1 2011-01-26 00:00:06 <luke-jr> jgarzik: that's not nice :P
  2 2011-01-26 00:00:48 <jgarzik> gavinandresen: therefore, it is common on Sparc and PPC to have _mostly_ 32-bit binaries on 64-bit OS
  3 2011-01-26 00:00:50 <andrew12^mac> luke-jr: well... you can make your mac boot in 64 bit by holding 6 and 4 while its booting (go figure)
  4 2011-01-26 00:00:57 <lfm> gavinandresen: prolly only want to build 64 bit binaries on 64 bit linux for now then
  5 2011-01-26 00:01:05 <jgarzik> gavinandresen: x86 muddies the waters a bit.  x86-64 adds many more registers, and makes code much more efficient (32-bit x86 is register-starved, even with tricks like register renaming)
  6 2011-01-26 00:01:10 <andrew12^mac> so... anyone here know suse?
  7 2011-01-26 00:01:14 <jgarzik> x86 is essentially a compressed instruction set
  8 2011-01-26 00:01:32 <lfm> gavinandresen: especially tricky would be to build 32 and 64 bit versions of wxwidget pacages
  9 2011-01-26 00:01:48 <gavinandresen> lfm, jgarzik:  so to produce 32 and 64 bit versions of Linux bitcoin I should create 32 and 64-bit VMs?  Or is it easy to get the 64-bit compiler to produce 32-bit binaries?
 10 2011-01-26 00:02:07 <jgarzik> gavinandresen: the bottom line for bitcoin:  a 32-bit Linux binary is just fine for UI and JSON purposes.  You will sacrifice some khash/sec in the built-in miner, versus a pure 64-bit build.
 11 2011-01-26 00:02:23 <jgarzik> gavinandresen: bitcoin does not need more than 32 bits of address space
 12 2011-01-26 00:02:28 <lfm> gavinandresen: it is possible to create 32 bit binaries onb 64 bit system but it can be pretty tricky I think
 13 2011-01-26 00:02:37 <luke-jr> gavinandresen: chroots would be enough
 14 2011-01-26 00:02:38 <davex__> What's a good mainboard for running two 5970's?
 15 2011-01-26 00:02:39 <jgarzik> lfm: it's trivial.  CFLAGS += "-m32"
 16 2011-01-26 00:02:48 <jgarzik> I do 32-bit builds on 64-bit all the time.
 17 2011-01-26 00:02:50 <tcatm> could something like scratchbox be used instead of a VM?
 18 2011-01-26 00:02:51 <jgarzik> windows and linux.
 19 2011-01-26 00:02:54 <doublec> yes, -m32 is all that's needed
 20 2011-01-26 00:03:04 <gavinandresen> tcatm: what's scratchbox?
 21 2011-01-26 00:03:05 <luke-jr> jgarzik: will that work for the assembly?
 22 2011-01-26 00:03:06 <lfm> jgarzik: is trivial to make wxwidgets 32 bit too?
 23 2011-01-26 00:03:11 <jgarzik> lfm: yes
 24 2011-01-26 00:03:19 <doublec> scratchbox is what I used to get the n900 build on arm
 25 2011-01-26 00:03:24 <lfm> jgarzik: ok I never tried it
 26 2011-01-26 00:03:48 <jgarzik> luke-jr: qualified yes -- you get the 32-bit instruction set, but assembler will always let you generate weird shit if you really try.
 27 2011-01-26 00:04:10 <luke-jr> well of course, I meant the asm actually working in miners :P
 28 2011-01-26 00:04:31 <jgarzik> luke-jr: bitcoin's built-in miner does not use assembler AFAIK
 29 2011-01-26 00:04:41 <tcatm> gavinandresen: a cross compiler build environment that simulates a target enviroment incl. cpu emulation (e.g. arm on x86) but uses native compilers. pretty easy to have a bunch of targets that require only a few hundred MB diskspace
 30 2011-01-26 00:04:42 <jgarzik> and that's the only miner relevant to this discussion.
 31 2011-01-26 00:05:00 <luke-jr> jgarzik: I thought it supported 4way O.o
 32 2011-01-26 00:05:08 <jgarzik> luke-jr: that's C code
 33 2011-01-26 00:05:12 <luke-jr> oh
 34 2011-01-26 00:05:20 <doublec> does it use intrinsics?
 35 2011-01-26 00:05:21 <jgarzik> luke-jr: special compiler built-ins that access SSE2 data types and instructions
 36 2011-01-26 00:05:26 <jgarzik> doublec: yes
 37 2011-01-26 00:06:31 <gavinandresen> Right.  So....   it sounds like Ubuntu 64bit is NOT the right environment to build a release bitcoin/bitcoind.  Ubuntu 32 would be better   (I don't care about trying to optimize the built-in CPU miner)
 38 2011-01-26 00:06:54 <jgarzik> gavinandresen: I would consider -not- doing a Linux build at all.  Consider letting the distro-specific packagers tackle that problem, and set up their own repos.  Punt on Linux altogether.  Windows is the only platform where you, as a project manager, _need_ to do binaries.
 39 2011-01-26 00:07:11 <luke-jr> gavinandresen: realistically, it's best to use your target OS to build packages for it
 40 2011-01-26 00:07:22 <jgarzik> gavinandresen: if you ask Fedora/Deb/Ubu package people, I bet that's what they'll tell you
 41 2011-01-26 00:07:33 <gavinandresen> jgarzik:  that's why I'm asking
 42 2011-01-26 00:07:36 <luke-jr> jgarzik: I would agree, but the original BitCoin client is notoriously difficult to build
 43 2011-01-26 00:07:37 <jgarzik> gavinandresen: freshrpms and other third party aggregator repos do builds from source
 44 2011-01-26 00:08:20 <jgarzik> gavinandresen: then the user process is simply "add $pkg_repo to list ; apt-get install bitcoin"
 45 2011-01-26 00:08:23 <lfm> gavinandresen: ya source only for linux might wel save you a few headaches
 46 2011-01-26 00:08:35 <gavinandresen> I'm all for fewer headaches
 47 2011-01-26 00:08:46 <jgarzik> people will pkg it for you
 48 2011-01-26 00:08:50 <jgarzik> put efforts into supporting that
 49 2011-01-26 00:09:26 <gavinandresen> Any advice on how to get the three or four different groups who have worked on various packaging to FIGURE IT OUT AND SUBMIT SOME FRICKIN PATCHES ???
 50 2011-01-26 00:10:13 <luke-jr> gavinandresen: you're welcome to merge my branch
 51 2011-01-26 00:10:22 <jgarzik> it doesn't help that bitcoin's build system is a bit ad-hoc.  Hate autotools all you want, but autotools ensures __automatic integration into package build systems__
 52 2011-01-26 00:10:36 <luke-jr> gavinandresen: it also has a bugfix to avoid throwing away sub-cent change
 53 2011-01-26 00:10:41 <gavinandresen> jgarzik:  uh huh.......
 54 2011-01-26 00:10:45 <jgarzik> the work it takes integrate a non-autotools pkg into Fedora or Deb is often 5x or 10x the work of a standard autotools pkg.
 55 2011-01-26 00:11:26 <luke-jr> gavinandresen: inside your Subversion checkout, you should be able to simply 'bzr pull lp:~luke-jr/+junk/bitcoin-tonal' I think
 56 2011-01-26 00:11:31 <luke-jr> assuming bzr and bzr-svn are install
 57 2011-01-26 00:11:32 <luke-jr> ed
 58 2011-01-26 00:11:52 <luke-jr> or 'bzr merge &' might work better, to review things before committing
 59 2011-01-26 00:12:03 <luke-jr> not sure how bzr handles a native svn checkout when you pull into it
 60 2011-01-26 00:12:51 <luke-jr> but if someone's done an autotools version, that might make more sense
 61 2011-01-26 00:13:01 <luke-jr> (in which case, let me know and I'll isolate my bugfix)
 62 2011-01-26 00:13:22 <gavinandresen> As you can tell, I don't know nothin bout cmake/autotools/etc..... I need somebody who DOES know something about it to step up and do it.
 63 2011-01-26 00:13:26 <gavinandresen> jgarzik: you willing?
 64 2011-01-26 00:13:37 <jgarzik> gavinandresen: I'd just post Linux source and call for packaging-related patches.  Let people motivated in that area take up the slack.
 65 2011-01-26 00:13:49 <jgarzik> gavinandresen: Yes, I would be willing.  I've done many autotools conversions.
 66 2011-01-26 00:14:07 <gavinandresen> jgarzik:  awesome, I'll add you to the team
 67 2011-01-26 00:14:27 <jgarzik> gavinandresen: building on MSVC++ seems like a sticking point though.  not sure how to proceed there.  I can do Linux + mingw builds with autotools.
 68 2011-01-26 00:14:34 <gavinandresen> luke-jr:  hmm?  I'm not going to learn Yet Another Revision Control System to pull your patches
 69 2011-01-26 00:14:45 <fabianhjr_> I got a quick question. How can you verify the BlockChain is legitimate? What signs each block? Could I in a LAN where someone starts to download the blockchain give him a rogue one?
 70 2011-01-26 00:15:14 <doublec> what are the windows builds done with? MSVC or MingW?
 71 2011-01-26 00:15:18 <luke-jr> gavinandresen: those single commands should work fine, but whatever. Subversion doesn't really work well in a distributed fashon
 72 2011-01-26 00:15:19 <jgarzik> doublec: both :)
 73 2011-01-26 00:15:27 <jgarzik> doublec: there are makefiles for each
 74 2011-01-26 00:15:41 <jgarzik> doublec: and I don't want to break MSVC if people really want that.
 75 2011-01-26 00:15:42 <doublec> ok, I've only used mingw to build it
 76 2011-01-26 00:16:10 <jgarzik> and that importance is greatly lessened today.
 77 2011-01-26 00:16:19 <doublec> I'm inclined to agree
 78 2011-01-26 00:16:21 <luke-jr> gavinandresen: how would you like me to prepare it for easier merging?
 79 2011-01-26 00:16:34 <doublec> on the other hand, getting all the required mingw packages and building them was a bit of a mission.
 80 2011-01-26 00:16:40 <gavinandresen> luke-jr: github pull request.
