would say there's probably 500 cfm of leakage (typically coming in around doors and windows, attic egress, and sometimes bathroom vents (and other things that may not have decent shutters on them).

Cracking a window or doing other things to help the intake side of your air path works great - and you'll also get a sense for what kind of pressure differential you're getting.  If you open the window, and you immediately feel it sucking the air in, then you know it's being starved to intake.  Also keep in mind that opening up multiple windows may not make any difference, because once the demand is met, it's not like it uses more.  And air is much like water in that it's demand is usually met by the source with the least resistance, so if the window right next to it is open and can supply demand, it might not even suck in air from other rooms.

If you really care, then I would highly recommend picking up a manometer - this is the one you see in all my builds:

http://amzn.to/1Tm1Y8P

Cheap and easy to setup, and no needing to guess about what the difference between supply and demand is.


GPU miners would be a bitch to handle - at least the open-air ones I've seen people using... The design to use the big box fan is cool, but it also means that there's really no way to control the airflow without doing some major ducting around the whole thing, which in turn would make it a pain to work on when things go wrong.

I've actually been toying around with the idea of designing an open-air case similar to the ones I've seen - most are using 80/20 or similar extrusion, which I work with all the time and have used for all my builds.  The big things I would do is make acrylic inserts that could alternatively be used to control the airflow... I'd also design it to use server PSU's instead of these lame PC ones - it blows my mind that all these GPU rigs are using off-the shelf PSU's.  I recently got one of optimizer's 4k boards to play with, and it's pretty sweet - plus he actually uses the sync pin, so it matches the voltage between the two PSU.  That really got me thinking about building a GPU rig around it, as the 4k could easily run a 6 GPU miner (a single 2k could probably run one, depending on the cards).

Any of you have any ideas how to exhaust the GPUs ?

Firstly, MarkAz is spot on with all points he mentioned. Kudo's

As to what your setup will do... for a start, depends on where the thermostat is followed by how makeup air is coming in. I believe most Codes now require some sort of provision to prevent high negative pressure in occupied areas but in most cases a 'leaky' building takes care of that

You said 'apartments' did you mean AC for physical apartments (units) are linked together or just the rooms in your apartment? If possible, short-circuit the airflow into the room by opening another window in the room away from the exhaust or a room/area close to the miner room and leave their doors open...

heheh, while I'm not in Saudi Arabia, I am in Arizona, and for at least one day last year we were the hottest place on the planet - and I can assure you, it's ALWAYS better to vent exhaust outside (except in really special conditions).    The big advantage we have is that it's a desert climate, so evap is an option and works quite well - you can get a 20-30f temp shift depending on your configuration.  I even have a high tech evap called a direct/indirect evap, that actually passes the evap air through a second stage of cooling in a water sealed membrane, and it's possible to achieve below dew bulb cooling with evap (not much, but any is amazing).  Either way, even with relatively inexpensive power here (~6-7c), AC isn't an option.

You're right to ask the question about where the air is coming from, but in my experience with most miners, you can address it in two ways - the first is lower the ambient temperature.  So the cooler you can make the space, the less air you have to move.  AC would be an example, but as I said, you'd need 4T of cooling running 24x7 to cover 15kw of mining.  The second way is just moving air - if you have move air through the device fast enough, then in many cases it doesn't matter if the air is 100f.  I'll just use a super simple example that's not really accurate, but you'll get the idea - if your miner moves 100cfm, and introduces 20f of temperature shift (so intake temp is 80f, exhaust is 100f), if you speed the fan up to move 200cfm, the temp shift would drop to 10f (not entirely true, as these things aren't linear, but close enough).  Of course, there's a limit as to how much air you're going to be able to move through a given space though, and that's where static pressure (SP) comes into play.  The other thing to consider if you were not exhausting air is the impact of the exhaust air in the closed system... With AC, it's probably less of a factor, but if you were trying to do a kind of close evap system, the humidity would increase with each pass through the system, and the cooling would decrease until you'd ultimately end up just making your room into some nasty high humidity sauna.

So, I would calculate the heat load of the room, in this case 15kw, the volume of the room, and then the amount of CFM it would take to exchange all the air in the room based on the exhaust size.  If intake or exhaust becomes a limiting factor, then you need to treat the air in some way in order to make up the difference (AC, evap, liquid cooling, etc).

