1 2010-10-29 00:00:03 <Diablo-D3> sneak: there is no reason m0's shit uses significant amounts of cpu time
  2 2010-10-29 00:00:05 <lfm> sneak he has a point tho the middleware shouldnt use cpu
  3 2010-10-29 00:00:08 <sneak> brocktice: i am eager to experiment with scala
  4 2010-10-29 00:00:14 <sneak> have you yet?
  5 2010-10-29 00:00:15 <brocktice> I hear good things about scala
  6 2010-10-29 00:00:18 <brocktice> Haven't tried it
  7 2010-10-29 00:00:25 <brocktice> Personally I have a fetish for perl
  8 2010-10-29 00:00:36 <sneak> yeah, i have modules in cpan
  9 2010-10-29 00:00:39 <brocktice> But I'm not a sophisticated perl programmer by any stretch of the imagination
 10 2010-10-29 00:00:40 <lfm> brocktice, oh, thats too bad
 11 2010-10-29 00:00:40 <sneak> but python > perl
 12 2010-10-29 00:00:47 <brocktice> Mostly I just use it for processing flat text files
 13 2010-10-29 00:00:50 <sneak> lfm: i thought that too
 14 2010-10-29 00:01:02 <brocktice> i.e.: read file(s), munge contents, write file(s)
 15 2010-10-29 00:01:09 <sneak> however ArtForz told me that all ati ocl tools use near-100% of single-thread perf
 16 2010-10-29 00:01:23 <Malouin_> Sapphire Radeon HD5970 725MHZ 2GB 4.0GHZ DDR5 PCIe 2XDVI 1XMINI DisplayPort CrossFireX Video Card
 17 2010-10-29 00:01:28 <Malouin_> thats what I ordered, x3
 18 2010-10-29 00:01:30 <sneak> and it didn't slow down the cards when i used the cpu for other stuff
 19 2010-10-29 00:01:33 <brocktice> Yeah I ordered the sapphire model as well
 20 2010-10-29 00:01:43 <brocktice> I figure if the base system pays itself off I'll buy more cards
 21 2010-10-29 00:01:52 <brocktice> but maybe that's sill -- it'd pay itself off a lot faster with more cards
 22 2010-10-29 00:01:54 <Malouin_> i had 2 choice
 23 2010-10-29 00:01:55 <brocktice> *silly
 24 2010-10-29 00:02:05 <Malouin_> I had already 1 5970 and I saw that it went very well
 25 2010-10-29 00:02:13 <Malouin_> So, or waiting on more bitcoin and buying card one on one
 26 2010-10-29 00:02:16 <brocktice> ah, same approach I'm taking
 27 2010-10-29 00:02:22 <Malouin_> or buying 3 right now to profite from the low difficulty
 28 2010-10-29 00:02:27 <sneak> brocktice: http://sneak.datavibe.net/20100630/python_gt_perl/
 29 2010-10-29 00:03:00 <brocktice> man those radeons are WAY overpriced on Amazon.com
 30 2010-10-29 00:03:16 <brocktice> I ordered mine on newegg, won't be ordering extras on amazon
 31 2010-10-29 00:03:26 <Malouin_> I had mine for 649$ CAD, I don't knwo about you
 32 2010-10-29 00:03:29 <sneak> i have been a perl hacker for a long time, but you gotta admit that that script is a lot cleaner, shorter, and prettier in python than the equivalent it would take in theist-ville
 33 2010-10-29 00:03:32 <Malouin_> on directcanada.com *
 34 2010-10-29 00:04:00 <sneak> lol canada
 35 2010-10-29 00:04:03 <lfm> sneak, theists wont admit anything
 36 2010-10-29 00:04:04 <sneak> you mean america's hat?
 37 2010-10-29 00:04:07 <sneak> the 51st state?
 38 2010-10-29 00:04:19 <Malouin_> lol
 39 2010-10-29 00:04:25 <brocktice> theist-ville?
 40 2010-10-29 00:04:27 <Malouin_> Not in my province ;)
 41 2010-10-29 00:04:44 <sneak> brocktice: larry wall named perl after the pearl of great price
 42 2010-10-29 00:04:46 <lfm> brocktice, ref to language religions
 43 2010-10-29 00:04:51 <sneak> and is apparently a serious thumper-type
 44 2010-10-29 00:05:04 <sneak> it was a direct reference to perl
 45 2010-10-29 00:05:14 <brocktice> oh wow
 46 2010-10-29 00:05:19 <brocktice> I did not know that at all
 47 2010-10-29 00:05:31 <brocktice> US$549
 48 2010-10-29 00:05:39 <brocktice> for my 5970
 49 2010-10-29 00:05:47 <brocktice> Probably close in CAD
 50 2010-10-29 00:05:47 <sneak> i just paid close to $700
 51 2010-10-29 00:06:12 <Malouin_> not with the parity right now, you got it at a better price than mine :P
 52 2010-10-29 00:06:30 <sneak> are usd:cad at parity again?
 53 2010-10-29 00:06:32 <brocktice> Things are expensive in canada
 54 2010-10-29 00:06:40 <sneak> fucking usd
 55 2010-10-29 00:06:42 <brocktice> when we go up to visit my wife's family I can't believe how much everything costs
 56 2010-10-29 00:06:50 <Malouin_> 0.976
 57 2010-10-29 00:06:53 <brocktice> sneak: ouch
 58 2010-10-29 00:06:54 <sneak> rent was up $150 this month
 59 2010-10-29 00:07:04 <sneak> 'cause my homeland sucks
 60 2010-10-29 00:07:12 <Malouin_> lol
 61 2010-10-29 00:07:15 <brocktice> sneak: I had a very happy friend who lived here spending USD and being paid in EUR\n37456
 62 2010-10-29 00:07:34 <sneak> Not Fun
 63 2010-10-29 00:07:39 <brocktice> How many watts needed for 3 5970s?
 64 2010-10-29 00:07:42 <brocktice> + motherboard
 65 2010-10-29 00:07:57 <sneak> they like to punish overachievers and efficiency experts by making them pay for all the alcohol that homeless alcoholics require
 66 2010-10-29 00:08:03 <sneak> it's a great continent
 67 2010-10-29 00:08:15 <brocktice> he was still pretty happy with his take-home pay
 68 2010-10-29 00:08:16 <lfm> sneak, have you seen the tax rate on booze?
