01:32:57fanquake_:fanquake_ is now known as fanquake
05:43:05[\\\]:[\\\] is now known as [\\\\\\\\\\\\\\\
05:43:14[\\\\\\\\\\\\\\\:[\\\\\\\\\\\\\\\ is now known as [\\\\\\\\\\\\\\]
05:45:49[\\\\\\\\\\\\\\]:[\\\\\\\\\\\\\\] is now known as [\\\]
06:10:09maaku_:maaku_ is now known as maaku
07:18:53kaptah:kaptah is now known as Guest80006
07:18:53rdymac:rdymac is now known as Guest12488
07:18:53jgarzik:jgarzik is now known as Guest56534
07:18:53austinhill:austinhill is now known as Guest80428
07:18:53grau:grau is now known as Guest42646
07:18:54maaku:maaku is now known as Guest40421
07:18:56gmaxwell:gmaxwell is now known as Guest84585
07:18:56midnightmagic:midnightmagic is now known as 14WACHCUQ
07:22:51Guest84585:Guest84585 is now known as gmaxwell
07:31:14marienz:[Global Notice] Services are back, running a recent backup of the database. If you (automatically) identified to nickserv without using SASL while services were split, your password might be compromised. Please change your nickserv password (/msg nickserv help set password) if this applies to you. The misconfiguration allowing this has been fixed. If you have questions, ask in #freenode. Thanks!
07:50:19Guest56534:Guest56534 is now known as jgarzik_
08:21:23Guest40421:Guest40421 is now known as maaku
10:10:5414WACHCUQ:14WACHCUQ is now known as midnightmagic
11:07:07HM:HM is now known as HM2
11:39:29c0rw|sleep:c0rw|sleep is now known as c0rw1N
11:58:37c0rw1N:c0rw1N is now known as c0rw1n
13:26:14Guest80006:Guest80006 is now known as kaptah
14:09:44mr_burde_:mr_burde_ is now known as mr_burdell
14:58:04mike4:mike4 is now known as money
15:33:36jgarzik_:jgarzik_ is now known as jgarzik
15:54:16Burrito:Burrito is now known as United_Kingdom
15:54:16United_Kingdom:United_Kingdom is now known as Burrito
17:03:57airbreather:There was some talk in #bitcoin-dev about making the whole blockchain essentially one big Merkle tree... has any alt actually tried doing that?
17:04:21vetch:not that I'm aware.
17:04:52vetch:most altcoins seem to ignore the block format entirely and just mess about the with POW.
17:18:11Elriel:airbreather: how would that work?
17:20:16airbreather:Elriel: from my understanding, each block is represented as a sub-tree that includes the current transaction merkle tree as well as the header fields
17:22:27airbreather:"header fields" would probably be a sub-tree themselves so that things like ASICs don't need to know too much about the protocol details
19:16:35gmaxwell:andytoshi: probably of interest to you https://bitcointalk.org/index.php?topic=567625.0
19:18:30gmaxwell:("CoinShuffle: Practical Decentralized Coin Mixing for Bitcoin", I haven't read it yet but I understand its a decenteralized coinjoin approach)
19:31:22vetch:gmaxwell: nothing particularly surprising in there
19:33:18vetch:their example source code seems to completely ignore peer discovery
19:36:16gmaxwell:vetch: if there is source code at all thats pretty great compared to normal!
19:37:22vetch:I guess we're already coming out on top then
19:37:47airbreather:I would think peer discovery is one of the most important things to pay attention to in a CoinJoin-based solution? Otherwise, it seems really vulnerable to something like a Sybil attack...
19:38:18andytoshi:peer discovery is not too bad a problem, what maaku and i were having trouble with was the N^2 communication scaling, so if they've solved that somehow that's great
19:38:28vetch:and a Sybil attack is fairly devastating, if you're the only person in the mix you gain absolutely nothing from it.
19:38:52vetch:(only person, i.e. everybody else is a single second party)
19:38:58andytoshi:that's not true, then the other guy knows your deal but no external analyst does
19:39:45gmaxwell:vetch: it's still never worse than not using CJ at all, and likely better since you're still hidden from the public. Interestingly, one observation is that sybil attacks probably make everyone who isn't a target of the particular sybil safer.