 81 2011-01-26 00:16:42 <doublec> I guess that's the same issue on msvc (or even more so)
 82 2011-01-26 00:16:58 <gavinandresen> luke-jr:  or send it to jgarzik :-)
 83 2011-01-26 00:17:01 <jgarzik> doublec: depends on distro.  We hired the mingw maintainer at Red Hat, so it's really easy on Fedora <grin>
 84 2011-01-26 00:17:04 <luke-jr> gavinandresen: has the project actually moved to git then?
 85 2011-01-26 00:17:11 <jgarzik> :_
 86 2011-01-26 00:17:13 <jgarzik> :)
 87 2011-01-26 00:17:15 <luke-jr> it's really easy on Gentoo as well
 88 2011-01-26 00:17:16 <doublec> heh, good plan
 89 2011-01-26 00:17:28 <luke-jr> crossdev --target mingw32 & done!
 90 2011-01-26 00:17:50 <fabianhjr> I oppened USDp/TCP 8883 port and got knocked offline. LOL xD Crappy ISP.
 91 2011-01-26 00:18:13 <jgarzik> there is even an easy script "mingw32-configure", which runs "./configure" with necessary cross-compile args for mingw.  it's great, and damned easy w/ autotools pkgs.
 92 2011-01-26 00:18:26 <luke-jr> gavinandresen: would a git tree off-github be feasable? I prefer not to use notioriously proprietary software, even if webapps
 93 2011-01-26 00:18:30 <jgarzik> that's how I build cpuminer for win32 + win32 installer.
 94 2011-01-26 00:18:43 <gavinandresen> jgarzik:  if nobody volunteers to maintain/build MSVC, then we should drop it.
 95 2011-01-26 00:19:08 <luke-jr> jgarzik: any plans for AVX support in cpuminer? ;)
 96 2011-01-26 00:19:40 <jgarzik> gavinandresen: cool.  in the meantime, my autotools conversion should not destroy the ability to do "make -f makefile.msvc"  autotools conversion merely implies changes to default makefiles etc.
 97 2011-01-26 00:19:58 <jgarzik> luke-jr: I accept patches!
 98 2011-01-26 00:20:09 <luke-jr> jgarzik: I failed miserably in my attempt to write one 9
 99 2011-01-26 00:20:22 <luke-jr> I blame lack of example of how to use AVX :P
100 2011-01-26 00:20:54 <luke-jr> Intel should publish some AVX examples
101 2011-01-26 00:20:56 <luke-jr> or something
102 2011-01-26 00:26:08 <fabianhjr> If I open my port(8883) how many connections can I expect and how fast would they fill?
103 2011-01-26 00:27:05 <tcatm> 8883 or 8333?
104 2011-01-26 00:28:57 <lfm> fabianhjr: Id guess no more that 25-30 at the worst and most of the time fewer, it goes up and down
105 2011-01-26 00:31:06 <gavinandresen> fabianhjr: the bitcoin faucet has 38 connections right now, after running for 24 hours or so.  I've seen it have as many as 60.
106 2011-01-26 00:31:30 <fabianhjr> tcatm: oh, really? xD
107 2011-01-26 00:32:53 <lfm> thats for a well known node, new nodes should be less
108 2011-01-26 00:34:59 <lfm> molecular: works for me, what versaion driver and sdk you have installed?
109 2011-01-26 00:34:59 <luke-jr> gavinandresen: can I put my git branch off-github? I cannot agree to their ridiculous ToS
110 2011-01-26 00:35:35 <jgarzik> luke-jr: gitorious should work; I think that's 100% FOSS
111 2011-01-26 00:35:54 <luke-jr> jgarzik: sure, that's fine.
112 2011-01-26 00:36:08 <luke-jr> I just don't care to obligate myself to make a legal defense of github :p
113 2011-01-26 00:36:31 <luke-jr> at my expense, even :/
114 2011-01-26 00:36:33 <jgarzik> luke-jr: really any HTTP webspace will do, even.
115 2011-01-26 00:36:47 <jgarzik> git supports HTTP pulls (although, they are slower than git pulls)
116 2011-01-26 00:36:54 <luke-jr> sure.
117 2011-01-26 00:37:00 <luke-jr> I have a git server myself
118 2011-01-26 00:37:10 <luke-jr> but I'll use Gitorious in hopes that more people migrate there ;)
119 2011-01-26 00:38:42 <fabianhjr> OMG no single block in over 20 minutes. What the hell? D:<
120 2011-01-26 00:39:43 <luke-jr> &
121 2011-01-26 00:40:45 <fabianhjr> C'mon don't stand there. Mine!
122 2011-01-26 00:41:16 <luke-jr> I am
123 2011-01-26 00:41:23 <luke-jr> at a whole 4.5 MH/s
124 2011-01-26 00:42:32 <fabianhjr> LOL, I am doing 300 MHashes/sec and it is no longer enough. xD
125 2011-01-26 00:44:44 <fabianhjr> Bitcoin going down to 0.40 BTC again. :P
126 2011-01-26 00:47:03 <jgarzik> and if only I'd held onto my early bitcoins
127 2011-01-26 00:47:42 <jgarzik> that 10,000 BTC that went to puddinpop for open sourcing his stuff would be worth a pretty penny
128 2011-01-26 00:48:03 <jgarzik> all generated w/ built-in miner or purchase around $0.06
129 2011-01-26 00:48:39 <gavinandresen> jgarzik: woulda coulda shoulda.... I wish I'd bought a bunch more at 0.005....
130 2011-01-26 00:49:01 <jgarzik> :)
131 2011-01-26 00:52:57 <jgarzik> gavinandresen: 'obj' dir in git seems weird.  would it be ok to add top-level obj/ directory to top-level .gitignore, and remove that tree from git?
132 2011-01-26 00:53:05 <fabianhjr_> gavinandresen: you can still buy and they are probably going to be worth a lot more in the future if this gets adopted mainstream. :P
133 2011-01-26 00:53:49 <Kiba`> fabianhjr:
134 2011-01-26 00:53:55 <Kiba`> we need to discuss the new editor
135 2011-01-26 00:54:00 <Kiba`> and next issue
136 2011-01-26 00:54:28 <fabianhjr> :/ I got to go almost. I got a class and have another 5 minutes.
137 2011-01-26 00:54:41 <Kiba`> I am thinking wednesday
138 2011-01-26 00:54:49 <Kiba`> as the next perma deadline
139 2011-01-26 00:54:52 <gavinandresen> jgarzik: ... but the makefiles don't create obj/ if it doesn't exist, so git checkout then make would fail, yes?
140 2011-01-26 00:54:53 <Kiba`> think you can handle?
141 2011-01-26 00:55:09 <fabianhjr> For this W or next one to this?
142 2011-01-26 00:55:14 <Kiba`> next one
143 2011-01-26 00:55:18 <Kiba`> for next issue
144 2011-01-26 00:55:35 <jgarzik> gavinandresen: correct.  hmmm....
145 2011-01-26 00:55:41 <Kiba`> the editor will choose the deadline for submission...
146 2011-01-26 00:55:47 <fabianhjr> Ok, I can handle it then.
147 2011-01-26 00:55:52 <fabianhjr> Ok
148 2011-01-26 00:55:53 <gavinandresen> (teaching makefiles to create obj/ would be OK with me, except I wouldn't know how to do that for WIndoze)
149 2011-01-26 00:56:01 <tcatm> mkdir -p obj/nogui/ in makefile?
150 2011-01-26 00:56:39 <jgarzik> gavinandresen: nod.   it's easy enough to create directories.  just asking why it was there at all in the git repo (and that question has been answered; thanks)
151 2011-01-26 00:58:22 <jgarzik> gavinandresen: FWIW, autotools always drops object output in the current directory, making the typical srcdir != objdir workflow slightly different:  mkdir /tmp/mybuilddir ; cd /tmp/mybuilddir ; /usr/src/bitcoin/configure ; make
152 2011-01-26 00:58:40 <jgarzik> cross-compiles and other arcane gadgetry work the same way
153 2011-01-26 00:58:51 <jgarzik> srcdir != objdir is a common workflow
154 2011-01-26 00:59:25 <jgarzik> of course, slackers like me often use srcdir==objdir for convenience
155 2011-01-26 01:03:38 <andrew12^mac> hm
156 2011-01-26 01:03:46 <andrew12^mac> is bitcoin compatible with db4.5?
157 2011-01-26 01:04:07 <Diablo-D3> use 4.7
158 2011-01-26 01:04:58 <molecular> lfm: ati driver 10.12, stream sdk 2.1
159 2011-01-26 01:05:02 <gavinandresen> Hmm, that's another release loose end-- I don't think any windows people said whether or not upgrading to db 4.8 would cause them problems.
160 2011-01-26 01:05:17 <jgarzik> andrew12^mac: I -think- so, but no guarantees.  Your main obstacle is to avoid building bitcoin with db version X, creating databases, then rebuilding with db version Y
161 2011-01-26 01:05:29 <jgarzik> gavinandresen: good point.  what is current windows binary using?
162 2011-01-26 01:05:33 <jgarzik> 4.7?
163 2011-01-26 01:05:39 <gavinandresen> Yeah, 4.7 I believe.
164 2011-01-26 01:06:23 <gavinandresen> 4.7.25.NC according to build-msw.txt
165 2011-01-26 01:06:39 <andrew12^mac> I found a 4.7 rpm so all is good
166 2011-01-26 01:07:05 <jgarzik> gavinandresen: boilerplate db disclaimer is that you should use db_upgrade when going 4.7 -> 4.8, 4.8->4.9, etc.  They attempt to make format backwards compatible, but do make changes from time to time, especially to tertiary files such as logs.
167 2011-01-26 01:07:40 <gavinandresen> jgarzik: yup-- I think all the problems I've seen have been with the log files as people DOWNgrade
168 2011-01-26 01:07:44 <jgarzik> yeah :/
169 2011-01-26 01:08:07 <prax> hey I didn't even read your issues yet, halfway through the thread, but I might write something
170 2011-01-26 01:09:49 <andrew12^mac> http://pastebin.com/KDwzETEB -- openssl and libopenssl-devel is installed...