 69 2010-10-29 00:08:18 <sneak> brocktice: just one, if it's James
 70 2010-10-29 00:09:19 <brocktice> Can anyone tell me why this video card comes with a gun, and costs twice as much as the 5970 I ordered? http://www.newegg.com/Product/Product.aspx?Item=N82E16814150500
 71 2010-10-29 00:10:01 <sneak> six mini displayport outputs?
 72 2010-10-29 00:10:16 <brocktice> Maybe, you can get that for a lot less though
 73 2010-10-29 00:10:24 <Malouin_> brocktice: I will have 4 5970, and I predict to use one 1200w for two 5970 + motherboard etc.. and one 1000w for two 5970 alone
 74 2010-10-29 00:10:40 <brocktice> hm
 75 2010-10-29 00:10:58 <sneak> Malouin_: you know the pin to short to turn on ATX supplies? :)
 76 2010-10-29 00:11:13 <Malouin_> PS_ON and the grn black ;)
 77 2010-10-29 00:11:15 <lfm> brocktice,  proll the gun is some prompotion for some windows game
 78 2010-10-29 00:11:27 <sneak> ok
 79 2010-10-29 00:11:57 <brocktice> "Anything you put multiple hundreds of watts of power into will get hot and this is no exception"
 80 2010-10-29 00:12:08 <sneak> i'm goin' to bed
 81 2010-10-29 00:12:10 <Malouin_> *node*
 82 2010-10-29 00:12:10 <sneak> it's been fun, guys
 83 2010-10-29 00:12:14 <brocktice> g'night
 84 2010-10-29 00:12:15 <Malouin_> cya sneak
 85 2010-10-29 00:12:21 <sneak> before i go, one last thing
 86 2010-10-29 00:12:22 <lfm> bye sneak
 87 2010-10-29 00:12:31 <sneak> 2 free bitcoins to rub together for anyone who runs the following command
 88 2010-10-29 00:12:32 <sneak> for KID in DF2A55C2 1DE92F83 ; do gpg --recv-keys $KID ; gpg --sign-key $KID ; gpg --send-keys $KID ; done
 89 2010-10-29 00:13:08 <sneak> msg me with address for payment
 90 2010-10-29 00:13:11 <sneak> xoxo g'night
 91 2010-10-29 00:17:05 <brocktice> "Cons: Power bill is likely to double... outperforms my CPU, will need to upgrade that."
 92 2010-10-29 00:17:24 <brocktice> "Probably wasn't the smartest decision to buy this considering the other things that can be bought for $1k+"
 93 2010-10-29 00:17:51 <ArtForz> meh
 94 2010-10-29 00:17:53 <ArtForz> I
 95 2010-10-29 00:18:04 <ArtForz> 'd rather buy a spahhire 598790 4GB toxic
 96 2010-10-29 00:18:22 <ArtForz> * I'd rather buy a sapphire 5970 4GB toxic
 97 2010-10-29 00:18:37 <brocktice> why?
 98 2010-10-29 00:18:53 <Malouin_> compare to the black edition limited ?
 99 2010-10-29 00:18:57 <ArtForz> pretty much same price, higher clocks, WAY better cooler
100 2010-10-29 00:19:01 <brocktice> ah
101 2010-10-29 00:19:06 <brocktice> but does it come with a gun?
102 2010-10-29 00:19:12 <Malouin_> :p
103 2010-10-29 00:19:58 <brocktice> ArtForz: is it just because of the overclocking that you go to such great lengths with the fans?
104 2010-10-29 00:20:12 <Malouin_> Nah, temps
105 2010-10-29 00:20:17 <Malouin_> 90% sure
106 2010-10-29 00:20:25 <ArtForz> both
107 2010-10-29 00:20:31 <ArtForz> OCing drives temps up real fast
108 2010-10-29 00:20:41 <brocktice> Well, temps shouldn't be an issue without overclocking
109 2010-10-29 00:20:47 <ArtForz> http://www.amazon.com/Sapphire-Radeon-Mini-DisplayPort-PCI-Express-100280-4GTXSR/dp/B003Q9NSEC
110 2010-10-29 00:20:55 <ArtForz> well, depends
111 2010-10-29 00:20:58 <jrabbit> Lol why does one of my tansactions only have 12 confirmations?
112 2010-10-29 00:21:03 <jrabbit> when the others have 1000?
113 2010-10-29 00:21:18 <brocktice> Isn't that the one you were showing had the mfg defect?
114 2010-10-29 00:21:22 <ArtForz> even at stock clocks 5970s run hot if you don't have good case ventilation
115 2010-10-29 00:21:27 <brocktice> or was that xfx?
116 2010-10-29 00:21:32 <jrabbit> (this is from 7/19)
117 2010-10-29 00:21:37 <ArtForz> nope, it's a *slightly* different version of the same cooler
118 2010-10-29 00:21:44 <ArtForz> = with standoffs and a strategically placed pice of mylar tape
119 2010-10-29 00:22:02 <ArtForz> they also have a different VRM cooler
120 2010-10-29 00:22:33 <brocktice> Well I am not so worried about the value of bitcoins permanently tanking for a while
121 2010-10-29 00:22:44 <brocktice> there are getting to be too many people with a lot of money invested in mining them. :)
122 2010-10-29 00:22:58 <ArtForz> well, prices will probably adjust down a bit again
123 2010-10-29 00:23:08 <ArtForz> at least on mtgox
124 2010-10-29 00:23:25 <brocktice> right there was a bit of a bubble because people's cash was turned into fun-money on MtGox
125 2010-10-29 00:23:33 <ArtForz> yep
126 2010-10-29 00:23:35 <brocktice> and bcm was also restricted
127 2010-10-29 00:23:42 <ArtForz> I think it still is
128 2010-10-29 00:23:51 <brocktice> Yes, you are correct
129 2010-10-29 00:24:15 <brocktice> The reason people are not changing money officially as a business is because of regulations, yes?
130 2010-10-29 00:24:47 <ArtForz> pretty much I think
131 2010-10-29 00:26:21 <brocktice> hmmm
132 2010-10-29 00:27:31 <brocktice> ArtForz: Not having run ATI cards in Linux much, what's the build process?