To put it in a more concrete example, in our mine with a ton of S5's and Terminator's last year, we typically saw temperatures 100f+, and only in a couple instances did we end up needing to underclock the machines to avoid the chance of thermal failures.  But at the same time, we had custom enclosures to all the miners, plenums to control the exhaust air, and enough air handling to cycle all of the rooms air in about 15 seconds.  We actually didn't even have evap until the last part of the year, and once we introduced that it made a huge difference in the temps in the room, but almost no impact in the operation of the mine.  I suspect it would have just saved us the couple days we had to underclock.  One nice side benefit of switching to 100% evap air is that it REALLY cuts down on dust, as the evap membranes work as really effective filters.

Unfortunately I didn't get to run the A6's or S7 in any real world high temp situations, but I did do some testing with them and they performed as I expected - except for the issues that VirosaGITS is having with the controller.  The other challenge that the A6's have in terms of high heat is that it's much more difficult (or at least it was) to control their usage - underclocking them didn't seems to control their temps nearly as easily as the S5's or S7's did.  I suspect this gets back to the same issue with stale air and the controller, but never looked into it further.

Well i'm doing 15kW atm and its easy since the air go strait out and in Canada its cold outside most of the year. Only sometimes in the summer, during heat waves, does the temp reach 30-40C outside. Which is what i'm preparing for.

The air used by the ASIC, dumped strait outside would be taken from the room, which would balance the pressure from the apartment, which has Vents that tie apartments together and also to the outside.

Mostly, the air that will be drawn into the room will be from my apartment which will be A/C cooled for my own living confort. Its hard to say the impact it will have.

I'm just trying to offset the heat generated on hot days with a combination of venting the exhaust out (which is definitively hotter than outdoor temp)... maybe evap cooling on dry days and A/C on bad days.

While I am sure it must be better to exhaust hot air to some extent, my question is always: What is the source of the air that's going INTO the miner and then ducted to the outside? While you are cooling off air that's at say 100F (exhaust), you are obviously going to have to draw outside air, cool it down from what ever it is to say 75F. If the outside air temp is roughly the same temp as the exhaust air temp, then it seems like a wash, doesn't it?

I clearly didn't do thermodynamics in college decades ago, but if the outside air temp is really hot (think Saudi Arabia), I could imagine that it MIGHT be better not to exhaust, wouldn't it?

I would also think that 10KW (or more) in an apartment would be a lot to deal with. I can't imagine going and buying 8-10 blow dryers and turning them on full blast 24/7 in any apartment I lived in. But I wasn't trying to mine Bitcoins then either.

Just a few random thoughts, from a non HVAC guy.

Its a bit more complicated than that. I'm planning on dumping the heat from the ASIC like that (A6, S5), but for GPU rigs, its not quite as doable unless i put them in some sort of tents, but then they would be so big i might just dump all of the room's air instead.

Those blower, i will keep that as an option, can they run quiet? Because well, the goal is to keep everything at 30-40dB which is easy to do with GPU, doable with ASIC underclocked and with not too noisy A/C.

At the moment i'm thinking about a combination of Evap cooler in the top left, window is in the top right. And Air going out through the window through the piped ASICs, so humidity should not accumulate in the room, but heat through heat in the air and humidity.

It hard to math out, but on dry days, this could be cool enough to keep everything working happy. And for bad, humid days, i'd need to flip on A/C if i want to mine, i believe.

I take it you must not be paying for power, otherwise I'm not sure what kind of rate you're getting to make S5's profitable.  


Boy, 10-15k of cooling is going to be a tough ask - I built out a model in Excel to give me estimates for my builds, and that works out (on the high side) to be about 47k BTU/hr, and would translate to about 4 tons of cooling with about  ~2200 CFM of airflow needed.  These are all worst case numbers, but I always designed for 100% operation.

Personally I would do ducting from the exhaust-side of your miners to that window - if you saw any of my previous builds, I'm the king of exhaust management, and that is definitely the best bang for the buck.  If you can control and isolate your exhaust, then not only are the operating temps in your environment much easier to maintain, but you aren't fighting against yourself as the miners keep generating more heat.  If you can construct a smooth enough channel, then the fans in the miners will be enough to move the air and there's no additional power hit.

In my builds, I had enough changes in the plenum to require assistance, and I would recommend looking into the squirrel cage fans for trucks - I bought several of these for a test, and they were exceptional:

http://www.spalautomotive.com/eng/products/view_centrifugal.aspx?id=007-A42-32D

I was paying about $40 each for them on eBay, and they have a ton of static pressure - this is really key in moving lots of air out of a little window.  They're also 12v so you can power them from the same PSU's you're using for your miners.    In your case you'll want to experiment with what speed to run them at (they're 3 speed), to get the correct blend of SP and CFM.

The cheapest way to build test plenums is just to go to home depot and buy 1/2" foam insulation sheets - you can cut these with a hot knife, and use aluminum tape to hole it together.  Unlike using a vinyl tube (like a dryer flex duct), these will generally restrict airflow substantially less and also will provide good insulation so the duct itself won't radiate heat back into the room.  A 4 foot x 8 feet sheet runs under $8 typically.