171 2011-01-26 01:11:31 <andrew12^mac> (that's a make error if you're too lazy to click it)
172 2011-01-26 01:12:14 <andrew12^mac> wait hmm
173 2011-01-26 01:17:33 <andrew12^mac> this is irritating
174 2011-01-26 01:18:03 <ne0futur> you could need berkeley db  -devel
175 2011-01-26 01:18:20 <andrew12^mac> that's probably why
176 2011-01-26 01:18:24 <andrew12^mac> because i found an rpm
177 2011-01-26 01:18:27 <andrew12^mac> its just not -devel
178 2011-01-26 01:18:37 <andrew12^mac> http://rpm.pbone.net/index.php3/stat/4/idpl/14272755/dir/mandriva_2010/com/libdbcxx4.7-4.7.25-6mdv2010.1.i586.rpm.html
179 2011-01-26 01:18:38 <doublec> some distro's package the c++ part of libdb seperately
180 2011-01-26 01:18:47 <ne0futur> use gentoo , you ll never have this type of problems ;)
181 2011-01-26 01:18:50 <andrew12^mac> heh
182 2011-01-26 01:19:12 <doublec> try downloading the libdb source from oracle - I'm still waiting for their page to load, sigh.
183 2011-01-26 01:19:48 <andrew12^mac> going to try http://rpm.pbone.net/index.php3/stat/4/idpl/14272748/dir/mandriva_2010/com/libdb4.7-devel-4.7.25-6mdv2010.1.i586.rpm.html first
184 2011-01-26 01:20:32 <ne0futur> google mandriva error: db_cxx.h
185 2011-01-26 01:20:48 <ne0futur> there are bugs and answers
186 2011-01-26 01:22:28 <ne0futur> https://qa.mandriva.com/show_bug.cgi?id=45257
187 2011-01-26 01:22:42 <ne0futur> you probably just have to add a msymlink
188 2011-01-26 01:22:54 <ne0futur> There's already a symlink for db.h in /usr/include.
189 2011-01-26 01:22:58 <ne0futur> Fedora and Debian are both providing symlinks for db.h db_185.h and db_cxx.h.
190 2011-01-26 01:24:22 <andrew12^mac> ne0futur: its suse
191 2011-01-26 01:24:53 <ne0futur> and you want to install a mandriva rpm on suze ?
192 2011-01-26 01:25:06 <andrew12^mac> well i dunno. it might work? lol
193 2011-01-26 01:26:45 <andrew12^mac> i dont know anything about rpm
194 2011-01-26 01:27:38 <Kiba``> so
195 2011-01-26 01:27:42 <Kiba``> anybody read the magazine yet?
196 2011-01-26 01:28:00 <andrew12^mac> im just gonna try building berkeley db from source
197 2011-01-26 01:32:28 <ne0futur> it might work yes, but it will probably break everything :p
198 2011-01-26 01:34:37 <andrew12^mac> hm
199 2011-01-26 01:34:57 <andrew12^mac> you don't need to have any arguments in configure for it to build the stuff for c++ right?
200 2011-01-26 01:35:29 <andrew12^mac> ah yes
201 2011-01-26 01:35:59 <andrew12^mac> because i backtraced it!
202 2011-01-26 01:36:46 <andrew12^mac> $ make me a sandwich
203 2011-01-26 01:40:01 <jgarzik> $ make out
204 2011-01-26 01:40:20 <andrew12^mac> lol
205 2011-01-26 01:40:30 <andrew12^mac> make: *** No rule to make target `out'.  Stop.
206 2011-01-26 01:49:15 <andrew12^mac> if i can get this to work, i'm going to figure out how to make an rpm for bitcoin and i'll make rpms for all these, and put them in a repo somewhere.
207 2011-01-26 01:49:25 <andrew12^mac> that is, if i decide to be less lazy for a bit
208 2011-01-26 01:50:25 <ne0futur> try to post your problems to build bitcoin on suse on the forums
209 2011-01-26 01:51:42 <andrew12^mac> i'm building db4.7 right now
210 2011-01-26 01:52:25 <ne0futur> also try updatedb ; locate db_cxx.h
211 2011-01-26 01:52:43 <ne0futur> db_cxx.h could be here but not in path and you just need a ln -s
212 2011-01-26 01:54:30 <andrew12^mac> pretty sure this is gonna work
213 2011-01-26 01:56:20 <andrew12^mac> hm
214 2011-01-26 01:56:24 <andrew12^mac> ;;ping
215 2011-01-26 01:56:24 <gribble> pong
216 2011-01-26 01:56:26 <andrew12^mac> weird
217 2011-01-26 01:56:48 <andrew12^mac> ssh connection froze :\n193345
218 2011-01-26 01:57:28 <andrew12^mac> its responding to ping though o_o
219 2011-01-26 02:03:55 <dooglus> andrew12^mac: try ssh -M user@host
220 2011-01-26 02:04:51 <luke-jr> FWIW, db_cxx.h isn't in the include path usually
221 2011-01-26 02:05:36 <andrew121mac> 21:56:41 -!- supybot [supybot@McInHost-3362242C.compute-1.amazonaws.com] has quit [Ping timeout]
222 2011-01-26 02:05:47 <andrew121mac> .-.
223 2011-01-26 02:05:59 <luke-jr> w00t, I won a 5850
224 2011-01-26 02:06:06 <luke-jr> $163
225 2011-01-26 02:06:17 <andrew121mac> nice
226 2011-01-26 02:07:44 <luke-jr> http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&rt=nc&nma=true&item=150553802314&si=xHhjaUb88oWHhtvkLwlQWY9FRko%253D&viewitem=
227 2011-01-26 02:08:11 <hacim> i got lucky with a 5970 today
228 2011-01-26 02:08:28 <luke-jr> ?
229 2011-01-26 02:08:30 <hacim> $280
230 2011-01-26 02:08:33 <luke-jr> :O
231 2011-01-26 02:08:46 <luke-jr> nice deal, but I wouldn't have the PSU to handle it possibly
232 2011-01-26 02:08:48 <hacim> +$9 shipping
233 2011-01-26 02:08:49 <luke-jr> only 500 W
234 2011-01-26 02:08:59 <luke-jr> I'd probably have grabbed that XD
235 2011-01-26 02:09:00 <hacim> should be fine its the 12v that matters
236 2011-01-26 02:09:39 <luke-jr> oh well
237 2011-01-26 02:09:41 <luke-jr> 5850 is its
238 2011-01-26 02:09:43 <luke-jr> it is*
239 2011-01-26 02:38:37 <Kiba``> an open source project
240 2011-01-26 02:38:39 <Kiba``> with no backing
241 2011-01-26 02:38:44 <Kiba``> no insitutional support
242 2011-01-26 02:38:49 <kiba> no money
243 2011-01-26 02:38:52 <kiba> launches a revolution
244 2011-01-26 02:39:32 <kiba> that spans the globe...
245 2011-01-26 02:39:34 <kiba> and that is...bitcoin
246 2011-01-26 02:43:27 <andrew12^mac> !!!!!
247 2011-01-26 02:46:54 <kiba> hmm
248 2011-01-26 02:47:09 <kiba> andrew12^mac: so you read the The Bitcoin Times yet?
249 2011-01-26 02:47:16 <andrew12^mac> no
250 2011-01-26 02:47:46 <andrew12> well i read the first one
251 2011-01-26 02:48:44 <kiba> we got the second issue out
252 2011-01-26 02:48:50 <kiba> available for...free!
253 2011-01-26 02:48:57 <andrew12> :O!
254 2011-01-26 02:49:16 <andrew12> http://twitter.com/theUSpresident
255 2011-01-26 02:49:30 <kiba> why a link to the US president?
256 2011-01-26 02:56:58 <andrew12> kiba: its not the actual president
257 2011-01-26 02:57:04 <andrew12> its a parody account
258 2011-01-26 03:13:24 <luke-jr> jgarzik: you're doing autotools? so no reason for me to re-do my qmake branch in git, right?
259 2011-01-26 03:44:32 <dooglus> jgarzik: have you seen cryptopp/sha.cpp in the source tree?  lots of assembler in there
260 2011-01-26 03:52:33 <jgarzik> dooglus: true.  I forgot about that.  It's in my cpuminer tree too, so I should have remembered :)
261 2011-01-26 03:56:32 <kiba> err
262 2011-01-26 03:56:35 <kiba> extension hacking
263 2011-01-26 03:56:37 <kiba> bitcoin
264 2011-01-26 03:57:20 <kiba> ah shit
265 2011-01-26 03:57:26 <kiba> not today
266 2011-01-26 06:13:13 <sipa> ;;bc,mtgox
267 2011-01-26 06:13:14 <gribble> {"ticker":{"high":0.425,"low":0.3904,"vol":4096,"buy":0.4001,"sell":0.4197,"last":0.4112}}
268 2011-01-26 06:48:12 <tcatm> Yay. It's snowing and GPU temps are <60???C :)
269 2011-01-26 06:48:46 <sipa> where do yoiu live?
270 2011-01-26 06:48:53 <tcatm> germany
271 2011-01-26 07:19:39 <molecular> good morning!
272 2011-01-26 07:20:09 <molecular> one should make a poll, many germans here ;)
273 2011-01-26 07:20:44 <tcatm> There's a poll somewhere in the forum
274 2011-01-26 07:21:35 <molecular> I know from three people here where they come from. all german.
275 2011-01-26 07:21:51 <molecular> one is me, of course ;) so that makes a low sample
276 2011-01-26 07:23:01 <tcatm> We could arrange a meetup.
277 2011-01-26 07:23:24 <molecular> yeah. I fear I'm the only one living in the northern part, though
278 2011-01-26 07:23:41 <molecular> artforz is from close to regensburg somewhere, I'm in Hamburg currently
279 2011-01-26 07:24:00 <molecular> and it's snowing where tcatm is, so probably southish?
280 2011-01-26 07:24:04 <tcatm> I'm near Kiel
281 2011-01-26 07:25:04 <molecular> couldn't have been more wrong ;) snow? hm...
282 2011-01-26 07:25:56 <tcatm> Still > 100km away so it's not surprising it's not snowing in Hamburg.
283 2011-01-26 07:27:01 <molecular> yeah
284 2011-01-26 07:27:42 <molecular> we _could_ have a meeting, I suppose. talk, brag about our miners, order overprized pizza for btc,...
285 2011-01-26 07:28:41 <tcatm> Is there a pizza service that accepts BTC?
286 2011-01-26 07:29:10 <molecular> I don't think so, but there are stories about people paying 10.000 for one pizza, respectively selling it
287 2011-01-26 07:29:43 <molecular> lyspooner told be he sold one for 6.000
288 2011-01-26 07:29:47 <molecular> well... stories.
