DeathAndTaxes
Donator
Legendary
Offline
Activity: 1218
Merit: 1079
Gerald Davis
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 05, 2011, 08:23:44 PM Last edit: October 05, 2011, 08:34:59 PM by DeathAndTaxes |
|
It would not (sic) be way more cheaper, easy and effective a "full immersion" oil cooling system? I am tempted to try a ghetto one if I find an almost free aquarium and car radiator. That is a solution but it isn't as cheap or easy as you think. 1) You need to use mineral oil unless you want a maintenance nightmare and gallons of mineral oil isn't cheap (high grade stuff is $15/gallon and the cheap stuff has enough impurities to conduct electricity). 2) Mineral oil has lower thermal conductivity than water so you will need a lot of it and will need to circulate it faster than water. 3) Mineral oil has higher viscoscity so that combined with #2 mean you will need a powerful pump and that isn't cheap. 4) You will need to ensure you design proper baffles so you get even circulation otherwise hotspots will develop. 5) You will need to filter the oil. Contaminants will increase electrical conductivity and without maintenance will eventually cause a short. If you have a free aquarium and car radiator you can reduce the cost but getting car radiator clean enough to avoid contaminating your oil isn't that easy. It doesn't take much contamination to get to a point where you get a short. Remember when the immersion fluid becomes conductive you likely are looking at a total loss. MB + powersupply + RAM + multiple GPU = $$$$. It is something I would like to try someday but it isn't that cheap or easy. Awesomely low? My 5870 mines at 970/330 at stock voltage on air cooling.
The advantage of water cooling is reduced noise, more control over the heat, and higher density. 8 of my rigs are open air in the garage where they create a nice bit of noise. Getting a single card to high clock is much easier than then getting 6 GPU inside a closed case to a high clock. The 3x5970 is my workstation in my office inside a closed case with less noise than a single small sized oscillating room fan. The radiator is mounted in the window to dump heat outside and when it gets colder I will dump heat inside instead. Water cooling does all that with 60C core temps (hopefully closer to 40C if I get a larger radiator). Nothing beats the cost of air cooling. Period. If you simply want max power at lowest possible price regardless of temps, noise, or space then you should air cool. Some of us for reasons due to wives, kids, pets, limited space, interest in water cooling look for a more elegant solution to the extreme thermal loads caused by 24/7 mining. Still I don't want to derail this thead so I will leave it at that.
|
|
|
|
conspirosphere.tk
Legendary
Offline
Activity: 2352
Merit: 1064
Bitcoin is antisemitic
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 05, 2011, 08:44:08 PM |
|
1) Yes. I wonder if some kind of motor oil would work too. It should be way cheaper than vaseline oil. 2) ? 3) Maybe not. There are tolerances. I use to run my diesel car on a mix of diesel and vegoil. I shouldn't and could't, according the instruction manual, but I already made 50,000 Km on it. Moreover, oil viscosity is a function of its temperature (the higher the temp, the less its viscosity). 4) I would leave the fans in. I've seen that they work in the oil, even if at much slower speed. Then there is the pump. 5) I think that you refer to distilled water (where microscopic amounts of contaminants make it counductive). I never heard of this problem about oil cooling. That is a solution but it isn't as cheap or easy as you think. 1) You need to use mineral oil unless you want a maintenance nightmare and gallons of mineral oil isn't cheap. 2) Mineral oil has lower thermal conductivity than water so you will need a lot of it. 3) Mineral oil has higher viscoscity so you will need a powerful pump and that isn't cheap 4) You will need to ensure you design proper baffles so you get even circulation otherwise hotspots will develop 5) You will need to filter the oil. Contaminants will increase electrical conductivity and without maintenance will eventually cause a short.
If you have a free aquarium and car radiator you can reduce the cost but getting car radiator clean enough to avoid contaminating your oil isn't that easy. High grade mineral oil runs ~$15 a gallon and you are going to need at least 10 gallons more likely 20 gallons (remember not just the aquarium but the radiator and cooling lines also).
