Topic: [ANN] Spondoolies-Tech - carrier grade, data center ready mining rigs - page 219. (Read 1260290 times)

Do you live in a winter climate?   do you have a home?  can you get cold air to flow from outside to the intake on the miners.

 open a window put box fans in the window blow in cold air to the intake of the miners then open a second window let the hot air blow out.

you would need 2 windows and a room with a shut door to do this build a pair of vents to direct the air.  you say you have 20 of these  

so they stack 5 high  make a 4 by 5 stack of them  direct the cold air in and the hot air out.

Just got off the phone with DHL regarding my GB SP20.
The miners are at their facility since 7:58 this am, but they won't deliver until tomorrow (Tue) because tomorrow is the "official" delivery day.
Did you ever hear this load of...? Sitting on a package for >24 hours because they can. I am sure that Spondoolies is unaware of such happenings.

Ok I'll give up... 

For me it's not hobby... It was a hobby with a Jupiter and some S1-s.

Maybe would be practical to buy one more unit for sure. But I have a lot of high cfm fans and other parts laying around....

But at this point I give up. You don't know a lot of details so you can't make a judgement what am i do and why. But trust me, I have the reasons.

So back to the original question.

Any ideas to improve the cooling? 

On average it takes ~2^32 (no exactly, it's actually 65535/2^48) hashes to find a diff1 share, and D diff1 shares to find a block. IE, it takes D*2^32 hashes to find a block.

Therefore, your expected chance of finding a block per second is HR/(D*2^32), and thus your expected reward is BR*HR/(D*2^32). 86400 seconds in a day, so expected BTC/day is BR*HR*86400/(D*2^32), or 0.00002011626*BR*HR/D. I just multiplied that by 10^9 to give the result in GH/s.

this is what I did 9 sp20E's   and 1 s-5 all underclocked a lot works best for me. and I have to tear the whole thing down in may as my rates go to summer price of 16 cents.

In the cold wither I pay 13 cents but I get a lot of heat which lowers my  13 cents to 10 cents.

So extra units with  underclock works best for me.

Yeah, I usually have 117-122V, but even with 110V I thought that 15A outlets (typical for US) are rated for constant 13A, hence 13X110=1440W

curious what you pay for power. I understand the aspect of clocking hardware to its limit if its profitable, but if you need to make >$10 in modifications to gain 50GH/60W per unit, plus have slightly more PSUs/TH it could be more practical to buy an extra SP20 for that extra 1.65TH/$500. If its a hobby to do modifications though thats another story

Whoa!  More than I expected, thank you again—and for the calculus refresher.  I'm paying more attention now than I ever did in high school (fun?!). One last more question, if I may: in my previous calcs, I used an empirical 0.012745 BTC/Th/day - you arrived there via block reward, difficulty, and 20116.26.  Where does 20116.26 come from? Representing luck?

Edit: Gh/Th

Sorry @knlondike but you misunderstood me.

You calculate with a fix price of the machine.

Calculate cost and profit.

I'm not speeking about the only 50ghs but when the weather gets warmer this will be 100ghs (from 1680 ghs) or more.

At 20 pieces it's 2000ghs....

And i have room much more than 20....

I hope you unerstand why i'm trying to find some solution.....

That should work if you have closer to 120V. If you really have only 110V, you'll need to limit the SP20 to more like 700W.

The constants are an amalgamation of things. You profit is really revenue - cost.
Revenue is BTC/day (20116.26 * BR / D) times the price (P)
Cost is the cost of energy (W * (1kW/1000W) * 24hrs/day * C ($/kWh)) in dollars
What I did was graph 2GOOD's efficiency numbers to get an exponential equation, Power(W) = 127*exp(0.001368351*Hashrate(in GH/s)). Sub in equation in place of power, and then your profit is simply a factor of your hashrate (if the other variable things are treated as constants each day).
Profit = 20116.26*(BR*HR/D)*P-127.0061827629*EXP(0.001368351*HR)/1000*24*C
Just take the derivative of the profit equation, and hashrate goes away in the first term, and the constant in the exponent comes out of the exponent equation.
dProfit/dHashrate = 20116.26*(BR/D)*P-0.00417093689255513*C*EXP(0.001368351*HR)
The constant 0.00417093689255513 is just combining all the constants in the second term, = 127.0061827629 * 0.001368351 * 24 /1000

To find the maximum, set the derivative to 0 and solve for the hashrate. It's a classic min/max problem that you might get to teach applications of derivatives. Calculus isn't just for fun in highschool.