19:40:05gmaxwell:Not to say that peer discovery isn't important, but to make progress we have to divide up problems.
19:42:07andytoshi:i don't like that the first few pages say stuff like "low overhead" and give time estimates instead of scaling behaviour
19:42:30vetch:you still wouldn't need much resources to completely poison the CJ network with Sybils. you'd need a lot of slush, sure, but that's easy to resell for no loss.
19:44:09vetch:I get that at best it's just ineffective not dangerous though.
19:45:17vetch:at worst I should say.
19:45:50andytoshi:looks like it's N^2 in bandwidth just from the graphs...but even having an implementation is a huge step forward :), so this is great news, i'll try to look into it more deeply
19:51:59gmaxwell:andytoshi: I'm not sure why quadratic scaling bothers you, durability to failure and transaction size limits keeps N relatively modest in any case.
19:55:34andytoshi:gmaxwell: for purely aesthetic reasons, it just felt suboptimal and made it discouraging to write code
19:55:45andytoshi:i.e., i'm just lazy and committing to my excuse means i have to keep whining about scaling
19:59:08airbreather:I can't prove it's impossible, but I can't wrap my head around how O(N) scaling would be feasible without adding centralization...
20:01:09andytoshi:airbreather: it's possible with multilinear crypto, so you won't find a 'proof' unless such crypto is necessarily insure
20:01:22andytoshi:(which it might be, but good luck showing it)
22:58:16phantomcircuit:gmaxwell, is it just me or is cert revocation almost entirely pointless
22:58:31gmaxwell:It's pretty much pointless.
22:58:46gmaxwell:It's also not very scalable.
22:58:51phantomcircuit:in a mitm ocsp can just be blocked and basically nobody treats that as a failure
22:59:09gmaxwell:It also leaks information about access patterns.
22:59:34phantomcircuit:gmaxwell, afaict the only way for it to actually work is for the CA's to revoke their intermediary signing certs
22:59:40phantomcircuit:which i kind of dont see happening
23:00:20gmaxwell:Thats what I promoted we encourage at work— basically reboot the CA infrastructure, though it didn't seem to be getting much traction.
23:00:29phantomcircuit:gmaxwell, ps did you ever pull power numbers for the ct box you down clocked?
23:00:33gmaxwell:ashame, it would be a fine time to fix a number of other things.
23:00:52gmaxwell:phantomcircuit: nope. don't have a meter I can easily deploy on it right now.
23:01:12phantomcircuit:gmaxwell, dont have a multi with a loop?
23:01:44gmaxwell:no, can't find the clamp on— and I don't have a split cable handy.
23:01:52gmaxwell:(I have one, but don't know where it is)
23:02:23phantomcircuit:gmaxwell, heh i have a split and stripped cable
23:02:26phantomcircuit:dat fire hazard
23:03:19gmaxwell:I can probably measure it using house voltage sag and comparing it to a fixed load, but meh.
23:03:32vetch:you need to get yourself one of those plug through meters
23:03:50vetch:I forget the name. it's something cute.
23:03:58phantomcircuit:vetch, split cables are so much more convenient
23:04:18phantomcircuit:you can monitor lots of stuff without needing a kill-a-watt for each thing
23:04:27gmaxwell:vetch: they're kinda overpriced ... similar to a clamp on multimeter in price but less flexible.
23:04:30vetch:phantomcircuit: harder to use while drunk. more chance for high voltage potential fingers.
23:04:50phantomcircuit:and if you're really feeling stupid you can strip a live wire
23:04:59phantomcircuit:just dont sue me when you die
23:05:39gmaxwell:lol I love the picture on this one: http://www.ebay.com/itm/P3-International-P4400-Kill-A-Watt-Electricity-Usage-Monitor-Used-/221414963916?pt=LH_DefaultDomain_0&hash=item338d5c3acc
23:05:40vetch:eh, earth leakage devices should save you. probably. wouldn't rely on them though.
23:05:59vetch:gmaxwell: oh god. wonder what got plugged into that.
23:05:59phantomcircuit:gmaxwell, why cant us utilities provide 240/415
23:05:59gmaxwell:"so I guess thats how it tells you the usage once the reading is out of range for the LCD"
23:06:08phantomcircuit:why the fuck do they supply 277/480
23:06:11vetch:"burnt face from heater. still works normally."