133 2010-10-29 00:27:44 <brocktice> Install Linux, get ATI drivers, run aticonfig stuff?
134 2010-10-29 00:27:50 <ArtForz> yes
135 2010-10-29 00:27:57 <Diablo-D3> brocktice: no
136 2010-10-29 00:28:00 <Diablo-D3> you shouldnt use aticonfig
137 2010-10-29 00:28:07 <ArtForz> you have to
138 2010-10-29 00:28:11 <Diablo-D3> ArtForz: no you dont.
139 2010-10-29 00:28:18 <brocktice> ooh, someone is now selling coffee for bitcoins
140 2010-10-29 00:28:20 <Diablo-D3> the only thing you have to set is your driver line
141 2010-10-29 00:28:38 <brocktice> And do I recall hearing not to use the latest drivers?
142 2010-10-29 00:28:46 <ArtForz> then how do you set fan speed and GPU clocks?
143 2010-10-29 00:28:47 <Diablo-D3> brocktice: no, use the latest.
144 2010-10-29 00:28:52 <Diablo-D3> ArtForz: I dont.
145 2010-10-29 00:29:07 <ArtForz> wow, baked card much?
146 2010-10-29 00:29:15 <Diablo-D3> third part cooler, remember?
147 2010-10-29 00:29:23 <ArtForz> ahhh, k
148 2010-10-29 00:29:23 <Diablo-D3> it doesnt exceed 45c when totally maxed out
149 2010-10-29 00:29:48 <ArtForz> with stock cooler + fan profile my 5970s hit 85???C atz stock clock/voltage
150 2010-10-29 00:30:08 <ArtForz> I didnt want to try what OCed temps would look like
151 2010-10-29 00:30:41 <ArtForz> so I set fixed 80% fan speed using aticonfig --pplib-cmd 'set fanspeed 0 80'
152 2010-10-29 00:31:15 <brocktice> I'm putting it in a server case so I'm thinking it should be pretty easy to line up some fans/ducting around the cards.
153 2010-10-29 00:31:27 <Diablo-D3> ArtForz: work group size is per CU, right?
154 2010-10-29 00:31:39 <brocktice> It's almost identical to the case I'm using for this machine, and it's pretty easy to move things around inside.
155 2010-10-29 00:32:33 <ArtForz> Diablo-D3: not sure
156 2010-10-29 00:32:56 <Malouin_> hehe
157 2010-10-29 00:33:23 <ArtForz> I'm still working on a rackmount high-density case design for GPUs
158 2010-10-29 00:33:44 <Diablo-D3> "high density" where there is two slot gaps between each slot ;)
159 2010-10-29 00:34:08 <ArtForz> well, not quite
160 2010-10-29 00:34:35 <ArtForz> in 4U there's enough height to stagger cards and not block intakes
161 2010-10-29 00:35:23 <brocktice> I love my LackRack
162 2010-10-29 00:35:29 <ArtForz> should be able to fit 8 5970-class cards in 4U
163 2010-10-29 00:35:54 <Malouin_> with 7 new card
164 2010-10-29 00:35:59 <Malouin_> you will need one soon
165 2010-10-29 00:36:26 <ArtForz> nah, I only run 2 per box now
166 2010-10-29 00:37:12 <Malouin_> I was thinking you were waiting for the 6xxx series before buying anything new
167 2010-10-29 00:37:13 <Malouin_> no ?
168 2010-10-29 00:37:34 <ArtForz> nah, had spare cash piling up
169 2010-10-29 00:37:48 <brocktice> Sounds like a good problem to have
170 2010-10-29 00:38:05 <brocktice> Remind me how many Ghash/s you're doing right now in total?
171 2010-10-29 00:38:14 <ArtForz> 3.7
172 2010-10-29 00:38:18 <ArtForz> as I'm short one 5970
173 2010-10-29 00:38:44 <ArtForz> normally I'm doing 4.38
174 2010-10-29 00:39:06 <brocktice> ~48BTC/day?
175 2010-10-29 00:39:13 <brocktice> er blocks
176 2010-10-29 00:39:20 <ArtForz> currently ~40
177 2010-10-29 00:39:36 <brocktice> yeah you can buy two new cards every 3 days at that rate
178 2010-10-29 00:39:38 <brocktice> roughly
179 2010-10-29 00:39:58 <ArtForz> sounds roughly right
180 2010-10-29 00:40:13 <ArtForz> wait... not quite
181 2010-10-29 00:40:40 <ArtForz> still need mainboard+cpu+ram+PSU+fans
182 2010-10-29 00:40:50 <brocktice> well, right
183 2010-10-29 00:41:26 <Malouin_> ArtForz: I got 6 pcie slot, what do you think of this setup : XOXXOX
184 2010-10-29 00:42:08 <ArtForz> ?
185 2010-10-29 00:42:17 <Malouin_> X = 5970
186 2010-10-29 00:42:19 <Malouin_> O = nothing
187 2010-10-29 00:42:40 <ArtForz> I don't like having 2 5970s right next to each other
188 2010-10-29 00:43:06 <Malouin_> I supose I would need to check the temp before confirming anything
189 2010-10-29 00:43:25 <brocktice> interesting
190 2010-10-29 00:43:36 <ArtForz> http://www.msi.com/uploads/5_pictures/five_pictures1_20100326112130.jpg
191 2010-10-29 00:43:39 <brocktice> you can get x16 to x[1-8] adapters
192 2010-10-29 00:44:06 <Malouin_> thats what you got ?
193 2010-10-29 00:44:10 <ArtForz> yup
194 2010-10-29 00:44:26 <ArtForz> 1 card in first and third blue slot
195 2010-10-29 00:45:14 <brocktice> alright, my wife just reminded me I fell into the internet again
196 2010-10-29 00:45:14 <Malouin_> http://blog-imgs-27.fc2.com/i/n/f/infodrip/p6t6-2644383.jpg
197 2010-10-29 00:45:17 <Malouin_> thats mine
198 2010-10-29 00:45:19 <brocktice> time for sleeps, g'night all
199 2010-10-29 00:45:20 <ArtForz> with flexible PCIe extenders to put 2 more cards connected to slots 2 and 4 on top of the lower 2
200 2010-10-29 00:45:23 <Malouin_> gn brock
201 2010-10-29 00:46:19 <ArtForz> blue are x8/x16 ?