|
|
|
|
DeathAndTaxes
Donator
Legendary
Offline
Activity: 1218
Merit: 1079
Gerald Davis
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 05, 2011, 08:57:40 PM |
|
Arrg I can't help myself. Come on even the lowest viscosity oil has much higher viscosity than water. Simple test. Pour some 0W-20 oil into a bowl. Now pour some water. The difference in viscosity is immediately obvious even to the eye. While oil thins at high temps we are looking for high temps. Oil @ 50C is pretty much the same as oil at room temp. Thermal conductivity is how efficiently the substance transfers heat. Water has very high thermal conductivity (there are very few low cost substances w/ higher thermal conductivity). There is a reason why your car doesn't use motor oil in the radiator. Does mean it won't work it simply means you need more fluid to remove the same amount of thermal energy. More fluid means not only more fluid to purchase but higher flow rate which means a more powerful pump. To get the kind of flow you need with a fluid as thick as even low viscosity oil is a serious pump. Contaminants are contaminants. It is what makes a material w/ low conductivity have higher conductivity. It isn't something special about distilled water. Contaminants can raise conductivity of the fluid (they can also lower it but we don't care about that). If the conductivity gets high enough you will have current flow between wires & circuits through the fluid = game over. That can happen in any fluid. The only way to protect against it is to use a) sealed clean system, b) replace fluid periodically, c) filter the fluid or more likely some combination. The nice thing about sealed liquid cooling is the conductivity of the coolant is the "last line of defense". If there is no leak then no circuit is exposed so the level of conductivity is immaterial. In immersion cooling there the circuits start out exposed by design. Look I am not saying it is impossible just that once you start getting down to details it isn't as easy or cheap as if first seems. Ok I promise now I am done. To everyone else I return you to your interrupted thread already in progress. Was nice to see that 3 hour block after that 9 day monster.
|
|
|
|
P4man
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 05, 2011, 09:20:33 PM |
|
Its a good topic, no need to change it ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) I read that motor oil would work, but Ive never seen anyone try it. To save money, the trick apparently is buy mineral oil thats used as a horse laxative (no joke) You can buy that in quantities of 5 liter and is not too expensive apparently. All the other problems remain though, just like d&t explained. If you look at experiments with this, you will see temperatures really are not good, even for desktop usage. If you want to do any non trivial overclocking, the only way to make that remotely possible would be chilling the oil. But obviously that costs a pretty penny, in hardware and electricity again. As for keeping the fans turning; how long do you think they will last if they encounter so much resistance, they barely spin at 10 rpm? anyway, dont let us stop you from trying. By all means, make pictures. Its not because its a bad or unpractical idea that its no fun, and Id love to see it ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif)
|
|
|
|
FalconFour
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 06, 2011, 12:50:00 AM |
|
Gahaha... OK, wow, what'd I start? ![Cheesy](https://bitcointalk.org/Smileys/default/cheesy.gif) Nah, the card's fried. I only found out about the PC on my way out the door to work today, I quickly hauled the parts of the system inside (PC, cables, monitor) and set them on the living room floor, then yanked out the card and it was drenched... I waved the card around to dry it off, and could hear water hitting everything in its path. For that much water (rainwater, no less, filled with whoknowswhat crap) to be in/on something electronic, AND for the computer to have turned itself off, something seriously bad had to have happened. Took the card to work, made sure it was completely dry, and tried it in another PC. The computer acted like there was no card in there at all... it just booted from the integrated graphics. Windows started, and it saw nothing either. Tried several things, cleaning the board of any residue or corrosion, looking over the whole board, and there were a few chips that looked like their leads had corroded somehow, they look like they were just eaten away. Very strange. No chips looked "popped" or anything, just eaten away a bit, somehow. When I got home, I took a second look at the PC itself. I don't dare plug it in yet - it was still totally soaking wet. Earlier, when I brought it inside, I held it with the open side downward to "drain" it... water just poured out. Probably not good for the carpet, but fuck it. But, yeah, it was soaked. And it still is kinda wet... For a computer to actually shut off... think about it. If it BSODs, it either reboots or hangs at the BSOD. If it hangs, it just hangs. If it overheats, that's about the only time I've known a computer to shut down. This thing was *off*, and with that kind of rain, I couldn't imagine it actually running. Yeah, it's pretty safe to say the card's toast. Probably even the computer, too. :/ edit: Can't keep a good falcon down. ![Cool](https://bitcointalk.org/Smileys/default/cool.gif) ![](https://ip.bitcointalk.org/?u=http%3A%2F%2Fhostfile.org%2Fthumbs%2FL_P1020564.jpg.png&t=665&c=pTIXO-SRbkkvJg)
|
feed the bird: 187CXEVzakbzcANsyhpAAoF2k6KJsc55P1 (BTC) / LiRzzXnwamFCHoNnWqEkZk9HknRmjNT7nU (LTC)
|
|
|
DrHaribo (OP)
Legendary
Offline
Activity: 2730
Merit: 1034
Needs more jiggawatts
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 06, 2011, 07:51:44 AM |
|
Seems noone could reach the server between 07:01 and 07:19 UTC. Sorry about that. Looks like a network outage at the datacenter. Everything running smoothly again since 07:19.
I have to move to a different datacenter anyway, so I'll do that soon. Will let you guys know ahead of time, as that will mean a little down time.
|
|
|
|
rTech
Sr. Member
![*](https://bitcointalk.org/Themes/custom1/images/star.gif) ![*](https://bitcointalk.org/Themes/custom1/images/star.gif) ![*](https://bitcointalk.org/Themes/custom1/images/star.gif)
Offline
Activity: 305
Merit: 250
Trust but confirm!