289 2011-01-26 07:29:53 <molecular> s/be/me
290 2011-01-26 07:31:34 <molecular> diablos miner went all "hog all cpu, hashrate drop way down" on me last night. I think I might switch back to m0's.
291 2011-01-26 07:33:49 <tcatm> Yep, his miner is strange sometimes
292 2011-01-26 07:33:50 <cosurgi> ;;bc,stats
293 2011-01-26 07:33:52 <gribble> Current Blocks: 104659 | Current Difficulty: 18437.64439217 | Next Difficulty At Block: 104831 | Next Difficulty In: 172 blocks | Next Difficulty In About: 23 hours, 47 minutes, and 36 seconds | Next Difficulty Estimate: 21952.52503258
294 2011-01-26 07:34:07 <tcatm> Found me 5 blocks this night, though so it's at least working
295 2011-01-26 07:35:23 <cosurgi> ;;bc,stats
296 2011-01-26 07:35:25 <gribble> Current Blocks: 104661 | Current Difficulty: 18437.64439217 | Next Difficulty At Block: 104831 | Next Difficulty In: 170 blocks | Next Difficulty In About: 23 hours, 28 minutes, and 10 seconds | Next Difficulty Estimate: 21973.30277294
297 2011-01-26 07:35:58 <cosurgi> 2 blocks on less than minutes
298 2011-01-26 07:36:53 <molecular> fortuna has been sitting on them
299 2011-01-26 07:50:37 <cosurgi> ;;bc,target
300 2011-01-26 07:50:38 <gribble> Error: "bc,target" is not a valid command.
301 2011-01-26 07:50:49 <cosurgi> ;;target
302 2011-01-26 07:50:50 <gribble> I do not know about 'target', but I do know about these similar topics: 'trade'
303 2011-01-26 07:51:33 <molecular> http://www.bitcoin.org/wiki/doku.php?id=gribble
304 2011-01-26 07:57:17 <cosurgi> ;;bc,diff
305 2011-01-26 07:57:18 <gribble> 18437.64439217
306 2011-01-26 07:57:28 <cosurgi> thanks.
307 2011-01-26 07:57:39 <cosurgi> I want to check the hex representation of target
308 2011-01-26 08:05:18 <sipa> http://blockexplorer.com/q/hextarget
309 2011-01-26 08:15:42 <cosurgi> thanks
310 2011-01-26 08:16:27 <molecular> I'm unclear about something: How often does the info returned by "getwork" change? only in between blocks or whenever there is a new transaction?
311 2011-01-26 08:16:47 <cosurgi> molecular: I think it changes every second.
312 2011-01-26 08:17:26 <molecular> but then a getwork interval of 5 seconds would make me hash on wrong data for 4 seconds of the 5.
313 2011-01-26 08:17:31 <tcatm> molecular: It should change whenever there's a new TX
314 2011-01-26 08:17:32 <cosurgi> I know that there is a unix time in inpu data for hashing. I'm just not sure if this time is provided by getwork or by client.
315 2011-01-26 08:18:04 <cosurgi> all P2P clients accept small deviation from current time - understanding that there must be some time spent on hashing.
316 2011-01-26 08:18:24 <cosurgi> there is a formula for maximum accepted time deviation, I just don't remember it.
317 2011-01-26 08:18:26 <Diablo-D3> molecular: it changes every time you pull it.
318 2011-01-26 08:18:50 <cosurgi> Diablo-D3: are you happy with you new miner?
319 2011-01-26 08:18:50 <molecular> I'm hashing in slush's pool at 540mhash/s. how often should I pull?
320 2011-01-26 08:18:59 <Diablo-D3> cosurgi: I... guess?
321 2011-01-26 08:19:02 <cosurgi> *your
322 2011-01-26 08:19:04 <Diablo-D3> molecular: the default.
323 2011-01-26 08:19:11 <molecular> ok
324 2011-01-26 08:19:34 <cosurgi> Diablo-D3: nice :) I wasn't sure, due to hasmeter, and also look at this: (let me find)
325 2011-01-26 08:19:40 <cosurgi> 09:31 < molecular> diablos miner went all "hog all cpu, hashrate drop way down" on me last night. I think I might switch back to m0's.
326 2011-01-26 08:19:47 <cosurgi> molecular: care to explain?
327 2011-01-26 08:19:52 <Diablo-D3> sounds like molecular was on drugs.
328 2011-01-26 08:19:59 <cosurgi> :)
329 2011-01-26 08:20:08 <Diablo-D3> I keep telling people to use sdk 2.1
330 2011-01-26 08:20:11 <Diablo-D3> but they dont listen
331 2011-01-26 08:20:15 <molecular> Diablo-D3, your miner had some bad behaviour last night. It went down to 10mhash/s (540 normal) and hogged the cpu (java thread).
332 2011-01-26 08:20:20 <molecular> no drugs, sadly
333 2011-01-26 08:20:30 <molecular> I am using sdk 2.1
334 2011-01-26 08:20:48 <Diablo-D3> if it does it again, just restart it
335 2011-01-26 08:20:53 <molecular> yeah
336 2011-01-26 08:21:10 <molecular> no bad feelings, just thought I'd tell you
337 2011-01-26 08:21:18 <Diablo-D3> the only time Ive ever been able to do that on purpose is when I could trigger the terminal change bug
338 2011-01-26 08:21:28 <cosurgi> tcatm: but you have found 5 blocks with Diablo's miner, right?
339 2011-01-26 08:21:28 <Diablo-D3> and I cant trigger that anymore
340 2011-01-26 08:21:30 <molecular> what's the terminal change bug?
341 2011-01-26 08:21:35 <Diablo-D3> molecular: switch vts
342 2011-01-26 08:21:56 <tcatm> cosurgi: yep
343 2011-01-26 08:22:10 <Diablo-D3> thats more than Ive found with my miner :<
344 2011-01-26 08:22:19 <cosurgi> so the merge with ArtForz code is 100% success :)
345 2011-01-26 08:22:35 <Diablo-D3> cosurgi: seeing as I pushed it to git yesterday, I'd assume so
346 2011-01-26 08:23:19 <molecular> cosurgi: I'm still skeptical about the new miner with art's kernel
347 2011-01-26 08:23:20 <molecular> dunno why
348 2011-01-26 08:23:29 <molecular> I'm using the older one
349 2011-01-26 08:23:31 <cosurgi> Diablo-D3, tcatm: what -f setting do you use?
350 2011-01-26 08:23:43 <tcatm> -w 128 -f 1
351 2011-01-26 08:24:00 <molecular> why 128? not the max 256 ??
352 2011-01-26 08:24:10 <Diablo-D3> 64 is faster
353 2011-01-26 08:24:12 <tcatm> too little registers on GPU
354 2011-01-26 08:24:21 <Diablo-D3> except for a very small number of people, 128 is faster
355 2011-01-26 08:24:53 <tcatm> The kernel uses 23 registers
356 2011-01-26 08:25:06 <tcatm> One CU has a total of 4096 registers for all threads
357 2011-01-26 08:25:13 <Diablo-D3> er, no?
358 2011-01-26 08:25:19 <Diablo-D3> it has 16384.
359 2011-01-26 08:25:51 <Diablo-D3> and thats 16384 vec4s too
360 2011-01-26 08:26:00 <molecular> the hasmeter averages over how long?
361 2011-01-26 08:26:13 <Diablo-D3> molecular: 30 seconds unless you use -f 1
362 2011-01-26 08:26:18 <Diablo-D3> well, below 5
363 2011-01-26 08:26:25 <molecular> and with -f 1 ?
364 2011-01-26 08:26:26 <tcatm> yes, you'Re right. 16k regs
365 2011-01-26 08:26:48 <Diablo-D3> molecular: forever
366 2011-01-26 08:27:21 <molecular> will have to wait quite a while for it to converge then
367 2011-01-26 08:29:49 <Diablo-D3> I find it hilarious, though, this is about 2% short of maximum theoretical speed on my card
368 2011-01-26 08:30:23 <molecular> how do you calc the "theoretical speed"?
369 2011-01-26 08:30:35 <Diablo-D3> art did
370 2011-01-26 08:30:57 <Diablo-D3> its 78.x something on a 4850
371 2011-01-26 08:31:02 <Diablo-D3> I get 76.x
372 2011-01-26 08:31:44 <molecular> actually, my skepticism against your new miner is starting to vanish
373 2011-01-26 08:32:02 <molecular> hashmeter @590, still rising (570 with old miner)
374 2011-01-26 08:32:12 <molecular> nice work!
375 2011-01-26 08:32:29 <Diablo-D3> its going to take awhile for it to settle if you're using -f 1
376 2011-01-26 08:32:45 <molecular> obviously
377 2011-01-26 08:32:48 <Diablo-D3> fast cards + long kernel execution lengths, it'll take awihle for the miner to finally reach that
378 2011-01-26 08:33:17 <molecular> what's nonce saturation??? tried all possible nonces?
379 2011-01-26 08:33:35 <Diablo-D3> yeah
380 2011-01-26 08:33:41 <cosurgi> molecular: yes, there are only 2^32 possible monces. Not much :)
381 2011-01-26 08:33:50 <Diablo-D3> it'll trigger a getwork when that happens
382 2011-01-26 08:34:02 <molecular> good idea, that ;)
383 2011-01-26 08:34:35 <tcatm> It could trigger before that happens to keep GPUs busy
384 2011-01-26 08:34:45 <Diablo-D3> tcatm: three execution queues
385 2011-01-26 08:34:45 <molecular> it will keep hashing while doing getwork, right?
386 2011-01-26 08:35:07 <Diablo-D3> so one being stalled for less than a second isnt going to effect anything
387 2011-01-26 08:35:44 <tcatm> Does it use one getwork for all GPUs?
388 2011-01-26 08:35:48 <Diablo-D3> no
389 2011-01-26 08:35:54 <Diablo-D3> one for each execution thread
390 2011-01-26 08:36:00 <Diablo-D3> so 3*gpu
391 2011-01-26 08:36:11 <tcatm> k
392 2011-01-26 08:36:27 <Diablo-D3> I tried a singular getwork for awhile, but it caused more problems than it cured
393 2011-01-26 08:37:33 <molecular> if I use "-f 2", will I get a "moving average" hashmeter?
394 2011-01-26 08:37:40 <Diablo-D3> no
395 2011-01-26 08:37:43 <Diablo-D3> gotta use 5 or up
396 2011-01-26 08:37:46 <molecular> ok
397 2011-01-26 08:37:54 <molecular> will that hurt performance a lot compared to -f 1 ?