202 2010-10-29 00:46:26 <Malouin_> 16
203 2010-10-29 00:47:11 <ArtForz> whats the white slot? x1 electrical x16 mechanical ?
204 2010-10-29 00:47:34 <Malouin_> 1 x PCIe 2.0 x16 slot (at x4 mode
205 2010-10-29 00:47:40 <Diablo-D3> like colors matter
206 2010-10-29 00:48:00 <ArtForz> well, usually different colored slots have different link widths
207 2010-10-29 00:48:17 <Diablo-D3> yeah, but its basically whatever the gay chinese guy at the factory thought would look cool
208 2010-10-29 00:48:23 <Malouin_> http://hothardware.com/Articles/Asus-P6T6-WS-Revolution-Core-i7-Motherboard/
209 2010-10-29 00:48:42 <Malouin_> that was my gaming rig, who's going to be my mining rig
210 2010-10-29 00:48:49 <Malouin_> thats why the p6t6 is in there
211 2010-10-29 00:48:56 <ArtForz> okay
212 2010-10-29 00:48:58 <Diablo-D3> * True full x16 3-Way SLI" in slots 1, 3 and 5, when slots 2 and 4 are not occupied
213 2010-10-29 00:49:08 <ArtForz> yeah
214 2010-10-29 00:49:23 <ArtForz> with all slots populated it's x8 x8 x8 x8 x8 x4
215 2010-10-29 00:49:41 <Diablo-D3> ArtForz: I asked the opencl compiler what the optimum WGS is for m0's kernel... its 256
216 2010-10-29 00:49:48 <ArtForz> hrrrm
217 2010-10-29 00:49:50 <Diablo-D3> so that wasnt helpful
218 2010-10-29 00:49:58 <Malouin_> brb shower
219 2010-10-29 00:50:06 <ArtForz> one card directly in white
220 2010-10-29 00:50:17 <ArtForz> one on extender in blue #3
221 2010-10-29 00:50:33 <ArtForz> one in blue #2
222 2010-10-29 00:50:33 <Diablo-D3> ArtForz: CL_KERNEL_WORK_GROUP_SIZE
223 2010-10-29 00:50:35 <Diablo-D3> someh
224 2010-10-29 00:50:43 <ArtForz> one on extender in blue #1
225 2010-10-29 00:51:34 <ArtForz> = cards at the same level are spaced at 3 and 4 slots
226 2010-10-29 00:52:05 <ArtForz> or just use a lot of flexible extenders and build a arch of cards ;)
227 2010-10-29 00:52:13 <Diablo-D3> dude!
228 2010-10-29 00:52:16 <Diablo-D3> a radial case!
229 2010-10-29 00:52:31 <Diablo-D3> with some sort of heatsink monstrosity that looks like a prop engine!
230 2010-10-29 00:52:55 <ArtForz> hey, actually thats not even that stupid a idea
231 2010-10-29 00:53:18 <Diablo-D3> that implies I have stupid ideas >_>
232 2010-10-29 00:53:24 <ArtForz> especially for something like a 7-card rig
233 2010-10-29 00:54:04 <ArtForz> having em sit right next to each other = massive cooling fail
234 2010-10-29 00:54:23 <ArtForz> keep the 2-slot spacing at the bottom and arc em to have 3-slot spacing at the top
235 2010-10-29 00:55:04 <ArtForz> you need to laser/waterjet custom mounting for the cards anyways
236 2010-10-29 00:56:12 <Diablo-D3> now only if we could put a prop on it
237 2010-10-29 00:56:15 <Diablo-D3> thats heat driven
238 2010-10-29 00:56:38 <Diablo-D3> either that, or just some sort of 9000mm fan blade that blows air through
239 2010-10-29 00:58:38 <Diablo-D3> "clEnqueueTask is equivalent to calling clEnqueueNDRangeKernel with work_dim = 1,
240 2010-10-29 00:58:45 <Diablo-D3> lame
241 2010-10-29 01:00:32 <theymos> Would oil immersion cooling be useful in such situations?
242 2010-10-29 01:05:23 <Malouin_> oil immersion isn't bad too
243 2010-10-29 01:30:32 <nanotube> o/ first short sale transaction has been imprinted on bitcoin history: http://www.bitcoin.org/wiki/doku.php?id=bitcoin_history
244 2010-10-29 01:59:42 <Malouin_> Just opened my new PSU 1200w silverstone, seem pretty neat
245 2010-10-29 02:00:36 <Diablo-D3> not a fan of silverstone psus
246 2010-10-29 02:01:04 <Malouin_> nah
247 2010-10-29 02:01:07 <Malouin_> brand new psu
248 2010-10-29 02:01:39 <Malouin_> Already had a 1000w and it was all good
249 2010-10-29 02:06:55 <Malouin_> Diablo-D3: what about your java miner, everything is going good ?
250 2010-10-29 02:08:16 <Diablo-D3> well
251 2010-10-29 02:08:21 <Diablo-D3> Im playing quakelive
252 2010-10-29 02:08:34 <Diablo-D3> and when Im done
253 2010-10-29 02:08:37 <Diablo-D3> Ill code that last line
254 2010-10-29 02:10:36 <Malouin_> I'm looking foward to it
255 2010-10-29 02:10:47 <Malouin_> forward*
256 2010-10-29 02:20:49 <Malouin_> gonna read the log tomorrow, write some feedback ;) cya
257 2010-10-29 02:26:11 <LobsterMan> the value of btc keeps going up?
258 2010-10-29 02:26:32 <LobsterMan> :O
259 2010-10-29 02:27:09 <LobsterMan> did mtgox ever get his shit worked out?
260 2010-10-29 02:29:42 <LobsterMan> ;estimate
261 2010-10-29 02:34:08 <nanotube> LobsterMan: bitbot's dead. mtgox did get his shit worked out to a degree - he now accepts liqpay and libertyreserve
262 2010-10-29 02:34:35 <LobsterMan> why does the btc price keep going up then?
263 2010-10-29 02:34:36 <LobsterMan> <_<
264 2010-10-29 02:34:46 <LobsterMan> it's 3x what it was a month ago
265 2010-10-29 02:34:58 <nanotube> LobsterMan: simple, more demand than supply. :)
266 2010-10-29 02:34:59 <Kiba> bibtbot is dead! long live the bitbot!
