Can u specific the hardware and the mhas per rig please ?
10 rigs with 5 5830s at 965/300. 1560Mh/s per rig. 15.6Gh/s total. 3 rigs with 5 6950s at 860/500. 1750Mh/s per rig. 5.25Gh/s total. 4 rigs with 4 5970s at 825/500. 2210Mh/s per rig. 8.84Gh/s total. 29.6Gh not including a lowly 5850 hashing away at my house. |
|
|
|
Whoa! How much power is that rack pulling?
My guess is just short of 16kw just by doing a card count and estimate. My spreadsheet says around 17Kw but it looks like i over estimated according to my power bill. You should post that at the Bitcoin Mining Rig Builders Competition , we are giving away 25 BTC!
Already have  -> http://btcnetwork.com/coinconnect/pg/pages/view/1302/gigavps-mining-operation  Man, that's crazy dense and clean setup. Is that 5 5850's on each rig? I'd really like a close up Nice furniture movers on the feet lol Here are some more pics. Below is a closer look at the rack.  Here are some rigs getting burnt in before being moved over to the DC.  |
|
|
|
|
I've entered. Let's see what happens...
|
|
|
|
|
Well, it looks like I found my answer. Seems that my home internet connection has something to do with it. When I moved the boxes to the DC, the stale rate moved back to where all of the other miners were at. Weird.
|
|
|
|
I've been waiting a while to post these, but here we go. It's still not complete, but it good enough shape to share.   |
|
|
|
In case anyone was wondering what I am cooling, here are some pics. |
|
|
|
I need an extra pair of hands. I am stressed to the max, and have more hardware coming in soon. I pay in bitcoin. LAX area in So-Cal. Multiple rigs need to be reworked and I just lost 1 to 2 6 card rigs. Assessing damage. Me physical, shaken and smoking.
Been there, done that. Popped 4 5970s, 3 6950s and 3 5830s is the span of two days. Suck it up, push through it, and you'll be fine. Here is a pic of my "close to" finished product to help you get on with it. |
|
|
|
It would be a fair amount of work. If you are offering to pay, sure.
What is your price? |
|
|
|
no.. i just make the window bigger.
I have tried this. My monitors are not large enough. I have 28" monitors. Is this something that can be added into mgpumon? |
|
|
|
Is there anyway to scroll down in mgpumon?  When I pull it up, there are too many computers and cards on the screen so they fall below the bottom of the display. |
|
|
|
If we find two more blocks here decently quick, we will have filled up all the spaces on the "Recently Found Blocks" page.
This excites me in ways I can't explain.
I need something to hump.
Bkkcoins, your leg. Bring it to me.
I am just as excited, but maybe in a little be less "i have to hump something" way. BurningToad, it would seem that the original server is not able to hand out work quickly enough and is causing miner idles. I have switched everything to server2 but this is probably effecting others without them knowing. |
|
|
|
So uh, we found a couple blocks... This has been be the best 1 day ever for ars. 16 blocks found on the 21st and the day even isn't over yet. Hopefully this will push us back into payouts every couple of hours with a big positive buffer. |
|
|
|
Well whats the results
Ok. Here's what I ended up doing. 1. Added the foam insulation onto the ceiling. This leveled out the temps in the building, cutting down the high temp by about 7 degrees. We had a high inside of 103 a couple of days and how the highest the temps have been is 96. This also cut out the lower temps we were experiencing in the evening. Instead of dropping to 78, we now drop to around 85. 2. I have redirected the entire 3 tons of AC to point directly at the equipment through a single 16" air duct. This has helped lower the temps at the GPU cores without lowering the temp in the building itself. I also added an LG portable AC unit in the office area which cut out 315 sqft of space that the 3 ton AC handled before. Next I am going to take the 16" air duct and attach smaller ducts to direct cold air directly onto the computers which should help even more. I am also going to build an exhaust manifold like in the video but instead of venting it to the outside, it will send air into the return of the 3 ton AC unit. This way I will not lose all of the conditioned air the AC worked to condition and the hot air will not mix with the cooler air in the building. The exhaust manifold will also be double sided, basically sandwiched between two shelves. The manifold will basically become the hot isle in the hot/cold isle diagrams above. |
|
|
|
Sorry for the lack of activity lately. Been a bit busy, and a bit less motivated thanks to BTC price ;P You may also be able to blame a friend of mine for getting me into league of legends....
Still busy the next few days, some people from out of town staying at my place. However, maybe near the end of this week, or on the weekend, I want to do a few pool improvements.
Items I want to look at first:
1) Server 2 stability. Pushpool still randomly decides to use ~100% cpu on Server2 occasionally. Haven't been able to figure out why. Doesn't seem load related. A simple pushpool restart fixes it most of the time. If I can't figure out the issue, I need to write a script to detect the condition quickly, and restart it automatically, or see if I can just consolidate back to using one server ( would have to improve Server 1 performance somehow.) The main server has been extremely stable in comparison, but a day or two ago we hit ~950 GH/s max, so I don't think just one server will cut it.
2) Fix block stats issues (missing blocks I have to add manually, and other such things.) I started this work as part of the PPLNS stuff, so I need to get back to it and implement some fixes. The main issues currently are related to finding blocks in very quick succession, or issues with blocks from server 2.
Seems like multiple pools are using poolseverj now and they seem to think it is more efficient for a large number of miners. Maybe this is worth looking into to keep a single server? https://bitcointalk.org/index.php?topic=33142.0 |
|
|
|
Deepbit's for suckers.
I would agree. Ars Bitcoin's uptime is way up there and the payouts for smpps have been the best of any pool I have tried. |
|
|
|
Before drawing any conclusions, make sure you
take your measurements on a long enough time
scale/number of shares. IMO, watching things for
a couple of hours isn't enough to draw conclusions.
I have tested the machines for multiple days and on different pools. Same results. Could this be something with the aggression setting in phoenix? |
|
|
|
|
I have a couple of machines running 3 5970s and am getting higher than usual stale rates. The stale rate for these machines seems to stay around .95% while other boxes with 5830s or 6950s are running rates from .2% to .35%. Has anyone else seen this and do you have an explanation for this?
|
|
|
|
A suggestion, if I may: Make the automatic payment feature payout the whole balance.
Currently, if I hit my threshold of say 3 BTC around noon, I don't get paid until 2300 (which is fine; I don't care about the delay). But, at that time I only get paid 3 BTC. This leaves behind a balance of whatever I made between noon and 2300, which complicates determining my hourly or 24-hour earnings.
I agree. There is quite a bit of work that needs to be done to tighten up the interface to the pool and to give miners the information that they need. |
|
|
|
|
Questions:
- Are you sure these spaces can handle the extra heat load? I am running 20.5Gh in 800sqft and it requires a full 3 tons of AC blowing *directly* over the rack to keep the computers cool.
- What is your strategy for monitoring servers, environmental conditions and networking? Do these rooms have connections to the internet?
I have talked to data centers and they will only let you do so many kilowatts per sqft without extra cooling equipment. This is a big can of worms unless you are trained in this kind of thing or are willing to learn the hard way.
I only say this from direct experience.
|
|
|
|
|
Show Posts