10-15kw in a small apartment with 1 small window in the summer is clearly nuts.

Well only the cost of the A/C need to be counted, the running cost will be 0. Or rather, the cost of running A/C is not being able to use that power to run a S5 or two instead.

The problem with where i'm moving is, the room the miners will be in is a medium bed room with only one window. The window is somewhat high but pretty thin so i can't even fit a 20 inch box fan there. During the cold seasons, so like 7-8 months a year this is not a problem, but during summer, i don't see a way other than using A/C or tons of inline fans dumping the air outside.

If you have a cooling solution for 10-15kW through a single, narrow window, i'm all ears. Using 2-3kW of that capacity for A/C seemed like the only option i know would work.

@Phil well running the controller outside is actually a pretty decent idea, i could also figure a way to lower the intake under 36-40C, its just that to be fully quiet (30dB) they had to be in the acoustic board box.

Another option would be to run them in another room but at 1k RPM fans, so very very underclocked but that sound kind of a waste of money. AC'ing them up and have the exhaust dumped outside with those vinyl tube seem like the way to go, so that the heat generated by the units do not have to be cooled by the A/C, the only loss would be the air from the cooled room being slowly dumped outside. Still better than 2kW being dumped strait in i believe.

That would be great - love to hear how it works.  I've used to before to manage heat, especially when reworking electronics, etc.  I would probably go from the front of the heatsink, to that first black line in your picture.  The only other thing that you might consider doing is replacing the metal standoff's with nylon ones, as those metal ones are also going to conduct heat right into the PCB.  That would be pretty localized, and the tape will probably make the biggest difference, but I'm always a fan of overkill.  

I could try the heat shield.



controller removed


point of interest  this controller has 4 jacks for 4 boards and 2 jack for fans
thus running four boards on one controller may work

BTH, I don't think using an AC in conjunction with it would be cost effective - even using evap, which is night and day more cost effective than AC moves the needle on power... And from my simple tests, it didn't seem to be necessary.  The big fail on their design was just not extending their grill up and over the controller and using part of that as air input.  I can get why they didn't, as right now the clean air flows almost exclusively over heat-generating parts/heatsinks - but I think they failed to think about the stale air getting stuck in the large open area above.

Running the controller outside of the case would be another interesting option (as philipma said), but trickier to pull off I think.  On the units that I did take apart, I don't recall there being a ton of extra cable length.  The other issue would be getting that ribbon cable outside of the case.  Depending on whether or not you care about how it looks, you could probably just use tin snips and make channels for then, and could easily drill 4 mounting holes so the PCB stat outside of the case instead of inside.  This would definitely solve the stale air issue - but I'm not sure if it would be better than venting inside the case, as in this configuration you're still leveraging the fan in the back of the unit to pull air over it.  Now you could obviously add a fan to the board also, but you're adding even more complexity, cost and power consumption to the build - so a couple holes in the front are still the best option I think.  

Ah, here's one more non-destructive thing you could consider doing first - what about adding Kapton tape on the head sinks directly under the controller, from the front to maybe 1" behind?  This is what I'm taking about:

http://amzn.to/1NDHlk8

The miner is like 6" across, so this roll would be enough to 30+ machines, but remove the controller than the screwdowns - apply the tape so it covers up the top of the heat sinks, from edge to edge, and then replace the screwdowns and put the PCB back in it.  The Kapton tape is a thermal insulator, so it should block a fair bit of the heat that's radiating up from the heatsink, and not really change any of the airflow characteristics, since there's only a small gap between the two, and you'll only be blocking a few inches of that.  This might be a quick/easy fix...

I also considered this  I have quite a few photos of this taken apart.

https://bitcointalksearch.org/topic/avalon6-safety-or-how-i-burnt-my-avalon6-1336377

I think you can pull the controller out and run it outside the units

If i don't find a buyer for them, i might do that, i'd have to see how exactly its placed, etc. Thanks for the idea. Considering what they are earning, plugging them in an A/C's exhaust for the summer might be a good idea to keep them running cool and quiet.

If you're so inclined, one thing I had considered doing when I was playing around with high temp issues was drilling a couple holes on the front right above the controller (probably wouldn't take too many, maybe 4 right across the front).  I think the problem has more to do with the hot air getting trapped in the area where the controller is, and just putting a bit of venting there would be enough to make sure it gets evac'd.  If you do consider doing it, make sure to fully disassemble it (pretty simple) so you don't get any metal filings on any of the PCB's.