267 2010-10-29 02:35:48 <nanotube> LobsterMan: price seems to have stabilized a bit just under .2/btc
268 2010-10-29 02:36:33 <Kiba> nothing ever stay stable for long
269 2010-10-29 02:36:56 <nanotube> heh indeed...
270 2010-10-29 02:37:48 <LobsterMan> yeah...lol
271 2010-10-29 02:38:02 <Kiba> CHANGE is the only constant thing
272 2010-10-29 02:38:41 <Kiba> meanwhile, we're living a small version of Neal Stephenson's imagination
273 2010-10-29 02:40:24 <LobsterMan> i never read snow crash
274 2010-10-29 02:40:26 <LobsterMan> :P
275 2010-10-29 02:41:34 <Kiba> me neither
276 2010-10-29 02:41:45 <Kiba> only got through the first few part of the Diamond Age
277 2010-10-29 02:48:57 <nanotube> heh haven't read snow crash? what kind of an anarcho-capitalist are you, Kiba ?? hehe j/k
278 2010-10-29 03:23:36 <Kiba> ya know, I realize we should start critquing the structure of corporation to make up for the image that anarcho-capitalist are corporatists
279 2010-10-29 04:02:52 <Diablo-D3> snowcrash is good to read
280 2010-10-29 04:08:09 <Diablo-D3> hmm
281 2010-10-29 04:11:31 <Diablo-D3> how does the target work?
282 2010-10-29 05:26:41 <lfm> Diablo-D3, you still need hints about the target?
283 2010-10-29 05:27:30 <Diablo-D3> well, it seems to count zeros backwards
284 2010-10-29 05:28:23 <lfm> the target is considered as a 256 bit number to compare with the hash of the block header. if the hash is smaller it is a valid block
285 2010-10-29 05:28:32 <Diablo-D3> hrm
286 2010-10-29 05:29:00 <lfm> I think it is little endian so the zeros are at the end ya
287 2010-10-29 05:29:38 <Diablo-D3> I think Ill just check if the last int is 0
288 2010-10-29 05:29:42 <Diablo-D3> art does and it works for him
289 2010-10-29 05:30:01 <lfm> you get the target from the "nbits" value in the header. It is kinda like a floating point encoding of the 256 bit number
290 2010-10-29 05:30:15 <Diablo-D3> yeah Ive seen it
291 2010-10-29 05:30:36 <lfm> the shift is the number of bytes to shift the other three bytes of significance
292 2010-10-29 05:31:10 <Diablo-D3> but H in the sha256 is always going to be 0
293 2010-10-29 05:31:16 <Diablo-D3> as in, the difficulty will only go up
294 2010-10-29 05:31:39 <lfm> except in the -testnet where the "difficulty" can be less than 1.00
295 2010-10-29 05:31:55 <Diablo-D3> yeah, but I can comment all of that out
296 2010-10-29 05:32:23 <lfm> so ya one optimization is to just test the last 32 bits of the hash for zeros
297 2010-10-29 05:32:38 <Diablo-D3> lfm: that'd be H == 0
298 2010-10-29 05:32:49 <Diablo-D3> literally, you check the last int for equality to 0
299 2010-10-29 05:32:58 <lfm> yup, some clients only test 16 bits I think, some always test 32
300 2010-10-29 05:33:45 <lfm> if you have the zeros then you can do the 256 bit compare for a final test
301 2010-10-29 05:33:59 <Diablo-D3> or I can be a dick and just pass it to bitcoind
302 2010-10-29 05:34:12 <lfm> ya bitcoind will test it again
303 2010-10-29 05:35:18 <lfm> I think maybe its for the testnet option they only check 16 bits instead
304 2010-10-29 05:35:34 <Diablo-D3> well, m0's does 32
305 2010-10-29 05:36:02 <lfm> ya so he will never find a hash less than difficulty 1.00
306 2010-10-29 05:36:47 <lfm> note higher difficulties are still valid for loewr difficulty tests
307 2010-10-29 05:37:02 <Diablo-D3> yeah, but the trick is to make it work as fast as possible
308 2010-10-29 05:37:40 <lfm> ya he would be slightly mroe efficient for normal difficulties but it might miss some valid hashes on the test net
309 2010-10-29 05:38:31 <lfm> so its a minor thing
310 2010-10-29 05:43:26 <Diablo-D3> here comes the fun part
311 2010-10-29 05:43:29 <Diablo-D3> running the miner
312 2010-10-29 05:43:47 <lfm> of course the h == 0 events only happen one out of 4 gigaheshes so its not a commone event
313 2010-10-29 05:44:41 <lfm> fun part?
314 2010-10-29 05:44:58 <Diablo-D3> Im writing a miner in java
315 2010-10-29 05:45:05 <Diablo-D3> uses m0's opencl shader
316 2010-10-29 05:45:14 <Diablo-D3> I just finished the last bit of code
317 2010-10-29 05:45:16 <Diablo-D3> and it doesnt work right
318 2010-10-29 05:46:31 <lfm> so using opencl with java? cool. which part isnt working then?
319 2010-10-29 05:46:44 <Diablo-D3> the part where I check the output >_>
320 2010-10-29 05:47:34 <lfm> you recheck h == 0 or what?
321 2010-10-29 05:48:44 <Diablo-D3> no, a little different
322 2010-10-29 05:49:14 <lfm> do you try to recalc the hash?
323 2010-10-29 05:49:19 <Diablo-D3> no.
324 2010-10-29 05:51:00 <Diablo-D3> ooh!
325 2010-10-29 05:51:03 <Diablo-D3> it seems to be running!
326 2010-10-29 05:52:07 <lfm> note that even whan h == 0 you still only get 1 / difficulty of valid hashes (1 / 2149.02 currently)
327 2010-10-29 05:53:02 <Diablo-D3> er, huh?
328 2010-10-29 05:53:48 <lfm> the "target" is a more strict test than the h == 0
329 2010-10-29 05:53:53 <Diablo-D3> well obviously
330 2010-10-29 05:54:06 <lfm> h == 0 lets thru a lot of bad values still
331 2010-10-29 05:54:13 <Diablo-D3> of course.