23:06:13phantomcircuit:how is that useful for anything
23:06:33phantomcircuit:gmaxwell, "used"
23:06:41gmaxwell:vetch: I've done that to outlets…
23:06:56vetch:gmaxwell: your country needs higher voltage mains.
23:07:27vetch:means you get to lug around less copper in your cables, if nothing else
23:07:46phantomcircuit:vetch, you can run 120/208 3P+E but it's only installed for washer/dryer and now some people have it for their electric cars
23:07:51gmaxwell:phantomcircuit: the psus in these miners can't take 277?
23:08:08phantomcircuit:gmaxwell, nobody sells a psu that will take 277
23:08:18phantomcircuit:they all have a maximum rating of 262-264
23:08:40gmaxwell:vetch: ~all homes in the us are wired as 240 with a center-tap neutral that lets you get 120.
23:08:47phantomcircuit:i actually tried running one at 277 and it exploded after about 10 minutes
23:09:02vetch:yeah, european countries are 230. the max rating on power supplies is just based on the accepted range on the higher side of 230v.
23:09:09gmaxwell:phantomcircuit: I think 277 arose because of carbon arc lighting or some crazy thing like that.
23:10:25phantomcircuit:gmaxwell, huh that's interesting
23:10:52phantomcircuit:the biggest nuisance is that you cant get anything designed for 230/240 in the us in a reasonable time frame
23:11:00phantomcircuit:it's all shipped by boat from the eu
23:11:15vetch:US frequency isn't the same either. must be weird to have all of your transformers humming 10hz different.
23:11:27phantomcircuit:so even if you're working with 15kV from the utility and installing your own transformers it's still a nuisance
23:12:48phantomcircuit:vetch, 60Hz is better for pretty much everything except industrial loads
23:13:10vetch:phantomcircuit: could be worse, you could be trying to run your bitcoin miners on an aeroplane. 400hz is used on there.
23:13:27vetch:reducing the size/weight of transformers I think
23:14:51phantomcircuit:yeah it does
23:15:43phantomcircuit:gmaxwell, it would be so very nice if psu's actually worked at 277
23:16:14vetch:gmaxwell: don't you get better higher efficiency with 240v power supplies?
23:16:18phantomcircuit:could move the pdus into the racks and save a ton of money on copper
23:16:41phantomcircuit:vetch, in general psu's operate better at 230/240 than at 120
23:16:59phantomcircuit:which is weird actually because they're dropping farther
23:17:01phantomcircuit:but whatever
23:17:13phantomcircuit:(typical output is 12vDC)
23:17:28vetch:imagine how much copper KNC has running to all of them
23:17:46phantomcircuit:that looks absolutely retarded
23:18:02phantomcircuit:there is zero active ventilation setup
23:18:08phantomcircuit:zero containment
23:18:33vetch:probably doesn't need it. it's in a cold country and hot air rises.
23:18:55vetch:I'm still trying to work out the power draw.. it's immense.
23:19:31vetch:interestingly if you look they're run in banks of six. KNC had that planned from the start, their original consumer run of the boxes miners had 6 headers on the controller FPGA
23:21:43vetch:each board seems to do 150GH @ 200W. there's 6 board in every bay, 6 bays to a stand, 20 stands to a row, 20 visible rows
23:21:46phantomcircuit:i just did the math
23:21:49phantomcircuit:that isn't very dense
23:22:09phantomcircuit:that's only liek 6kW per rack
23:22:36vetch:that's still like 2.1PH for the visible part of the photo, it looks like the facility goes a lot further in each direction
23:22:40phantomcircuit:vetch, the boards they had that worked perfectly did 150 @ 200
23:22:58phantomcircuit:their average is more like 100 @ 250
23:23:19vetch:oh, still a mammoth hash rate.
23:23:19phantomcircuit:and they get worse over time as died fails
23:23:33vetch:bet they churn them out cheaply though
23:23:44phantomcircuit:vetch, that's about 3/4th of the facility in that picture
23:24:00vetch:what are you basing that on?
23:24:13phantomcircuit:there's another one that has the entire length in it
23:24:26vetch:oh I just noticed there's laptops sitting on the rows controlling it. that's weird.