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 06, 2011, 09:39:23 AM Last edit: October 06, 2011, 10:10:23 AM by rTech |
|
With latest beta: Factory OC, XFX nvidia geforce 260 gtx (216 model) : Worksize 256 / Break interval 75 = 46/47mhash without any own overclocks. Previously i had Guiminer/OpenCL with 43/44 mhash. So this 2-3mhash is quite positive for me ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) I know this nvidia card uses more electricity than it produces bitcents, but every hobby cost some money.. DrHaribo, thanks for making great tool, latest beta is rockin'
|
|
|
|
Turbor
Legendary
Offline
Activity: 1022
Merit: 1000
BitMinter
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 06, 2011, 06:46:30 PM |
|
I know this nvidia card uses more electricity than it produces bitcents, but every hobby cost some money
Here in Switzerland it's almost impossible to mine with profit. But money and profit is not everything. I mine because i like the idea of bitcoin ![Cool](https://bitcointalk.org/Smileys/default/cool.gif) .
|
|
|
|
SgtMoth
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 06, 2011, 09:48:46 PM |
|
With latest beta:
I know this nvidia card uses more electricity than it produces bitcents, but every hobby cost some money
try telling my wife that! I keep telling her, "well, I could go sit in the bar and drink all day!"
|
|
|
|
FalconFour
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 06, 2011, 10:31:25 PM |
|
With your nVidia and the CPU-usage bug of OpenCL-over-CUDA, you should create a power profile and limit your CPU to 50%, which actually cuts the power consumption anywhere from 50% to 10% (50-90% savings). Really should try it. Meanwhile, I love Amazon. I'm so sorry about the problem you had with your item. I've placed a new order for you at no charge. Here are the details: Shipping Speed: One-Day Shipping Estimated Delivery Date: October 7 To return the defective item, first click the following link to print your return mailing label: ... When you return the item, follow the instructions listed on the mailing label page.
|
feed the bird: 187CXEVzakbzcANsyhpAAoF2k6KJsc55P1 (BTC) / LiRzzXnwamFCHoNnWqEkZk9HknRmjNT7nU (LTC)
|
|
|
DeathAndTaxes
Donator
Legendary
Offline
Activity: 1218
Merit: 1079
Gerald Davis
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 07, 2011, 01:26:58 AM Last edit: October 08, 2011, 08:08:04 PM by DeathAndTaxes |
|
Oops sorry. No that would cook the card, any card. A typo, it should have been 1.1625V although I have it down to 1.148V now.
|
|
|
|
P4man
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 07, 2011, 04:53:23 PM |
|
ONe of my 5850s just died :/. I guess my trouble keeping its temperature under control was related to it. Shame I just bought VRM cooler and what not for it. Anyone got a reference 5850 or 5870 for sale?
|
|
|
|
DrHaribo (OP)
Legendary
Offline
Activity: 2730
Merit: 1034
Needs more jiggawatts
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 07, 2011, 06:06:50 PM |
|
DrHaribo, thanks for making great tool, latest beta is rockin'
Thank you, that's very kind of you, sir ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) Sorry to hear of your misfortunes, FalconFour and P4man. Looks like your box did contain awesome, though, Falcon ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif)
|
|
|
|
FalconFour
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 08, 2011, 03:28:17 AM |
|
Well, the Amazonian gods did smile down upon my desk this morning, and upon it landed a subtle gift from the gods, ![](https://ip.bitcointalk.org/?u=http%3A%2F%2Fhostfile.org%2Fthumbs%2F322430_2377250584625_1051431182_.jpg.png&t=665&c=8eSpSmOPj2jdnQ) ... a replacement. Free! Including return postage label, too. Sheesh. I almost feel bad. ... * looks at water-filled PC * ... Nap, don't feel bad anymore.
|
feed the bird: 187CXEVzakbzcANsyhpAAoF2k6KJsc55P1 (BTC) / LiRzzXnwamFCHoNnWqEkZk9HknRmjNT7nU (LTC)
|
|
|
SgtMoth
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 08, 2011, 08:00:30 PM |
|
is there anyway we can transfer our balance out faster, seems like it takes 10mins before it goes.
|
|
|
|
DrHaribo (OP)
Legendary
Offline
Activity: 2730
Merit: 1034
Needs more jiggawatts
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 09, 2011, 09:25:39 AM |
|
is there anyway we can transfer our balance out faster, seems like it takes 10mins before it goes.