398 2011-01-26 08:38:06 <Diablo-D3> no
399 2011-01-26 08:38:20 <Diablo-D3> but if you're benchmarking its pointless
400 2011-01-26 08:40:04 <molecular> what is best for benchmarking?
401 2011-01-26 08:41:53 <molecular> it just said "nonce saturation" 13 times in a row. is that explainable?
402 2011-01-26 08:42:28 <molecular> keeps saying nonce saturation like 30 times now. still finding some shares, though
403 2011-01-26 08:43:00 <cosurgi> your cards are too fast, compared to getworks. You must getwork more often
404 2011-01-26 08:43:09 <molecular> default is 5?
405 2011-01-26 08:43:20 <cosurgi> I don't know
406 2011-01-26 08:43:34 <cosurgi> you are using slush1's pool?
407 2011-01-26 08:43:38 <molecular> yeah
408 2011-01-26 08:43:54 <molecular> is the nonce smaller when using the pool?
409 2011-01-26 08:43:55 <cosurgi> I don't know how to configure how often to getwork from pool
410 2011-01-26 08:44:10 <cosurgi> it's not. the target is smaller: only 32 bits.
411 2011-01-26 08:44:27 <slush1> I don't think 'nonce saturation' is a problem
412 2011-01-26 08:44:37 <molecular> hey slush
413 2011-01-26 08:44:40 <slush1> it just tell you that new getwork is needed
414 2011-01-26 08:45:23 <molecular> well, but after a new getwork (which it says it does), there should be 2^32 more nonces to mine?
415 2011-01-26 08:46:28 <molecular> slush1: love your cooperative, btw
416 2011-01-26 08:46:35 <molecular> nice page, too
417 2011-01-26 08:47:07 <slush1> yes, new getwork is 2^32 space to hashing
418 2011-01-26 08:48:15 <Diablo-D3> erm
419 2011-01-26 08:48:20 <Diablo-D3> molecular: nonce saturation is meaningless
420 2011-01-26 08:48:23 <Diablo-D3> its just information
421 2011-01-26 08:48:29 <molecular> ok
422 2011-01-26 08:48:31 <Diablo-D3> theres a reason why it says it under debug only
423 2011-01-26 08:48:34 <molecular> :)
424 2011-01-26 08:48:36 <Diablo-D3> if it saturates, it getworks.
425 2011-01-26 08:49:38 <molecular> ok, but still: why does it keep doing that once it starts happening?
426 2011-01-26 08:49:45 <Diablo-D3> because it keeps saturating
427 2011-01-26 08:49:59 <Diablo-D3> I am not surprised a card going almost 600 mhash/sec saturates in 5 seconds
428 2011-01-26 08:50:17 <molecular> this only happened after like 10 minutes of running
429 2011-01-26 08:50:31 <molecular> restart: no nonce saturation message
430 2011-01-26 08:50:54 <Diablo-D3> 2**32 is 4200 million
431 2011-01-26 08:50:58 <Diablo-D3> well, 4294
432 2011-01-26 08:51:11 <molecular> 4200/600 = 7
433 2011-01-26 08:51:16 <molecular> 7 seconds
434 2011-01-26 08:51:51 <Diablo-D3> yes, and it triggers a little early
435 2011-01-26 08:52:13 <Diablo-D3> it really triggers at 2**31
436 2011-01-26 08:52:23 <Diablo-D3> so you have 3.6 seconds.
437 2011-01-26 08:52:25 <molecular> you said before you might implement "early triggering". did you?
438 2011-01-26 08:52:26 <molecular> ah
439 2011-01-26 08:53:36 <Diablo-D3> since you have 3 execution threads, you dont see this like FPS*3 times a second
440 2011-01-26 08:54:24 <Diablo-D3> no thats wrong
441 2011-01-26 08:54:31 <Diablo-D3> it'd be 3 times before 5 seconds are up
442 2011-01-26 08:55:11 <molecular> I should really turn off debug mode, just makes me nervous for no reason
443 2011-01-26 08:56:05 <Diablo-D3> you'd have to have 2 5970s and a 5870 to consistently hit every saturation
444 2011-01-26 08:56:06 <slush1> Diablo-D3: For some strange reason, my miner is crunching quite old jobs. When I have one 5970, average time from getwork to submit was under 5 seconds
445 2011-01-26 08:56:21 <slush1> Diablo-D3: When I have two 5970 in one box, I have times around 10 seconds
446 2011-01-26 08:56:34 <slush1> Maybe it is related to job queues in miner?
447 2011-01-26 08:56:46 <Diablo-D3> slush1: probably not.
448 2011-01-26 08:57:04 <slush1> but it's quite consistent
449 2011-01-26 08:57:10 <Diablo-D3> yeah, but it could be your pool
450 2011-01-26 08:57:14 <slush1> how?
451 2011-01-26 08:57:29 <Diablo-D3> that'd be 12 threads getworking simultaniously
452 2011-01-26 08:57:56 <slush1> when miner ask for getwork and calculate it immediately, I should have very tiny times per job
453 2011-01-26 08:58:22 <Diablo-D3> yes it should
454 2011-01-26 08:58:22 <slush1> so it looks like miner ask for getwork, but don't calculate it immediately
455 2011-01-26 08:58:46 <Diablo-D3> the execution thread asks for getwork, blocks waiting for it, and then goes back to mining
456 2011-01-26 08:58:56 <slush1> with one 5970, I had also times under second. Now when it hit 5 second per job, it is rare
457 2011-01-26 08:59:18 <slush1> yes, but roundtrip is 40ms
458 2011-01-26 08:59:58 <Diablo-D3> you're measuring only submits being 10 seconds old?????????????????????????????????
459 2011-01-26 09:00:04 <Diablo-D3> er, goddamnit ? key stuck
460 2011-01-26 09:00:19 <Diablo-D3> IRC + -f 1 may not be best combination
461 2011-01-26 09:00:26 <slush1> no, I'm measuring all submits from my miner right now
462 2011-01-26 09:00:33 <Diablo-D3> hrm
463 2011-01-26 09:00:35 <slush1> and they are typically >7s
464 2011-01-26 09:00:42 <Diablo-D3> 7s I can almost see
465 2011-01-26 09:01:06 <slush1> my getwork rate is 5 sec, but there should be also much faster submits, you know
466 2011-01-26 09:01:17 <Diablo-D3> slush1: well
467 2011-01-26 09:01:25 <Diablo-D3> you have 3 execution threads EACH taking, say, a second
468 2011-01-26 09:01:26 <slush1> but I have no explanation for 14 second from getwork to submit...
469 2011-01-26 09:01:29 <Diablo-D3> when doing -f 1
470 2011-01-26 09:02:12 <Diablo-D3> so 5 seconds almost are up, kernel is queued, two other kernels complete before it, it completes, you're at second 8
471 2011-01-26 09:02:13 <slush1> yep, I have -f 1
472 2011-01-26 09:02:35 <sipa> 14s delays = 2.5% extra chance for stale blocks...
473 2011-01-26 09:03:32 <Diablo-D3> slush1: see if it still does it when doing -f 1000
474 2011-01-26 09:04:19 <molecular> I dont know how this relates to "gewort to submit"-measurements: just did 50 shares in 390 seconds. that's 7.8 seconds per share on avg. seems a bit low considering slush said something about getting <5s average ?
475 2011-01-26 09:04:35 <sipa> what's your hashrate?
476 2011-01-26 09:04:41 <Diablo-D3> molecular: whats your hashrate?
477 2011-01-26 09:04:45 <molecular> 606mhash/s
478 2011-01-26 09:04:49 <sipa> ;;bc,calcd 606000 1
479 2011-01-26 09:04:50 <gribble> The average time to generate a block at 606000 Khps, given the supplied difficulty of 1, is 7 seconds
480 2011-01-26 09:04:58 <molecular> hmm, ok
481 2011-01-26 09:05:11 <Diablo-D3> ;;bc,calcd 75000 1
482 2011-01-26 09:05:11 <gribble> The average time to generate a block at 75000 Khps, given the supplied difficulty of 1, is 57 seconds
483 2011-01-26 09:05:17 <Diablo-D3> that calc is off
484 2011-01-26 09:05:30 <molecular> how the hell did you do <5s average on a 5970?
485 2011-01-26 09:05:49 <gribble> (bc,calcd <an alias, 2 arguments>) -- Alias for "echo The average time to generate a block at $1 Khps, given the supplied difficulty of $2, is [time elapsed [math calc 1/((2**224-1)/$2*$1*1000/2**256)]]".
486 2011-01-26 09:05:49 <sipa> ;;bc,calcd
487 2011-01-26 09:05:55 <Diablo-D3> wait, its right
488 2011-01-26 09:06:19 <sipa> that formula is actually not entirely correct
489 2011-01-26 09:06:40 <ArtForz> close enough for all insensitive purposes
490 2011-01-26 09:06:55 <molecular> moin artforz
491 2011-01-26 09:07:14 <sipa> ok 0.0015% off
492 2011-01-26 09:07:16 <sipa> good enough :D
493 2011-01-26 09:07:22 <ArtForz> moin moin
494 2011-01-26 09:07:26 <Diablo-D3> molecular: 2**31/(606m*3)
495 2011-01-26 09:07:34 <Diablo-D3> er
496 2011-01-26 09:07:56 <Diablo-D3> /3
497 2011-01-26 09:08:02 <ArtForz> huh?
498 2011-01-26 09:08:16 <Diablo-D3> molecular: thats how long it'd take you to do nonce saturation, not find a block
499 2011-01-26 09:08:27 <ArtForz> ah, k
500 2011-01-26 09:08:32 <molecular> 3.54
501 2011-01-26 09:08:39 <Diablo-D3> and thats to guarantee nonce saturation
502 2011-01-26 09:08:46 <Diablo-D3> no, 10.63
503 2011-01-26 09:09:08 <molecular> <Diablo-D3> /3
504 2011-01-26 09:09:24 <molecular> the 606  is the combined rate of the queues ?
505 2011-01-26 09:09:28 <sipa> 2**32/(606M/3)
506 2011-01-26 09:09:34 <Diablo-D3> sipa: 31
507 2011-01-26 09:09:38 <sipa> oh, yes
508 2011-01-26 09:09:44 <sipa> 2**31/(606M/3)
509 2011-01-26 09:09:52 <ArtForz> ignoring the top bit or something?