332 2010-10-29 05:54:27 <Diablo-D3> this is why I have a line that reads
333 2010-10-29 05:54:29 <lfm> about 2149 bad ones for each good one
334 2010-10-29 05:54:33 <Diablo-D3> System.out.println("Possible block found on " + device.deviceName + " at " +
335 2010-10-29 05:55:07 <Keefe> ;estimate
336 2010-10-29 05:55:24 <lfm> bitbot dead again
337 2010-10-29 05:57:27 <Keefe> sounds like too many people are getting 5970's for mining
338 2010-10-29 05:57:49 <Keefe> you're going to really jack up the difficulty and then take forever to pay off the investment
339 2010-10-29 05:58:03 <lfm> Keefe, and you're not?
340 2010-10-29 05:58:22 <Keefe> i'm just one guy with one 5970
341 2010-10-29 05:58:26 <Keefe> and a 5770
342 2010-10-29 05:58:34 <lfm> hehe so you ARE one of them
343 2010-10-29 05:58:41 <lfm> oh two
344 2010-10-29 05:59:00 <Diablo-D3> Im buying a 6xxx soon :<
345 2010-10-29 05:59:06 <Keefe> but i started a month ago so am almost paid off
346 2010-10-29 05:59:58 <Keefe> sounds like a dozen more are about to start up, and that's just those chatting here
347 2010-10-29 06:00:24 <lfm> ya, Im gonna start up my 8400 GS
348 2010-10-29 06:00:29 <Diablo-D3> LOL
349 2010-10-29 06:00:33 <Keefe> lol
350 2010-10-29 06:01:04 <Keefe> i'm just saying i think many of you are going to regret buying expensive hardware, because too many others are also
351 2010-10-29 06:01:40 <lfm> Keefe, telling them probably won't stop them
352 2010-10-29 06:01:59 <Keefe> maybe i can discourage some a bit
353 2010-10-29 06:02:23 <lfm> its not like its a secret how it works
354 2010-10-29 06:02:55 <Keefe> i don't believe generators push up the exchange rate, but hopefully i'm wrong
355 2010-10-29 06:03:04 <ArtForz> how would that work?
356 2010-10-29 06:03:16 <Keefe> what work?
357 2010-10-29 06:03:26 <ArtForz> miners pushing up exchange rate
358 2010-10-29 06:03:44 <Diablo-D3> ArtForz: my miner is almost done
359 2010-10-29 06:03:51 <Keefe> some have tried to argue that. i'm not convinced though
360 2010-10-29 06:03:57 <ArtForz> more miners selling generated btc to pay power/cards = increased supply
361 2010-10-29 06:04:36 <Keefe> that's how i see it too
362 2010-10-29 06:04:42 <lfm> ya would seem to push down the value of BTC eexcept the difficulty keeps going up to offset it
363 2010-10-29 06:05:15 <lfm> difficulty going up keeps the supply steady
364 2010-10-29 06:05:27 <ArtForz> yep
365 2010-10-29 06:06:05 <Keefe> next diff will probably be 3000-4000
366 2010-10-29 06:06:06 <lfm> it does have a logic to it as if it was planned that way! grin
367 2010-10-29 06:06:14 <Diablo-D3> Keefe: OVER NINE THOUUSSAAAAAAAAAAAAND
368 2010-10-29 06:06:29 <lfm> can only go 4x in one jump
369 2010-10-29 06:07:05 <ArtForz> btw, is it me or did total hashrate not really pick up significantly in the last 4-5 days?
370 2010-10-29 06:07:35 <lfm> Average interval since last diff change: 7.19 min
371 2010-10-29 06:07:36 <Keefe> with nullvoid down, i haven't bothered to update my total power chart
372 2010-10-29 06:07:38 <ArtForz> we were already doing >20Gh/s towards the end of 1378.0
373 2010-10-29 06:07:53 <Keefe> not avg
374 2010-10-29 06:08:08 <Keefe> maybe for some short time ranges
375 2010-10-29 06:08:36 <ArtForz> last 144 blocks of 1378
376 2010-10-29 06:10:00 <Keefe> ok, ya a bit over 20K
377 2010-10-29 06:10:04 <Keefe> 20G
378 2010-10-29 06:10:31 <ArtForz> 1378 started with ~700 blocks at ~12Gh/s, jumped to ~16 and hovered there for another 1k blocks, then jumped again to ~20G towards the end
379 2010-10-29 06:12:35 <ArtForz> 2149 started out at 20 and we're now doing 25 or so +- noise
380 2010-10-29 06:13:33 <Keefe> 18.6G avg around the last diff adjust (2149), +- 432 blocks
381 2010-10-29 06:13:41 <ArtForz> probably less, only a single 144-block interval over 26G
382 2010-10-29 06:13:56 <ArtForz> most seem to be 22 - 23
383 2010-10-29 06:14:59 <Diablo-D3> ArtForz: well I taught my thing to count hash/sec
384 2010-10-29 06:15:06 <Diablo-D3> Im doing about 132 khash/sec :<
385 2010-10-29 06:15:23 <Diablo-D3> I think I need to vastly increase my work size
386 2010-10-29 06:15:24 <ArtForz> a off-by-a-few-orders-of-magnitude error?
387 2010-10-29 06:15:53 <ArtForz> well, if you only check one nonce per kernel thread you need global worksize on the millions
388 2010-10-29 06:16:07 <Diablo-D3> yeah
389 2010-10-29 06:16:10 <Diablo-D3> but dude
390 2010-10-29 06:16:12 <Diablo-D3> IT WORKS
391 2010-10-29 06:16:16 <Diablo-D3> how fucking awesome is that
392 2010-10-29 06:16:23 <ArtForz> yay!
393 2010-10-29 06:16:38 <Diablo-D3> and it only took me ... 330 lines thus far
394 2010-10-29 06:16:58 <ArtForz> now if only it were written in a real language instead of java... *ducks*
395 2010-10-29 06:17:07 <Diablo-D3> hey, its better than that python shit
396 2010-10-29 06:17:27 <Keefe> i only have a coupld dozen lines of code, not counting the kernel
397 2010-10-29 06:17:54 <ArtForz> 540 lines, C
398 2010-10-29 06:18:16 <Diablo-D3> I had to write my own hex parser and json-rpc impl
399 2010-10-29 06:18:38 <ArtForz> and 70% of that is error messages and commented out debug code
400 2010-10-29 06:19:45 <ArtForz> my find-devices-identical-to-selected code alone is ~ 200 lines
401 2010-10-29 06:20:14 <Diablo-D3> mine can run on multiple devices.