Interesting, I got 1587Gh for 0.11/kwh electricity.
I can't use it at this setting, though, because I would have to run the fan at 80-100, which would not be optimal inside the house, especially for three units.
Another consideration: i am thinking of running one SP20 and one S5 on each 110V/15A circuit (this way each circuit would produce ~2.5Th, which is awesome). I thought that it is possible if SP20 is limited to ~800-820W+~600W for S5.

Thank you for this.  Much more elegant, and useful, than my profit table I cobbled together at incremental power costs versus a single over-clock and under-clock setting.  Can you tell me about the constants?  Thanks again.

SP Tech, so what is new in FW 2.5.61 other than the changes with the setup features that were introduced in a previous version? I have been running the current version for the last six hours and it seems to be working better than 2.5.58 or 2.5.59 which I see were taken off. As it is rather painless to upgrade, and these are my first SP units, I have been pushing the upgrades as they came in. The 2.5.58 did cause the miner to cycle cgminer on its own, something 2.5.59 did not do I am happy to say. I will continue to run with 2.5.61 and report back if there are an anomalies.

Take a look at the formula I posted earlier. With $320/BTC, $0.15/kWh and current difficulty, your maximum profit would actually be at 1370GH/s @ 827W, making revenue of $5.59 a day with a power cost of $2.98, for a profit of $2.60/day.

Obviously individual preferences for prioritizing BTC production, etc might change that, or if your efficiency numbers vary much from 2GOOD's, but at least it gives an idea that for a lot of people trying to run at 1700W might not make a bunch of sense.

^This is really making a big issue over pulling an extra 50GH out of a unit - a value of less than $18. voiding warranty or buying a second $20 fan and spending the time to install it is not worth it.

At the peak speeds, any improvements you make come at the cost of about 1-1.2W/GH. Running at 1500GH uses approx 210W less than running at 1700GH. Unless you have very cheap power it will likely make sense by summertime to throttle back the hardware to improve the overall efficiency - or have heat problems.

I personally have 3 arriving this week that I hope to achieve ~1400GH/800W so that the fan can be kept at a reasonable volume. At $0.15/kwh it doesnt make much sense for me to try and get another 300GH/400W out of the unit, as the added hashrate gains are entirely spent on added power draw. Same reason as Ive throttled my SP10 units from 1400Gh to 1260GH to save 200W

The problem is that you can't fit them on the heatsinks.

I will check how many space is avaliable. But if the space is enough, first have to stabilize the heatsinks and I think this is the harder part.

The heatsink is fixed with 2 screws and they moving on the asics.

To improve the cooling, high cfm small fans needed but now I can't see a chance to fit them on the heatsinks. (Maybe work something out after the holiday season but I see a litle chance to solve it.)

Edit: I will be more clever after New Year's eve. Now i'm just writing by heart from under the christmas tree  .

Sure it's possible.

BUT

I don't want to do that more than 20 machines... (Maybe more if I change my other gear)

I don't need 1725ghs, my problem is not here. I made my calculations with 1670-1680ghs.

1650 is still no problem, not a big diffrence but if you calculate with 20 machines (round number) it's 600ghs less.

I'm testing the gear, when I make 13-22C intake the hasrate is between 1630-1680.

With 22-26C inlet I got only 1590-1600Ghs. (At summer I'll have about 28-32C)

It's winter here now, but with this machines I will have some problem at summer. (I won't able to do 10C for sure.  )

To turn back to the calculation: 20*80ghs= 1600ghs less due to overheating.

Don't misunderstood me, I'm not whining about the machines, I like them and the engineering behind them.

I will break even until summer.

But the profit will be less then estimated.

On the other hand.

I'm running a good ammount (I don't write huge ammount because I think these days it's the petahash range) of an other manufacturer's machines.

They were running in 40C at the summer and they worked rock solid as advertised. (I had to improve the cooling meanwhile but it's an other story.)

So I'm looking a DOABLE fix cooling solution for the machines. Some push pull configs, custom heatsinks, or everything else what somebody tried out.

I'm working on this me myself too but I don't want to invent the wheel if somebody have a good idea.

yeah you want 1700gh fans at 100 and air temp at 10c with a solid 1200 want psu.

Now  if you really are truly desperate for that 1700 mark  build an open case and use a fan on each heat sink.  make sure the air is 10c.

I can do it but my power cost makes it stupid for me to do.

lay the 1st board then the controller then the second board                    
 
1 board                                                  2 board
_________                                           ________
 HS HS           ----------controller--------    HS HS
_______                                              ________
  

 ^^                                                             ^^
fan fan                                                     fan fan



the above lay out  would allow for much better cooling.  you void your warranty but have the satisfaction of 1725gh at 1250 watts

No room for it..... 

I will cool them with ice or something....

Tried push pull with an about 240cfm fan on the front, I got 4 C better front temp but it's not a high improvement...

The asics on the back  are still hot...