Heat is the enemy of most electronics including FPGA's. Overclock them 20% and increase their heat output by 50% and you are reducing their life. Properly cooled and stock speeds/specs and they should last more then ten years.
Stock clock for a miner, or stock clock as meant by chip mfg
|
|
|
Yeah, it's quite soon to get reliable figures, but a general direction is possible to calculate. With enough of them out there, it will be possible to calculate quite early on quite precisely. That's how manufacturers get the MTBF readings - not by testing a huge amount of the product for years upon years.
|
|
|
I'm trying to figure out what should be the next mining related purchases i do, but one question remains: What is the life expectancy of FPGA based miners? Has people have them fail? How long were they running before they failed? and how many do you got which are still running, and how long has each been running? That data would help me calculate the annual failure rate, from which one can derive the expected average lifetime Versus GPUs it seems people expect GPUs to have an average lifetime of 3years, correct me if i'm wrong.
|
|
|
Dude, obviously you do not have a clue about the underlying physics going on. FYI, i've even made blocks myself BEFORE watercooling became anywhere near so popular that you could buy anything watercooling related. Back then i were already using compressors to drive -34C coolant to the CPU. Did you know that you can get by just fine with low pressure, low volume pump (cheapest 12V you can find) if your block is properly massive and TDP is below 100W?
Btw, "stop-leak" and other stuff you put in to stop leaks is the stupidest idea ever.
Do you know what is the coolant of choice of extreme high output racing engines which actually partake in sanctioned races? Now i'm talking about engines like 2L Inline 4 boosted to some 4bar, Cummins 5-8L diesel boosted to 7+bar? Nope, not distilled water.
TAP WATER, with maybe some water wetter. And these machines need to be capable of removing heat in the multiple MEGAWATT range. Are you telling me water is not sufficient to remove your couple hundred watts of heat?
As for the dirt accumulation, i've only seen it once, and it was with off the shelf system with which came the "water additive", very small amount to be used, i was shocked to see that accumulation and i somehow doubt that would happen if proper glycol was being used. I suspect the additive that came with it was just some coloring.
The thing is with water if you add a little bit of heat, but not much, and circulate it constantly, it WILL have bacterial growth, that's why you need to add enough glycol.
Still, WTF is dead water? Other than a sea water phenomenon known to few sailors, which might just be a myth as likely.
|
|
|
Copper should not be used with Aluminum it causes Galvanic corrosion, don't use Aluminum with anything actually. I tried EK coolant the first time, I was worried about leaks. About 3 months down the road my cpu block got plugged with little fibers, was a major pain figuring out what caused the flow to drop. I'm not sure if it was the fluid or not. It definitely leaves residue on everything that it touches. Now I use distilled water, along with silver coils and a few drops of dead water. Seems to be working great. Here is a simple guide: http://www.overclock.net/t/913181/water-cooling-guide-for-noobs#post11984918Yet cars have been utilizing both metals mixed in the same system, in the same cooling system for decades upon decades. Those work wonderfully, and the coolant channels in cylinder heads are WAY smaller than your CPU block has, often a diameter of just 5mm in many parts! So now you are using something which is a strange physical phenomenon known by some sailors to happen in sea? Coolant, proper glycol, is to be used aswell to minimize the build up caused by bacteria etc. Some glycols themselves will build up, and still causes dirt to apper in the coolant, hence, the coolant ought to be swapped annually, maybe accompanied by a flushing the system. Aluminium does oxidize, but it doesn't matter at all in a cooling system, especially if you use red glycol which has properties to protect from that. Might be a worthwhile read: http://www.gewater.com/handbook/cooling_water_systems/ch_32_closed.jspSpeaks in length about corrosion in cooling systems. All you really need is a bottle of cheap glycol and tap water, and annually change the coolant.
|
|
|
You know, I spent to much time researching mixtures. One site started that 3/4 Ethealyne-glycol 1/4 distilled h20. Is there a mix you would recommend for effiency ?
Useless to spent all that time researching, pretty much use any you feel comfortable, or which gives the nicest color in your senses Forget distilled water, useless waste of money & time for cooling purposes. Just add at least 1%, but don't exceed 50%. I'm not a fan of watercooling computers, too much work to maintain and hassle in regular desktops. For servers it makes sense tho. I recommend the red glycol (i keep forgetting the names lol) because it has the highest boiling point and best protective for modern engines, but for computer watercooling any automotive glycol will work just as fine.
|
|
|
I smell homebuilt BFL PSU's to be constructed. I bet it's going to be really annoying in a cluster if you have so many little external power bricks.. It'd be cool to see a "BFL Cluster Powerbrick" with like.. 5-10 12v lines coming out of it, Could prolly save alot of money, And on shipping alone too I smell a business opportunity for them here, supplying very high efficiency PSUs with modular cabling for clusters
|
|
|
Copper + Aluminium in the same loop is not a fatal problem. Those are combined in automotive applications for decades. Radiators in standard cars used to be copper! Sometimes copper head gasket is used etc. Newer cars tend to use aluminium cores for radiators due to their lower price and weight.
