I kept getting that until I put all the config options in one file, called "lightcoin.conf", and invoked Reaper as "reaper.exe lightcoin.conf".  Then it loaded the same file twice, but seemed perfectly happy about it and ran just fine.



Side note, on the off chance mtrlt peruses this thread...

I've poked a bit at the .cl file for Lightcoin support in Reaper, and although I have a few questions (like what the hell the gpu_thread_concurrency and lookup_gap "mean", aside from their role in memory block alignment), I have one "real" question...

Why does padcache (aka lookup in scrypt_core()) use workitem-major order (ignoring z) rather than y-major order which (for non-CL C code) would better preserve locality of reference?  I refer to the output of the coord() macro here.  I honestly don't know if it makes any difference in performance given the overall pattern of memory access, but I wondered if on some cards that serves to improve vector performance (by using the dot product rather than a straight mult)?

Okay, I don't care to get sucked back into this, but in the interest of extending an olive branch...


Ambient doesn't ever hit 80C under my scenario.  It peaks at "real" ambient plus 6.5F (3.6C).  So you have no worries about your PSU or VRMs or CPU or NB or SB or anything except the GPU cores themselves.  I don't point this out defensively, but as a positive, because...



Aside from the fact that I never claimed 110F (okay, a bit defensive), you can read my numbers on a more positive note relevant to your project.  I didn't prove it as "impossible", but rather, a function of the GPU cooling device itself.  Your idea of using a water cooler should nicely get around that limitation, easily buying you another 15-20C with which to play.  That puts you solidly in the feasibility range up to an ambient of 134.5F

Again, though, you make a good point about cost.  It still makes more sense to me to just throttle your mining rigs down over a certain temperature, than to go for a complex, expensive per-GPU cooling solution.

Okay, sick of this discussion, so I did the math.  Go ahead and carry on arguing with it, without me.

Repeating my earlier point, the average bedroom measures a mere 86ft^2, or roughly 600ft^3, Meaning a 2400CFM box fan will completely remove the air from the room every 15 seconds.  Two will then do the job every 7.5 seconds.

1 BTU will heat 55 ft^3 of air (at STP and H) by 1F.  20KW = 68290BTU/hr.

20KW will therefore heat 3755950 ft^3 per hour, or 62600 ft^3 per minute, or 7825 ft^3 per 7.5 seconds, by a single Fahrenheit degree.  Or, put more usefully, it will heat 600ft^3 by 13F every 7.5 seconds;  At 4800CFM (one full air-exchange per 7.5 seconds), that gives us a mean-over-time rise of 6.5F - Pretty damned close to my "feels 5F warmer" estimate from the woodstove.  Funny, that.

For the rest of this, I will refer to the Radeon 6990, the hottest single card (AFAIK) in common use.  Your "preferences" of running cooler than spec aside, a Radeon HD 6970 (and the 6990 just contains two downclocked 6970 Cayman chips) has a tMax of 194F.

Now we have an interesting result here - With the stock cooler and under full load, a 6990 will violate its own spec at ANY ambient over 73F.  This would make the naysayers appear correct...

But!  We've also demonstrated that two box fans will clear 20KW of heat from a room, with a 6.5F steady-state rise.  That reduces the problem to airflow at the ambient/heatsink interface.

Since we've limited this discussion to airflow, I'll limit the following to fan/heatsink combos only.  Using a halfway decent aftermarket cooler, you can drop the under-load temperature to below 160F (you can find similar numbers in dozens of articles on Tom's Hardware or AnandTech, if you like - For my exact source, I found http://www.kitguru.net/components/graphic-cards/zardon/arctic-cooling-accelero-twin-turbo-6990-cooler-review/5/ helpful in that it described the test conditions sufficiently well to extrapolate from them), for a plausible under-load delta-T of a mere 87F.

This, then, gives us a "feasibility" result of 107F at the intake, or a raw outside temperature of 100.5F.

Call that a win or a lose, as you like.  Where I live (if I didn't pay a fortune for electricity), it would put me over my "thermal budget" for two days in the past decade (no, DeathAndTaxes, not talking about winter, nor did I ever).  And I don't live in Sweden.   

A typical window fan moves 2000-2500CFM.

The average bedroom measures a mere 86ft^2, or roughly 600ft^3 - Meaning a $20 window box fan will completely remove the air from the room every 15 seconds.

A typical woodstove puts out 3.4KWH/Kg, or somewhere around 10-15KW.  Mine does closer to 15.

And put bluntly, if I put a box fan in the room with my stove, with a window opposite it open to allow air in - I would freeze to death in the winter.  And I don't speculate on this, I've done it when I "burned in" my stove.  Two years ago, in late autumn, at 60F outside, I fired that sucker up to a good 700F with a fan blowing out the nearest window to out-gas the enamel... And it didn't make a damned bit of difference in the ambient temperature of my house.  Standing directly between the stove and the window, you could feel perhaps a +5F difference.