510 2011-01-26 09:09:57 <Diablo-D3> ArtForz: yeah, its signed
511 2011-01-26 09:10:00 <Diablo-D3> yay java
512 2011-01-26 09:10:08 <ArtForz> *headdesk*
513 2011-01-26 09:10:10 <Diablo-D3> its not like it matter
514 2011-01-26 09:10:29 <Diablo-D3> on a 5870 speed chip, you're not going to see nonce saturation often
515 2011-01-26 09:10:43 <ArtForz> you shouldnt really see it at all
516 2011-01-26 09:11:11 <Diablo-D3> ArtForz: molecular doesnt
517 2011-01-26 09:11:17 <Diablo-D3> he was just confused at why my miner said it
518 2011-01-26 09:12:23 <ArtForz> btw, tweaked my CAL kernel a bit yesterday, still same speed, but 3% lower power consumption
519 2011-01-26 09:12:28 <Diablo-D3> LOL
520 2011-01-26 09:12:42 <Diablo-D3> ArtForz: you said my card maxes at 78.x right?
521 2011-01-26 09:12:58 <ArtForz> 4850?
522 2011-01-26 09:13:01 <Diablo-D3> yeah
523 2011-01-26 09:13:17 <ArtForz> havent done the math for 4xxx with precalculating as much shit as possible
524 2011-01-26 09:13:23 <Diablo-D3> heh
525 2011-01-26 09:13:28 <Diablo-D3> its doing 76.x now
526 2011-01-26 09:13:38 <ArtForz> with full double-sha2356 it's 75.3Mh/s
527 2011-01-26 09:13:50 <ArtForz> *256
528 2011-01-26 09:14:26 <ArtForz> 78.67 leaving out the first and last 3 rounds
529 2011-01-26 09:14:26 <Diablo-D3> yeah, and I was getting 74.6 with 6 rounds chopped off
530 2011-01-26 09:14:33 <Diablo-D3> molecular: not worth it
531 2011-01-26 09:14:40 <molecular> why not?
532 2011-01-26 09:14:44 <Diablo-D3> because its meaningless
533 2011-01-26 09:15:38 <ArtForz> 80.1 or so with precalculating partial rounds
534 2011-01-26 09:15:39 <molecular> it's directly related to BTC/day
535 2011-01-26 09:16:20 <ArtForz> thats theoretical peak assuming 100% VLIW utilization
536 2011-01-26 09:16:51 <ArtForz> so 76.x is really pretty damn good
537 2011-01-26 09:17:15 <ArtForz> thats like 95% peak
538 2011-01-26 09:17:32 <molecular> If I overclock too hard, might I get wrong hashes or will it always freeze?
539 2011-01-26 09:17:41 <ArtForz> you might get wrong hashes
540 2011-01-26 09:17:51 <Diablo-D3> molecular: you probably WILL get wrong hashes
541 2011-01-26 09:17:56 <Diablo-D3> which is why my miner has a check for it
542 2011-01-26 09:18:00 <molecular> so I might still see 600mhash/s but no shares found ?
543 2011-01-26 09:18:08 <Diablo-D3> unlikely
544 2011-01-26 09:18:14 <Diablo-D3> my miner will bitch very loudly
545 2011-01-26 09:18:20 <ArtForz> depends
546 2011-01-26 09:18:28 <ArtForz> it only complains on false positive
547 2011-01-26 09:18:39 <ArtForz> false negative is a LOT more likely
548 2011-01-26 09:18:46 <ArtForz> by about a factor of 2**32
549 2011-01-26 09:18:48 <Diablo-D3> yes, but you wont get away with zero false positives
550 2011-01-26 09:19:21 <ArtForz> depends really
551 2011-01-26 09:19:46 <ArtForz> assuming randomly distributed errors, int's about 1 : 2**32 false pos : false neg
552 2011-01-26 09:20:01 <Diablo-D3> yeah
553 2011-01-26 09:20:02 <Diablo-D3> but
554 2011-01-26 09:20:21 <molecular> you mean this: "error("Invalid block found on " + deviceName + ", possible driver or hardware issue");" ?
555 2011-01-26 09:20:30 <Diablo-D3> hes doing a h == 0 every 7 seconds
556 2011-01-26 09:20:34 <ArtForz> yup
557 2011-01-26 09:20:38 <Diablo-D3> molecular: yes
558 2011-01-26 09:20:49 <ArtForz> srsly, why not add a test mode?
559 2011-01-26 09:20:52 <Diablo-D3> if your hardware is fucking up at that speed, _you will know about it_
560 2011-01-26 09:20:57 <Diablo-D3> ArtForz: meh
561 2011-01-26 09:20:59 <molecular> what's that: "if(G <= currentWork.target[6]) {" (around the H==0)
562 2011-01-26 09:21:12 <ArtForz> feed 1 or 2 known good block headers with varying starting nonce, see if you get the expected result
563 2011-01-26 09:21:24 <Diablo-D3> molecular: checks if it meets the target.
564 2011-01-26 09:21:44 <molecular> might a hardware error cause the target to not be met?
565 2011-01-26 09:21:48 <ArtForz> thats exactly what I'm doing with my burn-in test
566 2011-01-26 09:22:02 <Diablo-D3> molecular: I check if H even == 0, if it DOESNT, then theres a hw issue
567 2011-01-26 09:22:07 <ArtForz> for known good block headers, see blk0001.dat ;)
568 2011-01-26 09:23:24 <molecular> but if ! (G <= currentWork.target[6])) it won't even check H == 0
569 2011-01-26 09:23:32 <molecular> but silently continue
570 2011-01-26 09:23:47 <ArtForz> I mangle the starting nonces so every enqueuekernel a different VLIW unit gets the "good" hash
571 2011-01-26 09:24:07 <molecular> that's why I asked if hardware trouble can cause (G <= currentWork.target[6]) == false
572 2011-01-26 09:24:23 <ArtForz> finds hardware problems a LOT quicker than just waiting for a false pos to show up
573 2011-01-26 09:24:23 <Diablo-D3> molecular: but the hardware doesnt check for G
574 2011-01-26 09:24:31 <Diablo-D3> molecular: it only checks for H
575 2011-01-26 09:24:44 <ArtForz> another good test: pyrit-calpp in benchmark mode
576 2011-01-26 09:25:11 <Diablo-D3> molecular: the only thing that both the host and the opencl do is check for H
577 2011-01-26 09:25:22 <molecular> ok, that answers the question, thx
578 2011-01-26 09:25:37 <ArtForz> if pyrit says hardware error, back down the clock
579 2011-01-26 09:25:43 <Diablo-D3> so yes, a hardware error can cause a wrong G
580 2011-01-26 09:25:55 <Diablo-D3> its just going to be very fucking hard to do without also misidentifying H
581 2011-01-26 09:26:30 <ArtForz> quite a few of my 5970s start producing incorrect reults 20MHz+ below "locks up" clock
582 2011-01-26 09:27:22 <ArtForz> thing is, even with some errors it might still be a net gain
583 2011-01-26 09:27:54 <Diablo-D3> ArtForz: heh
584 2011-01-26 09:28:02 <molecular> btw: restarting X causes the fan-control to go back to auto -> nice
585 2011-01-26 09:28:03 <Diablo-D3> theres an entire school of thought that believes that
586 2011-01-26 09:28:18 <Diablo-D3> molecular: yes, it reinitializes the chip
587 2011-01-26 09:28:26 <sipa> is there no pplib command to let the fans go back to auto?
588 2011-01-26 09:28:31 <ArtForz> nope
589 2011-01-26 09:28:37 <ArtForz> well, there is one, but it's broken
590 2011-01-26 09:28:41 <ArtForz> it sets fan speed to 0%
591 2011-01-26 09:28:56 <molecular> I asked how to put it back to auto a couple times here. someone said: not possible, need reboot. glad it's not true
592 2011-01-26 09:28:56 <sipa> which is rarely what you want
593 2011-01-26 09:29:06 <ArtForz> weird thing is, it works fine with ADL lib calls
594 2011-01-26 09:29:54 <molecular> btw artforz: gave up on the fan_control linking problem for now. thanks for the code anyways
595 2011-01-26 09:30:23 <molecular> I think the function signatures are different, but don't know how to read them form the .so
596 2011-01-26 09:30:29 <ArtForz> really weird, it compiles and links fine on debian sid x86 and amd64 here
597 2011-01-26 09:30:44 <molecular> yeah, someone else tried it yesterday, worked for him
598 2011-01-26 09:31:06 <molecular> I think it was tcatm
599 2011-01-26 09:31:27 <molecular> I might have some weird version of the ati-driver. gentoo says "ati-driver 10.11"
600 2011-01-26 09:31:37 <ArtForz> shouldnt really matter
601 2011-01-26 09:31:46 <molecular> but that's where libatiadlxx.so comes from
602 2011-01-26 09:31:51 <ArtForz> the ADL interfaces I'm using havent changed in years
603 2011-01-26 09:32:08 <molecular> puh. them maybe some gcc issue
604 2011-01-26 09:32:16 <molecular> *then
605 2011-01-26 09:32:52 <molecular> hm, you could send me your libatiadlxx.so, just to check if I can link to that...
606 2011-01-26 09:33:38 <ArtForz> wild guess: you're missing some dynlib libatiadlxx.so links against
607 2011-01-26 09:33:52 <ArtForz> or at least miss it in standard lib path
608 2011-01-26 09:33:52 <sipa> molecular: still that problem with linking?
609 2011-01-26 09:34:00 <sipa> about the missing symbol
610 2011-01-26 09:34:15 <ArtForz> it's not missing, thats the weird thing
611 2011-01-26 09:34:33 <ArtForz> what does ldd /usr/lib/libatiadlxx.so say?
612 2011-01-26 09:35:07 <molecular> a lot of stuff like: "libGL.so.1 => /usr/lib/opengl/ati/lib/libGL.so.1 (0xb760b000)"
613 2011-01-26 09:35:15 <sipa> are you both using the same architecture? (32-bit, 64-bit, ...) ?
614 2011-01-26 09:35:24 <ArtForz> does it say that for ALL linked libs?
615 2011-01-26 09:35:48 <molecular> what do you mean "All linked libs"?