23:24:38phantomcircuit:they took the picture from that angle and cut off the end of it specifically to make it appear like it goes on forever
23:24:44vetch:got it.
23:25:00vetch:hate to think how loud that place is.
23:25:03phantomcircuit:it's like a bar with mirrors on all the walls
23:25:25vetch:I think the optimal thing would be to open a fruit drying / bitcoin mining farm.
23:26:07phantomcircuit:vetch, lets say that they're using 6kW per rack
23:26:13phantomcircuit:vetch, did you count how long each row is
23:26:22vetch:20 racks per row
23:26:41phantomcircuit:so that's 120kW per row
23:26:54phantomcircuit:240kW per row for both sides
23:27:23phantomcircuit:assuming 20 rows
23:28:01phantomcircuit:general rule of thumb is that to remove 1kW of heat you need 80 CFM of airflow (typically assuming 70F)
23:28:44phantomcircuit:so that's 384k CFM
23:29:06phantomcircuit:assumign it's 65F there in summer
23:29:24phantomcircuit:and the heat coming out hte back is 105F
23:29:36phantomcircuit:105-65 = 40F delta
23:30:27phantomcircuit:that's about 250 CFM/ton of air
23:31:38phantomcircuit:they need to move 1536 tons of 65F air flowing through the facility every minute
23:31:54phantomcircuit:and those are numbers assuming containment
23:32:01phantomcircuit:so lets say 2000 tons
23:33:31phantomcircuit:vetch, that's like 10 industrial chillers
23:37:18phantomcircuit:vetch, so yeah basically that equipment is almost certainly going to overheat in the summer
23:37:28vetch:* vetch shrugs
23:37:36vetch:I'm sure they've thought of all that
23:37:56vetch:aren't they based in sweden?
23:38:05phantomcircuit:vetch, they might not be expecting the stuff to even be running in 2 months
23:38:07phantomcircuit:so maybe not
23:38:49vetch:don't really think they'll be needing chillers. fans would do.
23:39:27phantomcircuit:vetch, sure
23:40:46adam3us:are petetodds diseconomies of scale kicking in yet? (harder to dissipate heat in bulk)
23:41:02vetch:I doubt they'll keep that running for a huge amount of time anyway, as you said phantomcircuit
23:41:15vetch:phantomcircuit: sell it on to someone else before it's totally useless
23:41:15phantomcircuit:adam3us, yes and no
23:41:49phantomcircuit:adam3us, yes they're kicking in, but the most cost effective solutions are still larger systems which use too much power for a home user anyways
23:42:30phantomcircuit:adam3us, so it's kicking in the sense of someone at 30Th/s vs someone at 3 Ph/s
23:42:46adam3us:phantomcircuit: think next gen. more density. more heat. govt scale. does it ever work practically for real constants
23:42:53phantomcircuit:but 30Th/s costs a lot
23:43:17phantomcircuit:adam3us, actually i think the next gen of gear will push back the diseconomies of scale for that a lot
23:43:43vetch:their next generation isn't meant to be more than 30% more efficient
23:43:54phantomcircuit:vetch, nothing nkc is working on is next gen
23:44:00adam3us:phantomcircuit: 30TH about $100k retail. should be able to get bulk discount not cheap tho. and your 100A house fuse is going to blow
23:44:12phantomcircuit:they have basically no hope of producing their neptune
23:44:22vetch:no, I reasoned that too
23:44:24phantomcircuit:which is why they have all those jupiters for their plan b
23:44:32vetch:it seems thermodynamically impossible
23:44:48phantomcircuit:vetch, to get a 30% improvement?