Just add high quality coolant or in other words: Glycol for protection. The red stuff is highest quality. You need this in any case, unless you want your loop to gather all kinds of dirt, bacteria etc. It also prevents corrosion.
In this utilization even 1:20 mixture is probably more than sufficient and it's cheap.
|
|
|
Really cool work, for what i understand this already offers around 30% more per cycle? That's simply awesome. If i were a miner with a significant any scale and investment into FPGAs i would definitely throw some BTC to your direction, especially if that meant i get unlimited access to the bitstream
|
|
|
LOL! Bitcoin pyramid in your sig was funny This has been always happening, getting far lower results than expected. It's odd. Oh well, got to see after upgrade to BAMT on all my rigs. Are pool hoppers hoping to gain bigger gain for their Mhash or why is that even happening? P2Pool subsidy? maybe i ought to setup P2Pool, took a glance and it seems a little bit more involved. To me it's starting to sound like P2Pool definitively gives the best pay, while being fair to other miners out there?
|
|
|
I mine on Deepbit with around 1.8 G/hash.
Some days I get 1.5 BTC. Others I get .95 BTC. Variance is a part of mining.
Also depends if you're doing PPS or proportional. If you're doing PPS, then the only thing that varies is your own individual hash rate.
If you're doing proportional, then your payout will be affected not only by variance in your individual hash rate, but also by the pool's hash rate.
My average is .95BTC, variance is small: 0.93-0.98 And i'm not really using deepbit at all anymore EDIT: I'm expecting variance, but the variance currently is a constant -20% vs. expected average.
|
|
|
I use Slush's pool almost solely, but the rewards has been far worse than expected always. Before i didn't really care, just few Ghash/s worth for fun running. Currently i have on Slush 1.77Ghash worth of capacity and the yield is 0.95BTC/day while expected is ~1.2BTC/day. This has always been true of Slush's pool, and deepbit seems to be even far more worse. All my systems are properly cooled, have good PSUs, only mild OC if any and efficiency (stales etc.) travels at around 97.8 to 99.2%. I use Linuxcoin, and even after updating to latest Phoenix which increased hash rate ~7.5% i saw actually a decrease on reported hash rate @ Slush's pool, and it made absolutely no effect on yield. Generally my GPUs are running under 100% utilization, somewhat below the average expected by the hardware comparison charts, except for some cards, ie. 6770s i'm running at 242Mhash/s average! (890Mhz, stock 850, temps ~65c, stock air cooler) About to change to BAMT (Linuxcoin has always sucked, but it was most convenient to use last summer) and see if that changes things. But anyone know why my returns are roughly 20-21% below the expected? Direct Ghash/s comparison to pool total should yield 1.214BTC/day. According to: http://www.alloscomp.com/bitcoin/calculator.php should make 1.19BTC/day and after taking into account pool luck should still be 1.1186BTC/day. Also someone mentioned that 1Ghash/s stays at around 4$ per day value on average over the long term, How the hell does one achieve that kind of results? :O
|
|
|
Thanks for the awesome photos! Exactly what i were hoping, big and detailed images Now i can finally belief this ain't vaporware long con, and put in a order for a few ...
|
|
|
The data sheet says "200 W power supply". No, you cannot run 32 Spartan6-150 FPGAs off a 200 W power supply while mining. Mining consumes about 8 to 9 W per FPGA. Didn't check in that much details. Simply swap the PSU. Then again, cooling probably hasn't been made to match
|
|
|
Well mine just arrived Pictures please so i can put in my order!
|
|
|
I'll just leave this here as eye candy for you guys (was just sent this by BFL via email, pic of their burn-in/testing rack right now) hrrr, i still have a hard time in believing BFL is not a long con.
|
|
|
Pretty cool. Each board has two 10 amp @ 240 VAC relays, and they are SPDT which means they don't have to be on all the time. Just turn the relays on for a few seconds when you want to kill the rig. The little ethernet board works directly with SNMP for you hardcore monitoring folks. Some models have 15A relays. Not that it matters that much, as usually you have 16A fuse for a regular 230-240V line and 2.3kW is already plenty for single rig. I bought a 16relay IP version for testing. Too cheap to pass up! BTW, are you actually a (the?) guy that works at pulsedmedia.com? Great seedboxes there. Yeah, i am
|
|
|
Pretty cool. Each board has two 10 amp @ 240 VAC relays, and they are SPDT which means they don't have to be on all the time. Just turn the relays on for a few seconds when you want to kill the rig. The little ethernet board works directly with SNMP for you hardcore monitoring folks. Some models have 15A relays. Not that it matters that much, as usually you have 16A fuse for a regular 230-240V line and 2.3kW is already plenty for single rig. I bought a 16relay IP version for testing. Too cheap to pass up!
|
|
|
|