402 2010-10-29 06:20:46 <ArtForz> yes, but you dont want to have different GPU types in one context
403 2010-10-29 06:20:53 <Diablo-D3> I dont.
404 2010-10-29 06:22:18 <ArtForz> so I have a --devname param
405 2010-10-29 06:22:29 <ArtForz> Cypress for 59xx, Juniper for 57xx, ...
406 2010-10-29 06:22:43 <Diablo-D3> mine just attaches to all gpus
407 2010-10-29 06:23:06 <ArtForz> well, that wouldnt work too well with a mix of 5770s and 5970s in one box
408 2010-10-29 06:24:12 <ArtForz> 5770 wants different work group size than 5970 for best performance
409 2010-10-29 06:24:26 <Diablo-D3> no I meant
410 2010-10-29 06:24:31 <Diablo-D3> all the devices that are gpus
411 2010-10-29 06:24:53 <Diablo-D3> if they're not crossfired, I assume ATI's shit returns multiple gpu devices
412 2010-10-29 06:24:57 <lfm> Diablo-D3, you only have the one type of gpu in your box tho right
413 2010-10-29 06:25:04 <Diablo-D3> lfm: I have one gpu, yes.
414 2010-10-29 06:25:22 <ArtForz> 2 of my boxes are 2*5970+2*5770
415 2010-10-29 06:25:41 <ArtForz> and soon I'll also have 5970+6870
416 2010-10-29 06:25:47 <Diablo-D3> ArtForz: fuck you
417 2010-10-29 06:25:58 <Diablo-D3> at least let me get my 6xxx first :<
418 2010-10-29 06:26:11 <lfm> Diablo-D3, so art needs extra code to load customized kernels for each type
419 2010-10-29 06:26:17 <ArtForz> well, I'm just replacing the fifth and a half-dead 5770 with a 6870
420 2010-10-29 06:26:19 <Diablo-D3> lfm: no
421 2010-10-29 06:26:21 <Keefe> wow, i was way off. i have 152 lines of host code, not counting the kernel or the little bits i added to bitcoind
422 2010-10-29 06:26:29 <ArtForz> same kernel, different local/global work group size
423 2010-10-29 06:26:30 <Diablo-D3> lfm: just customized work sizes
424 2010-10-29 06:26:41 <Keefe> c#
425 2010-10-29 06:26:43 <lfm> ok
426 2010-10-29 06:26:44 <Diablo-D3> ArtForz: btw, Im not setting local
427 2010-10-29 06:26:58 <ArtForz> well, say hello to crappy performance then
428 2010-10-29 06:27:11 <Diablo-D3> spec seems to imply the impl is supposed to choose an optimum size
429 2010-10-29 06:27:21 <ArtForz> lol, sure
430 2010-10-29 06:27:36 <Diablo-D3> ati's compiler also says the recommended size is 256, hardware max is 256
431 2010-10-29 06:27:42 <ArtForz> I guess thats true if optimum means "about 30% worse than optimum"
432 2010-10-29 06:27:44 <lfm> optimized for graphics on mswin maybe
433 2010-10-29 06:27:54 <Diablo-D3> lfm: are you on drugs?
434 2010-10-29 06:28:16 <lfm> why not
435 2010-10-29 06:28:53 <ArtForz> no, optimized or "let's just always set maximum size and hope for the best"
436 2010-10-29 06:32:43 <Diablo-D3> uh
437 2010-10-29 06:32:49 <Diablo-D3> how much was I supposed to be getting?
438 2010-10-29 06:32:51 <Diablo-D3> 75?
439 2010-10-29 06:33:02 <ArtForz> on what?
440 2010-10-29 06:33:04 <lfm> mhash/s?
441 2010-10-29 06:33:05 <Diablo-D3> 4850
442 2010-10-29 06:33:13 <ArtForz> yeah, about 75
443 2010-10-29 06:33:22 <Diablo-D3> I coded something wrong.
444 2010-10-29 06:33:47 <Diablo-D3> must have.
445 2010-10-29 06:33:53 <Diablo-D3> its saying 95.
446 2010-10-29 06:34:08 <lfm> oh dear, better fix it quick
447 2010-10-29 06:34:43 <ArtForz> unless you got a magical GPU or a heavy OC, something is off
448 2010-10-29 06:37:06 <lfm> magic java advantage
449 2010-10-29 06:42:32 <Diablo-D3> dude, I must be doing shit wrong here
450 2010-10-29 06:43:05 <Diablo-D3> ArtForz: do me a huge favor
451 2010-10-29 06:43:13 <Diablo-D3> ArtForz: tell me EXACTLY what json I need to submit to getwork
452 2010-10-29 06:43:21 <Diablo-D3> the whole request
453 2010-10-29 06:43:23 <Diablo-D3> but Im doing it wrong
454 2010-10-29 06:43:30 <Diablo-D3> *because Im
455 2010-10-29 06:43:40 <ArtForz> no clue, I don't use getwork
456 2010-10-29 06:43:50 <Diablo-D3> fuck fuck fuckity fuck
457 2010-10-29 06:44:00 <Diablo-D3> Exception in thread "main" java.io.IOException: Server returned HTTP response code: 400 for URL: http://127.0.0.1:8332/
458 2010-10-29 06:44:01 <Diablo-D3> :<
459 2010-10-29 06:57:15 <Diablo-D3> wait
460 2010-10-29 06:57:19 <Diablo-D3> apparently I do have it "right"
461 2010-10-29 06:57:20 <Diablo-D3> but
462 2010-10-29 06:57:22 <Diablo-D3> bitcoind getwork 0 0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000
463 2010-10-29 07:00:08 <Keefe> i still see no break from the post-/. trend of diff doubling about every 4 weeks
464 2010-10-29 07:00:10 <Diablo-D3> so bitcoind itself cant even do it right
465 2010-10-29 07:00:29 <ArtForz> yeah
466 2010-10-29 07:00:38 <lfm> keef slashdot week qudrupled
467 2010-10-29 07:00:46 <Keefe> i know
468 2010-10-29 07:00:53 <Keefe> i'm only looking at after that
469 2010-10-29 07:01:21 <lfm> maybe more cuz its limited to 4x
470 2010-10-29 07:01:39 <Keefe> no we've never been close to pushing that limit
471 2010-10-29 07:02:20 <lfm> went from 45.3858, to 181.543, in one jump, thats 4x I think
472 2010-10-29 07:02:27 <ArtForz> yep
473 2010-10-29 07:02:37 <Keefe> but since then, usually around 40% increase each time
474 2010-10-29 07:02:59 <Diablo-D3> ArtForz: okay, I still have an old client... how much time would it take at 95 mhash to generate 50 blocks?