So, if your box will run, at ambient, at full load without cooking itself, for a mere 15-30 seconds - Really, the outdoor temperature will suffice to cool it.  A crappy Lasko window fan (and two open windows in an area you can conveniently close off from the rest of the house) will do the job juuuuust fine.



phorensic... Look, I don't think we substantially disagree on this, and might even respect each other IRL.  But 15-20KW of misused gaming rigs still doesn't come anywhere near datacenter loads, even if they may have a higher spatial density.  If you really want to extend the point to "miners" pulling half a megawatt, I'll concede the point.

Then, as a data center designer, could you kindly explain to me what, exactly, fails to "scale up" when using outside ambient air to cool?  Just how high of a server density do you need before "outside" experiences a significant increase in temperature?




Yes, they have.  And you have apparently read them, because you specifically mention their impressive improvements on the boring ol' "swamp cooler".  So no doubt, you know all about Facebook's newest DC in Lulea, where they expect to need less than two weeks of supplemental active cooling per year.

Now compare the form factor and uptime demands of a hardcore miner against a Facebook datacenter - And try to tell me with a straight face that you don't see just the teensiest difference between a room full of mid-tower PCs loaded with GPUs that can go down for an hour or two in mid-afternoon on the hottest days of the year with no real harm done, vs row after row after freakin' row of 24core x 42U racks with a contractually guaranteed six-nines uptime?

I know, right?  Lookit this moron, who answered the initial poster by telling him the right pinout to properly convert between 2/3/4 wire fans off the top of his head, yet seems blissfully unaware of the high fan failure rate experienced by those unwise enough to run a stock bottom-of-the-barrel cooling fan.  Whadda maroon!

I do have to agree with you about the truly... "incredible"... level of expertise on this forum, however.

Where does it come from?  No, not your air conditioner - You open... wait for it... Another window.
Hot enough to kill the whole idea?  Seriously?  Do you folks live in Death Valley?



I don't design datacenters for a living, but Microsoft has a few guys you might need to set straight:
http://blogs.msdn.com/b/the_power_of_software/archive/2008/09/19/intense-computing-or-in-tents-computing.aspx
Let me know how they respond to your concerns.   

Excellent point - Consider my previous post as not applying to cheap sleeve bearings.  When buying cooling fans, don't skimp on the extra $0.50.

Sure they do - Just vent them outside.  One 20" box window fan will move the heat of a LOT of GPUs outside (unless you live somewhere with an outside temperature already over 110F).

DAMN but I wish I lived somewhere other than one of the highest $-per-KW states right about now.

First, don't worry about the amps.  You want a 12V fan, pick the one with the highest CFM (and/or the lowest db, depending on your needs).  plus or minus a few Watts won't make a damned bit of difference.

That said... The speed of a cooling fan may slightly decrease its useful life, but basically speed doesn't matter except for noise level.
 
They also don't tend to die while on, only when spinning up.  So for a 24/7/365.24 machine, you really don't need to worry about it.  Seriously, looking at 5+ years even at full speed before it dies (somewhat less if you have pets, and halve that if you smoke indoors).  As it gets near the end of its useful life, pay more attention when you power it up to make sure it actually starts (they usually start making a buzzing/grinding noise for a few seconds on startup, for a few months before completely failing).
 
As for the 3 vs 4pin connector, you've chosen the "wrong" 3 pins.  You may need to actually split out the wires to get it right, buy you want R/B for +12V/Gnd power, and yellow for the fan signal.  That said, fans cost nothing while good GPUs don't come cheap.

I would take that personally, if you had actually understood what I said.  Let me know when you learn to code, and I'll try to act hurt over your opinion of me.

I have no favorite horse in this race.  You would do well not to mock me when I point out that yours has a broken leg.

Not quite true - I would sincerely love to see a modern decentralized cryptocurrency make old-school notions of fiat currency like the dollar and euro completely obsolete.  I didn't join this community back when BTC had a value of $0.04 each (nor do I mine LTC, currently worth less than a US penny each) hoping to strike it rich.   

At worst, you could say that I believe in hedging my bets.  I won't buy $4000 in GPUs just to help seed a cyptocurrency, and if I make "some free money from your old stuff", hey, cool.

That said, we post in a community who already understands all of this (and still has legitimate disagreements - bordering on holy wars - on how it should all work).  But fiat currency or not, any "money" other than food, shelter, and sex, has no value except what a sufficiently large population agree it has.  We need to consider what the average Joe sees in this - And Joe has no reason to take an interest and play along if he can't join us without an unreasonable barrier to entry.