Payment (from BitMinter to your own bitcoin wallet) is once per hour, currently 10 minutes past the hour. Perhaps it would be less confusing if it was exactly on the hour? ![Wink](https://bitcointalk.org/Smileys/default/wink.gif) I am planning to add an "instant payout" button as well. But that will probably have to cost something to use. Alternatively, when I get a donation system implemented, if you donate over a certain percentage it could be free. I don't want to pay more of those 0.0005 BTC transaction fees than I have to. I'll need some income to balance out those expenses. One thing I could do, probably without incurring more transaction fees than today, is to run payments also immediately when a block we made is confirmed. So you get your bitcoins immediately when they are added to your account, if the payout threshold is low enough. And if you, at any time, change the threshold from above to below the current balance, you have to wait 60 minutes in the worst case. I think I will also lower the default payout threshold set for new users from 1 BTC to 0.1 BTC or maybe even 0.01 BTC. I don't really like handling other people's money longer than I have to.
|
|
|
|
FalconFour
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 09, 2011, 11:35:36 AM |
|
Maybe you should look into figuring out how sites like dailybitcoins.org "collect a bunch of transactions together to minimize cost" (iirc) so they can send a lot of people bits and pieces of 0.005 BTC by the hundreds... I think it's possible to chunk things down in increments, and have, say, a group of 1 BTC payouts that each go to address ......, address ...., address ....., etc, then a 0.75 BTC chunk to address...., ....., ....., ....., and a 0.5 BTC chunk to ....., ...., ..., and a 0.25 BTC chunk to ....., ....., ...., which would be "payments in 0.25 BTC increments" and you would only (theoretically) have to pay 3 or 4 transaction fees for your custom transactions to multiple recipients. Then we could also lower the automatic payment threshold* since it would really nearly eliminate the transaction fees. * rather, eliminate it, since it would be "in 0.25 BTC increments when each block is minted". Since people accumulate balances only when blocks are minted anyway ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif) That's just based on the assumption that what sites like dailybitcoins.org do is possibly applicable to our scenario ![Smiley](https://bitcointalk.org/Smileys/default/smiley.gif)
|
feed the bird: 187CXEVzakbzcANsyhpAAoF2k6KJsc55P1 (BTC) / LiRzzXnwamFCHoNnWqEkZk9HknRmjNT7nU (LTC)
|
|
|
DrHaribo (OP)
Legendary
Offline
Activity: 2730
Merit: 1034
Needs more jiggawatts
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 09, 2011, 11:59:42 AM |
|
I put these changes in right now, for higher and faster payments: - Pay out income from transaction fees - the pool now only keeps rounding errors. This goes for block #24 (next block we create) and up.
- Use "half down" rounding. The last "bit dust" of income cannot always be split fairly among users. I think this is the most fair way without the pool losing money. 0.000000006 is rounded up, but 0.000000005 is rounded down, just like ...4. This leaves a tiny amount of income which is kept by the pool.
- Make an extra payment run immediately after a block generated by the pool is confirmed
- Run the scheduled hourly payments exactly on the hour
|
|
|
|
DrHaribo (OP)
Legendary
Offline
Activity: 2730
Merit: 1034
Needs more jiggawatts
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 09, 2011, 12:25:12 PM |
|
"collect a bunch of transactions together to minimize cost"
Yep, we already do that. That's why it is better to run payments every hour (and now on every confirmed block). If you run payments when a user clicks a button, then you can't batch transactions together and you run the risk of paying lots and lots of transaction fees. This is why a pool needs to be a little smart about an "instant payout" button. An option would be to completely remove the minimum payout threshold and just pay out every user's complete balance every time a block is confirmed and money comes in. Upside: - You get your money in your wallet sooner
- The servers become a less profitable target for hackers. I can sleep well at night and not be responsible for (as much of) your money.
Downside: - The transaction history in your wallet (the bitcoin program or other wallet) becomes full of tiny transactions
- Negative for the bitcoin system in general: when you next send money from your wallet, the transaction will consist of many sources, each of a tiny amount of bitcoins. This means bitcoin has to handle more large (in number of bytes) transactions. We will contribute to that 600+ megabyte blk0001.dat file growing on everyone's harddisk.
![Tongue](https://bitcointalk.org/Smileys/default/tongue.gif)
Any thoughts?
|
|
|
|
DeathAndTaxes
Donator
Legendary
Offline
Activity: 1218
Merit: 1079
Gerald Davis
|
![](https://bitcointalk.org/Themes/custom1/images/post/xx.gif) |
October 09, 2011, 03:47:00 PM |
|
DrHaribo,
"An option would be to completely remove the minimum payout threshold and just pay out every user's complete balance every time a block is confirmed and money comes in."
I like this option. Makes it simple. If the pool gets very large you may need to change it but currently I would assume (you can check records) most miners get > 0.1 per block reducing the number of tiny transactions.
A hybrid could work. All miners are paid INSTANTLY (once block is confirmed) as long as the balance is >= 0.1 BTC. Miners that earned less than 0.1 on the block will have reward rolled over till next block.
BTW. Keep up the good work. As a show of confidence I moved another rig (+ 2GH/s) to bitminter.
|
|
|
|
|