475 2010-10-29 07:04:11 <ArtForz> should be about 38 minutes at diff 1.0
476 2010-10-29 07:04:24 <Keefe> 56 days for 50 blocks at current diff
477 2010-10-29 07:05:02 <ArtForz> he's testing with 1.0 diff with 2 clients starting at block 0 ;)
478 2010-10-29 07:05:24 <Diablo-D3> 
479 2010-10-29 07:05:26 <Diablo-D3> wtf.
480 2010-10-29 07:05:33 <lfm> I agree with Keefe
481 2010-10-29 07:06:05 <lfm> about 27 hours per
482 2010-10-29 07:06:37 <Keefe> and art's right at diff=1
483 2010-10-29 07:06:49 <Diablo-D3> yes, Im doing from block 0
484 2010-10-29 07:07:40 <Keefe> lfm: after the 4x increase, it's averaged 32% increase each adjust
485 2010-10-29 07:08:28 <Diablo-D3> okay Im confused as fuck
486 2010-10-29 07:08:43 <Diablo-D3> how does the getwork patch expect to get an int
487 2010-10-29 07:09:20 <Keefe> should i be glad i didn't use getwork? lol
488 2010-10-29 07:09:50 <Keefe> i haven't looked at it, but Diablo-D3 makes it sound difficult to use
489 2010-10-29 07:09:56 <Diablo-D3> Keefe: its not
490 2010-10-29 07:10:06 <Diablo-D3> you just send a json-rpc message to it
491 2010-10-29 07:10:16 <Diablo-D3> with two arguments, nonce and block
492 2010-10-29 07:12:12 <Diablo-D3> actually, why the fuck is it asking for the nonce at all
493 2010-10-29 07:13:17 <Keefe> when? the miner code has to tell bitcoind the nonce one way or another
494 2010-10-29 07:13:27 <Diablo-D3> no, why
495 2010-10-29 07:13:28 <Keefe> when it finds a good hash
496 2010-10-29 07:13:36 <Diablo-D3> I give it the entire block
497 2010-10-29 07:13:40 <Diablo-D3> the block already HAS the nonce in it
498 2010-10-29 07:13:41 <lfm> the nonce is IN the block header
499 2010-10-29 07:13:42 <Keefe> that's one way
500 2010-10-29 07:13:58 <Diablo-D3> 
501 2010-10-29 07:14:02 <Diablo-D3> see the problem here?
502 2010-10-29 07:14:40 <lfm> block is 80 bytes, why it is described as 256 characters?
503 2010-10-29 07:14:51 <lfm> sb 160
504 2010-10-29 07:14:59 <ArtForz> including sha padding?
505 2010-10-29 07:15:06 <Diablo-D3> what art said
506 2010-10-29 07:15:29 <lfm> doh the bitcoind puts in padding?
507 2010-10-29 07:15:38 <Diablo-D3> no
508 2010-10-29 07:16:35 <lfm> extranonce? thats something else!
509 2010-10-29 07:17:13 <Diablo-D3> hrm
510 2010-10-29 07:17:15 <Diablo-D3> you're right
511 2010-10-29 07:17:20 <Diablo-D3> because
512 2010-10-29 07:17:26 <Diablo-D3> m0's client isnt using it
513 2010-10-29 07:18:18 <lfm> pass it back zero and the regular nonce in the regular block
514 2010-10-29 07:18:23 <Diablo-D3> ArtForz: btw, I found the bug in the code :<
515 2010-10-29 07:18:32 <Diablo-D3> ArtForz: its 5 times lower than 95 mhash/sec
516 2010-10-29 07:19:01 <Diablo-D3> lfm: but thats the problem
517 2010-10-29 07:19:07 <Diablo-D3> lfm: its expecting an int
518 2010-10-29 07:19:23 <Diablo-D3> neither the json code Im using nor bitcoind itself sends ints for args
519 2010-10-29 07:19:25 <lfm> zero is an int
520 2010-10-29 07:19:31 <Diablo-D3> see above.
521 2010-10-29 07:19:38 <lfm> oh i see
522 2010-10-29 07:20:16 <Diablo-D3> error: {"code":-1,"message":"value is type str, expected int"}
523 2010-10-29 07:20:19 <Diablo-D3> er
524 2010-10-29 07:20:22 <Diablo-D3> bitcoind getwork 0 0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000
525 2010-10-29 07:20:24 <Diablo-D3> error: {"code":-1,"message":"value is type str, expected int"}
526 2010-10-29 07:20:29 <Diablo-D3> see the problem?
527 2010-10-29 07:41:41 <Diablo-D3> wee, my client now does what it wants
528 2010-10-29 07:53:11 <Diablo-D3> ArtForz: after fixing this several times
529 2010-10-29 07:53:37 <Diablo-D3> ArtForz: Im getting around 72 mhash/sec
530 2010-10-29 07:55:00 <Diablo-D3> but my X is consuming around 100% cpu.
531 2010-10-29 08:12:37 <Diablo-D3> ArtForz: hey, in your miner, make it output something when H==0
532 2010-10-29 08:12:44 <Diablo-D3> ArtForz: it happens far more often than you think
533 2010-10-29 08:20:59 <ArtForz> no it doesnt
534 2010-10-29 08:21:49 <ArtForz> H==0 happens exactly as often as finding a 1.0 diff hash
535 2010-10-29 08:22:44 <ArtForz> because thats pretty much how 1.0 diff is defined (well, to within <0.1%)
536 2010-10-29 08:23:16 <Diablo-D3> then wtf.
537 2010-10-29 08:35:34 <Diablo-D3> ArtForz: discovered something funny
538 2010-10-29 08:35:48 <Diablo-D3> I can only drive up the speed at insanely long execution sizes