Bitcoin passed that barrier quite a few months ago, and today, "Joe" sees it as a subject of ridicule - "Oh, you mean that thing where the Russian mafia buys up all the ATI cards so they can launder their money?" (and don't even get me started on how many idiots haven't the faintest clue about the real meaning of a Ponzi scheme).  A CPU-only (or at least, GPU/ASIC unfriendly) BTC variant, I believe, has a lot of potential to let all the latecomer Joes try to get into the game again.



Haven't you (all) realized that if you can do it for LTC, you can do it for SC?  Though different core algorithms, they use the same underlying assumptions about the cost of memory accesses to limit GPU performance - Which mtrlt may have proven false (we still have no proof of that).  I mentioned both LTC and SC in my post because I mean to refer to both.  Nothing short of a "hard" serial algorithm will prevent GPUs from doing it better than CPUs; that, however, would tend to break the whole idea of a distributed currency (if 200 people can mine independently more efficiently than one person with a bigger rig, then most likely so can 200 ALUs vs one full CPU).

I don't really have an answer to this, but I think we need to discuss it without disintegrating into "my chosen BTC variant works better than yours".

Even if the underlying concept has some merit, I don't think any of the major miners actually use a random process - They all just increment the nonce in a very deterministic manner.  And even if they did, computers don't do very well at truly "random" numbers without an external source of them.

You could probably modify a published CPU miner to pick random nonces;  You would need, at least, a kernel-level hardware-derived entropy driver, and preferably a radioisotope-based RNG (something like you can build as described at http://www.inventgeek.com/Projects/alpharad/OverView.aspx).  But overall the idea wouldn't fit well with GPU mining, so if you manage to get a factor of 100 improvement by power of will, you should contact James Randi to collect your reward and forget about gaming BitCoins.   

Those of you welcoming GPU mining for LTC/SC have kinda missed the point of why those currently hold the #2/#3 spot in the cryptocurrency world...

I have a decent gaming rig, which I can put to work GPU-mining BTC.  I have no need for more than one decent gaming rig, nor do I have even the faintest excuse to deck it out with quad 6990s.  I do, however, have access to somewhere around twenty i5/i7 cores, all sitting around doing nothing very useful 95% of the time.  I suspect, in that regard, I represent a fairly typical BitCoin enthusiast.

I became interested in mining a second cryptocurrency only because tennebrix (and then others) came up with a way to make CPU mining viable again... I figured, why not put the two or three decent machines I run at home, as well as a couple strays at my shop, onto a useful task instead of simply letting them sit around growing slowly more obsolete by the day?  I suspect I represent a fairly typical LTC/SC enthusiast, in that regard.

Put bluntly, if GPU mining becomes viable for LTC and/or SC, their entire raison d'etre vanishes.  We don't need to decide if we should react if mtrlt releases Reaper13 and it really does work 25x faster on GPUs; We need to decide how to react, and preferably have our changes ready to roll out on the very same day.  And while "do nothing" always counts as an option, in this case it counts as an option that effectively dooms this entire branch of BTC variants.

And yes, before anyone points out the various other problems with BitCoin that various spinoffs address, I understand this doesn't reduce solely to mining hardware issues; But really - It does, so far as most people care.

First off, new to the pool, thanks for hosting it!

That said - I have a help/feature request...

Any way to change the password and/or name for a worker?  Or just disable worker PW checking altogether (if someone else wants to mine for me, they have my blessing! )

I realize I could retire a worker and just make a new one, but it seems silly to clutter up your DB with stale workers just because I accidentally hit "Add" a bit prematurely.

Thanks!

Ah... I did not, actually.  So I guess that makes it only a 2x speedup (the reported numbers went from 2.1 on just about any miner running two threads, to 8.0 reported by ufasoft's miner also running two threads).  Still... A 2x speedup!  Not complaining.   

Wow... Your miner took my Core2 machine from a mere 2.1MH/s to over 8!

Even weirder, it performs slightly better than the pure-CPU miners on my Athlon64X2 - Which completely sucks at the standard 4-way SSE version.

I don't have many bitcoins yet (finally passed a whole BTC just this past week), but I've sent a dime your way.  Thank you!

So... Put your mining rigs in the garage or some similarly outside-but-not-really area of your premises...  A guest bedroom where you can close the door and AC vents and open the windows; a garden shed with an extension cord; even on the porch, if you live in an area where you wouldn't expect to come home and find them missing.

Though still admirable to only mine at night, you don't need to increase your air conditioning load.

I had a hell of a time getting a GPU miner to run on Windows until I switched to PuddinPop's miners.  Only gotchas, make sure you have the NVidia CUDA driver installed, and on XP, you'll need the MSVC80 runtime environment (free from Microsoft, requires no configuration, and you have a good chance of already having it installed from something else).

Not to self-promote, but if you find it interesting, I just posted a table of all the currently-available Radeons, with prices (looked up personally at NewEgg about an hour ago) and assorted derived stats in another thread:
http://bitcointalk.org/index.php?topic=3814.msg55620#msg55620