616 2011-01-26 09:35:50 <ArtForz> iirc it says "missing" or something if one linked lib isn't found
617 2011-01-26 09:36:02 <sipa> maybe that extern defined symbol in the C code has some types wrong, but are accidentaly correct for the platform ArtForz is using
618 2011-01-26 09:36:19 <molecular> http://pastebin.com/bdQGsv38
619 2011-01-26 09:36:24 <ArtForz> my code compiles and links on x86 and x64
620 2011-01-26 09:36:33 <sipa> ok, nvm then
621 2011-01-26 09:36:40 <ArtForz> against 10.9, 10.10, 10.11 and 10.12
622 2011-01-26 09:37:09 <ArtForz> hurrrrm
623 2011-01-26 09:37:21 <ArtForz> libGL.so.1 => /usr/lib/opengl/ati/lib/libGL.so.1
624 2011-01-26 09:37:36 <ArtForz> I'm pretty sure thats not in the standard lib path
625 2011-01-26 09:38:03 <molecular> can't be gcc version, can it? (4.4.4)
626 2011-01-26 09:38:04 <molecular> does anyone know how to check the complete type-list (signature?) of a function in a lib?
627 2011-01-26 09:38:23 <molecular> I tried objdump, mn, dumpelf
628 2011-01-26 09:38:29 <sipa> is it C or C++?
629 2011-01-26 09:38:34 <ArtForz> C
630 2011-01-26 09:38:49 <sipa> then types arent encoded in the function name
631 2011-01-26 09:38:51 <ArtForz> try adding a -L /usr/lib/opengl/ati/lib/
632 2011-01-26 09:38:53 <lfm> libGL.so.1 => /usr/lib/libGL.so.1 (0x00007f7743d98000) works ok on my system
633 2011-01-26 09:38:55 <sipa> so no way to know
634 2011-01-26 09:39:01 <ArtForz> yeah, same here
635 2011-01-26 09:39:02 <molecular> gentoo does some shit allowing you to switch opengl implementation
636 2011-01-26 09:39:07 <UukGoblin> molecular, I don't think it's exactly possible without disassembling
637 2011-01-26 09:39:10 <ArtForz> yay for putting libs i nweird places
638 2011-01-26 09:39:33 <UukGoblin> there's patchelf that can change the location of the libs in an elf
639 2011-01-26 09:39:46 <molecular> "-L /usr/lib/opengl/ati/lib/" when linking?
640 2011-01-26 09:39:49 <ArtForz> yep
641 2011-01-26 09:39:52 <ArtForz> my lGL is in /usr/lib
642 2011-01-26 09:40:29 <ArtForz> it's a ati-specific libGL
643 2011-01-26 09:41:53 <molecular> still no luck: http://pastebin.com/cEcFfC32
644 2011-01-26 09:43:23 <ArtForz> looks like you're picking up a cpp lib comewhere
645 2011-01-26 09:44:24 <molecular> where do you see that?
646 2011-01-26 09:44:28 <UukGoblin> there's these -rpath and -rpath-link options
647 2011-01-26 09:44:43 <ArtForz> -_gxx_personality_v0 is a libstdc++ symbol
648 2011-01-26 09:45:43 <UukGoblin> passed via -Wl,rpath IIRC
649 2011-01-26 09:45:44 <molecular> oh f.ck
650 2011-01-26 09:46:00 <molecular> hhhhrm. I named the file ati_fan_control.cpp initially
651 2011-01-26 09:46:06 <molecular> then used some generic Makefile
652 2011-01-26 09:46:24 <molecular> renamed to .c last night, still didn't work
653 2011-01-26 09:46:31 <ArtForz> clean your objects, use plain ole gcc
654 2011-01-26 09:46:31 <molecular> now I cleaned everything and it works
655 2011-01-26 09:46:34 <molecular> thanks for that hint
656 2011-01-26 09:46:48 <molecular> how does gcc detect cpp/c ?
657 2011-01-26 09:47:17 <molecular> puuuh, thanks a lot
658 2011-01-26 09:47:19 <ArtForz> no clue really
659 2011-01-26 09:47:37 <molecular> maybe that "generic Makefile" did that somehow
660 2011-01-26 09:47:53 <ArtForz> you probably still had a c++ .o floating about
661 2011-01-26 09:48:18 <molecular> I recompiled a lot after renaming to .c
662 2011-01-26 09:48:49 <molecular> well, don't really care... breakfast now, then will look at your fan_control ;)
663 2011-01-26 10:08:18 <molecular> well, now I remember... in cpp you have to use 'extern "C" ...'
664 2011-01-26 10:08:42 <Diablo-D3> so
665 2011-01-26 10:08:47 <Diablo-D3> today's fredo and pidjin
666 2011-01-26 10:13:24 <pozic> My miner suddently uses 100% CPU. Does that mean it found something?
667 2011-01-26 10:13:41 <pozic> I use the poclbm miner.
668 2011-01-26 10:14:37 <pozic> It had been running for > 40 hours without any issue.
669 2011-01-26 10:15:53 <Diablo-D3> ati?
670 2011-01-26 10:20:28 <noagendamarket> did you delete system 32 ?
671 2011-01-26 10:20:54 <cosurgi> I just noticed one thing: with old Diablo's miner all GPUs load was almost always 99%, with new one - one of three GPUs load is 96%
672 2011-01-26 10:21:44 <cosurgi> other two stay at 99%
673 2011-01-26 10:22:42 <pozic> Diablo-D3: yes
674 2011-01-26 10:23:04 <pozic> noagendamarket: no idea what you mean.
675 2011-01-26 10:23:19 <cosurgi> it's a hole on the bottom, where the second GPU is, the case is lying on its side.
676 2011-01-26 10:27:45 <lfm> what sorta case is it? not a fancy one I assume
677 2011-01-26 10:30:40 <xelister> anyone seen too high clock?
678 2011-01-26 10:30:45 <xelister> what does too high clock cause on 5970?
679 2011-01-26 10:30:58 <xelister> or - how do I know that given clock is too high (for mining). The box is also used as desktop
680 2011-01-26 10:31:32 <lfm> erratic stability mostly
681 2011-01-26 10:31:49 <noagendamarket> the ultimate case is no case at all ?
682 2011-01-26 10:32:18 <lfm> like the antec skeleton case?
683 2011-01-26 10:32:19 <noagendamarket> air cooled to the max...
684 2011-01-26 10:32:40 <lfm> not the quietest tho
685 2011-01-26 10:32:48 <Diablo-D3> pozic: 10.12?
686 2011-01-26 10:33:16 <Diablo-D3> cosurgi: gpu load is not a measurement of performance
687 2011-01-26 10:33:55 <xelister> is there any GPU miner that does verify all the time validity of all calculated things?
688 2011-01-26 10:35:13 <lfm> if you match the referance imlemnation for one hash the odds of your implementation being wrong are astronomically small
689 2011-01-26 10:35:38 <xelister> lfm: but I ment to test gpu / opencl stability-validaity
690 2011-01-26 10:35:48 <pozic> Diablo-D3: Fedora 14
691 2011-01-26 10:37:08 <cosurgi> lfm: I go the cheapest route, all componentes except for PSU are the cheapest ones. sempron CPU, cheapest 5850 that I could find (2 year warranty, though).
692 2011-01-26 10:37:46 <cosurgi> of course cheapest case :)
693 2011-01-26 10:38:02 <lfm> so you might consider springing for a slightly better case next time?
694 2011-01-26 10:38:13 <pozic> A case is a good investment.
695 2011-01-26 10:38:25 <pozic> You can reuse it for a different machine later on.
696 2011-01-26 10:38:56 <cosurgi> no, I have nothing against cutting holes in the case. I did this already to my backup server, which serves me 4 years already and is doing rsnapshot backups three times a day of all my 7 boxes scattered around town.
697 2011-01-26 10:39:00 <lfm> otoh I like the coormaster elite 330, cheap and seems to have decent circulation
698 2011-01-26 10:39:27 <lfm> and you dont pay for silly clear sides and so on
699 2011-01-26 10:40:20 <lfm> coormaster -> coolermaster
700 2011-01-26 10:40:25 <cosurgi> well, I put them far away, where noise is not a concern, add lots of fans and and UPS, and forget. They keep running. backup server has about 10 fans in it, and 9 HDDs :) raid6.
701 2011-01-26 10:41:06 <cosurgi> its not a clean way, but cheap and robust.
702 2011-01-26 10:41:54 <cosurgi> and it's actually a P3 :)
703 2011-01-26 10:42:20 <lfm> oh never mind then, no one and nothing can help a p3
704 2011-01-26 10:42:27 <cosurgi> um, no. sorry.
705 2011-01-26 10:42:30 <cosurgi> model name      : Intel(R) Pentium(R) 4 CPU 2.00GHz
706 2011-01-26 10:42:34 <cosurgi> a P4.
707 2011-01-26 10:42:54 <cosurgi> I wonder where did my P3 go?
708 2011-01-26 10:44:23 <lfm> ok I have a soft spot for p4s. I have had several of them given to me and I didnt even tear all of them apart for parts yet
709 2011-01-26 10:50:22 <molecular> artforz: fan_control working nicely... don't see a need to fiddle with Kp, Ki, Kd
710 2011-01-26 10:50:34 <ArtForz> well, try step load change
711 2011-01-26 10:50:49 <molecular> step load?
712 2011-01-26 10:50:56 <ArtForz> idle to 100% load and back
713 2011-01-26 10:51:17 <molecular> ok
714 2011-01-26 10:51:22 <cosurgi> ;;bc,stats
715 2011-01-26 10:51:24 <gribble> Current Blocks: 104680 | Current Difficulty: 18437.64439217 | Next Difficulty At Block: 104831 | Next Difficulty In: 151 blocks | Next Difficulty In About: 20 hours, 55 minutes, and 49 seconds | Next Difficulty Estimate: 21928.87509190
716 2011-01-26 10:51:37 <ArtForz> thats why I added the "force 100% fan if current temp > targettemp + 5???C" safety
717 2011-01-26 10:51:59 <lfm> ya it seems to work fine for me too but 77 degrees made fan zero so I reduced temp to 65 for result fan 30 percent
718 2011-01-26 10:52:00 <molecular> yeah, saw that
719 2011-01-26 10:52:10 <molecular> fan at 0% now
720 2011-01-26 10:52:18 <ArtForz> with decent case ventilation and card idle it likes to settle for 0% fan speed
721 2011-01-26 10:52:21 <pozic> The difficulty is going up 10fold?