23:44:56adam3us:phantomcircuit: i mean not short-term/literal. i mean give it a few years whcihever manuf has best density & power efficiency
23:45:01vetch:no, to power a chip with their stated power consumption
23:45:13vetch:it would be literally hellfire coming out of a slither of silicon
23:45:27vetch:copper cables the size of your fist to carry the amperage
23:45:30phantomcircuit:adam3us, in a few years it will basically be a question of capital vs operating cost
23:45:39adam3us:phantomcircuit: i mean centralization is a threat. so i was interested at what point if ever diseconomy of scale actual works with real life constants
23:45:41phantomcircuit:very efficient chips will be slower
23:45:53vetch:very efficient chips will be small and numerous
23:45:58phantomcircuit:since a lot of the cost is per chip regardless of how fast the chip is
23:46:03phantomcircuit:you have a higher upfront cost
23:46:23phantomcircuit:but it's already to the point that you can acquire hardware which will cost more to operate than to purchase over the course of 1 year
23:46:32phantomcircuit:(if you're buying like 10m in hw)
23:46:40adam3us:phantomcircuit: efficient is lowest W/GH
23:47:02phantomcircuit:adam3us, yes im aware, there is a fairly strong link between W/Gh and $/Gh
23:47:14midnightmagic:cost also helps.
23:47:28phantomcircuit:mostly they have both been falling together, but it's getting to the point where you have to design for one or the other
23:47:36vetch:wouldn't pumping out 110nm chips be cheap but very inefficient?
23:47:36adam3us:phantomcircuit: this is good. gmaxwell was viewing electricity as limiting factor better than capital as limiting factor. otherwise capital rich win
23:48:28phantomcircuit:adam3us, for example, you can produce more smaller chips and drop the voltage to get better W/Gh, but your $/Gh will go up considerably
23:48:39adam3us:phantomcircuit: its strange but the current W/GH leader is on 40nm. maybe other just did weaker system optimization jobs
23:48:45midnightmagic:at some point there is a cost:electricity:speed factor which is optimal. I'm pretty sure we haven't reached it yet
23:49:00phantomcircuit:adam3us, all of the current 28nm chips are suboptimal designs
23:49:07midnightmagic:At least one or two of the "leaders" is just doing fpga hardcopies.
23:49:11adam3us:phantomcircuit: in fact they have 193 chips per miner (sp10)
23:49:27phantomcircuit:adam3us, yeah that's what i just said
23:49:45phantomcircuit:they can run those at lower voltage than the 28nm chips people are making
23:49:51phantomcircuit:so they're more efficient
23:50:23vetch:not that node size is everything
23:50:44vetch:55nm bitfury is more efficient than 28nm knc
23:50:57midnightmagic:That's because knc is doing lame hardcopies
23:51:07midnightmagic:(while simultaneously defrauding their customer base)
23:51:31vetch:not as badly as hashfast seems to be doing, i.e. not delivering customer hardware at all
23:51:42vetch:while selling the same stuff commercially.
23:51:47phantomcircuit:vetch, hashfast is afaict completely broke
23:51:58midnightmagic:HF did deliver *some* hardware. Just not all of it, or even a complete order, and not any of the upgrade packages.
23:51:59vetch:I don't doubt it.
23:52:06phantomcircuit:apparently they owe suppliers millions
23:52:22midnightmagic:They are also the subject of at least one bitcoin-freeze injunction.
23:52:35midnightmagic:.. or is the term object?
23:53:31phantomcircuit:subject is right in context
23:54:25vetch:they goofed either way. there's not really a reliable creator of these things.
23:54:52vetch:bit main seems to be fairly consistent I suppose.
23:55:05midnightmagic:Right, so they're subject to at least one bitcoin-freeze injunction for the total amount under dispute. Link here: http://www.deanslyons.com/PDFs/2014-03-27-Temporary-Restraining-Order.pdf
23:55:17[\\\]:bitmain's last batch has sucked on the manufacturing side
23:55:19[\\\]:total rush job
23:55:34phantomcircuit:bitmain is a reseller
23:55:36vetch:[\\\]: yeah, I saw some photos of mushed heatsinks
23:55:41phantomcircuit:those are obviously bitfury chips
23:56:14[\\\]:Are they?
23:56:18vetch:don't think they are
23:56:28vetch:the package is different
23:56:36phantomcircuit:the ant miner things?
23:56:47phantomcircuit:yeah those are definitely bitfury chips
23:57:10vetch:they're branded bitmain
23:57:20phantomcircuit:vetch, so?
23:58:02vetch:bitfury sells his/their chips on reels, branded with bitfury. there's lots of things using those chips and they all have the BF branding
23:58:53phantomcircuit:vetch, they also sell a license to manufacture the chips
23:58:59vetch:in their forum post they call it "BM1380" and mention 7 months of development, not buying a bunch of chips and lasering on their own branding