722 2011-01-26 10:52:28 <ArtForz> and then takes ages to ramp back up normally
723 2011-01-26 10:52:28 <molecular> hm, actually pplib still reports 19%
724 2011-01-26 10:52:33 <pozic> Never mind
725 2011-01-26 10:52:46 <ArtForz> it's because the fan controller internally ramps speed
726 2011-01-26 10:53:19 <ArtForz> setting fan speed 0->100% or 100->0 takes about 30 seconds to actually get there
727 2011-01-26 10:53:46 <molecular> which might be too slow when I start the miner?
728 2011-01-26 10:53:48 <lfm> oh its not due to zero speed stall start mechanical effect you think?
729 2011-01-26 10:54:00 <ArtForz> nope
730 2011-01-26 10:54:14 <ArtForz> fan controller hardware has a slope limit
731 2011-01-26 10:54:39 <molecular> ok, the safety kicked in
732 2011-01-26 10:54:44 <ArtForz> yep
733 2011-01-26 10:54:56 <lfm> prolly a noise control measure so the jet blast doesnt startle people
734 2011-01-26 10:55:00 <molecular> one could shorten the buffer or increase the frequency
735 2011-01-26 10:55:09 <ArtForz> 1.0/0.5/1.0 is on the sloooow side to avoid oscillating
736 2011-01-26 10:55:49 <molecular> does this way of regulating have a name?
737 2011-01-26 10:55:58 <ArtForz> yes
738 2011-01-26 10:56:02 <ArtForz> http://en.wikipedia.org/wiki/PID_controller
739 2011-01-26 10:57:27 <ArtForz> probably increase Kp and lower Ki
740 2011-01-26 10:58:09 <molecular> if I adjust the refresh-rate (sleep(3) -> sleep(1)) will I have to adjust the P_x to get the same result?
741 2011-01-26 10:58:13 <ArtForz> yep
742 2011-01-26 10:58:23 <molecular> uhm, divide all three by 3?
743 2011-01-26 10:58:31 <pozic> So, poclbm does not run stable here. After 40 hours it uses 100% CPU. Did anyone else also experience that?
744 2011-01-26 10:59:00 <ArtForz> iirc leave p and k alone, i / 3
745 2011-01-26 10:59:11 <ArtForz> errr, leave p and d alone
746 2011-01-26 10:59:55 <molecular> ah, of course... makes sense
747 2011-01-26 11:01:25 <ArtForz> it's kinda fun to play with params
748 2011-01-26 11:02:02 <ArtForz> once could also factor GPU load into it
749 2011-01-26 11:02:20 <ArtForz> which would kinda make sense
750 2011-01-26 11:02:22 <molecular> one could also adjust frequency and have noise level as goal ;)
751 2011-01-26 11:02:37 <ArtForz> yep
752 2011-01-26 11:02:55 <ArtForz> adjust frequency to regulate card temp at constant fan speed :P
753 2011-01-26 11:03:07 <molecular> and power consumption, price and mtgox exchange rate, hehe
754 2011-01-26 11:03:37 <molecular> stichwort "smart meter"
755 2011-01-26 11:04:26 <molecular> well, with difficulty rising and assuming btc value will rise, the only sane way is to go full power anyway
756 2011-01-26 11:04:41 <ArtForz> yep
757 2011-01-26 11:05:52 <lfm> molecular full pwer or off
758 2011-01-26 11:05:59 <molecular> actually, initially I was just going to simply _buy_ some btc
759 2011-01-26 11:06:14 <molecular> then I thought it might be more entertaining to mine a little
760 2011-01-26 11:06:24 <molecular> little late to get in the game, though
761 2011-01-26 11:07:41 <ArtForz> depends really on electricity price
762 2011-01-26 11:09:04 <molecular> yeah, well, I live in germany
763 2011-01-26 11:09:05 <molecular> also depends on btc/eur value
764 2011-01-26 11:09:20 <ArtForz> yeah, probably not a too good idea
765 2011-01-26 11:09:45 <ArtForz> $0.30+/kWh = suck
766 2011-01-26 11:10:01 <molecular> jup
767 2011-01-26 11:10:29 <molecular> might send miner to the states or somewhere when time comes. probably just sell the card for like nothing and let it go, though
768 2011-01-26 11:10:33 <ArtForz> it's like $0.08/kWh in canuckistan
769 2011-01-26 11:10:39 <pozic> How can the German economy run on 0.3euro/kWh?
770 2011-01-26 11:10:42 <molecular> ca-what?
771 2011-01-26 11:10:47 <ArtForz> canada :P
772 2011-01-26 11:10:53 <pozic> er dollar
773 2011-01-26 11:11:08 <ArtForz> how it runs? fine. :P
774 2011-01-26 11:11:29 <pozic> In France it is like 0.08 euro
775 2011-01-26 11:11:33 <molecular> exporting alternative energy solutions
776 2011-01-26 11:11:57 <lfm> france exporting neuclear solutions
777 2011-01-26 11:12:00 <molecular> until one of your reactors blows ;)
778 2011-01-26 11:12:09 <pozic> molecular: they cannot blow anymore.
779 2011-01-26 11:12:22 <molecular> no?
780 2011-01-26 11:12:35 <molecular> bad things can't happen? then the only problem is waste
781 2011-01-26 11:12:38 <pozic> molecular: there are also fusion designs which are even safer with a higher yield.
782 2011-01-26 11:12:39 <lfm> pozic kinda depends how stupid the operators can get
783 2011-01-26 11:12:42 <ArtForz> my chart says ~0.12EUR/kWh in .fr
784 2011-01-26 11:12:53 <pozic> ArtForz: your chart is wrong.
785 2011-01-26 11:13:16 <pozic> It depends probably on how much you use, though.
786 2011-01-26 11:13:23 <lfm> may vary by province
787 2011-01-26 11:13:46 <ArtForz> probably does
788 2011-01-26 11:17:44 <pozic> Alternative energy sounds nice, but as long as we don't have near free paint for solar and a way to store the energy it doesn't work.
789 2011-01-26 11:18:01 <ArtForz> depends
790 2011-01-26 11:18:08 <pozic> MIT invented something like that, but if it was so easy to produce I don't see why we don't have it yet.
791 2011-01-26 11:18:13 <Diablo-D3> http://feedproxy.google.com/~r/boingboing/iBag/~3/TWUGjs_8d0M/provocative-metaphor.html
792 2011-01-26 11:18:17 <Diablo-D3> what.
793 2011-01-26 11:18:23 <molecular> I kinda like the idea using electric cars for decentralized storage
794 2011-01-26 11:18:41 <Diablo-D3> ArtForz: btw
795 2011-01-26 11:18:45 <Diablo-D3> the real question is
796 2011-01-26 11:18:48 <Diablo-D3> how fucking long is that 6990
797 2011-01-26 11:18:54 <Diablo-D3> the photo makes it look huge
798 2011-01-26 11:19:17 <ArtForz> looks shorter than a 5970
799 2011-01-26 11:19:27 <noagendamarket> http://www.net-security.org/secworld.php?id=10493   heh
800 2011-01-26 11:19:52 <ArtForz> hard to tell at that weird angle
801 2011-01-26 11:20:11 <Diablo-D3> shorter? seriously?
802 2011-01-26 11:20:25 <ArtForz> yep
803 2011-01-26 11:20:44 <Diablo-D3> I dont think that card would fit in my case
804 2011-01-26 11:21:02 <ArtForz> http://it.oc.com.tw/images/1101/201163044121194732630447638.jpg
805 2011-01-26 11:21:06 <ArtForz> check length of card relative to PCIe connector
806 2011-01-26 11:21:08 <Diablo-D3> no theres a better image out
807 2011-01-26 11:21:10 <noagendamarket> http://www.blogcdn.com/www.engadget.com/media/2011/01/11x0126h83amd.jpg
808 2011-01-26 11:21:18 <Diablo-D3> http://www.engadget.com/2011/01/26/amd-radeon-hd-6990-shows-up-in-its-metallic-flesh-looking-large/
809 2011-01-26 11:21:22 <noagendamarket> it looks huge
810 2011-01-26 11:21:23 <Diablo-D3> yeah that one
811 2011-01-26 11:21:29 <Diablo-D3> it looks like a fucking xbox
812 2011-01-26 11:21:33 <noagendamarket> lmao
813 2011-01-26 11:21:55 <noagendamarket> you could fit the whole computer inside the card lol
814 2011-01-26 11:22:04 <ArtForz> eww
815 2011-01-26 11:22:09 <noagendamarket> maybe theyare gonna sell them as cases
816 2011-01-26 11:22:21 <Diablo-D3> small mexican families could live in those
817 2011-01-26 11:22:30 <noagendamarket> lol
818 2011-01-26 11:22:33 <ArtForz> if thats really the cooler design, say goodbye to decent cooling performance
819 2011-01-26 11:22:45 <lfm> seems like the new single slot width versions could be more power dense
820 2011-01-26 11:23:00 <ArtForz> XFX custom 5970s had the same braindead arrangement
821 2011-01-26 11:23:39 <noagendamarket> maybe you can remove the cover....
822 2011-01-26 11:23:42 <molecular> looks like 5970 sizewise to me
823 2011-01-26 11:24:30 <Diablo-D3> its gotta be bigger
824 2011-01-26 11:24:36 <Diablo-D3> because Im pretty sure I can fit a 5970 in my case
825 2011-01-26 11:24:39 <Diablo-D3> theres no way thats fitting
826 2011-01-26 11:24:53 <noagendamarket> you could stick wheels on it and carry your luggage to the airport
827 2011-01-26 11:25:09 <ArtForz> let's see
828 2011-01-26 11:25:17 <Diablo-D3> dude, it'd stick wings on it and fly it away from the airport
829 2011-01-26 11:26:57 <noagendamarket> the engine might overheat though...it doesnt look like much air cooling for such a huge card
830 2011-01-26 11:27:28 <ArtForz> length of PCIe conn in that pic 101 px, card 292 px
831 2011-01-26 11:27:29 <noagendamarket> maybe its a dyson and it  sucks lie a vacuum cleaner?
832 2011-01-26 11:27:35 <noagendamarket> lol
833 2011-01-26 11:28:24 <ArtForz> random 5970 pic, PCIe conn 173px, card 511px
834 2011-01-26 11:29:11 <ArtForz> so 2.9 PCIe conn length for 6990, ~2.95 for 5970
835 2011-01-26 11:30:08 <ArtForz> so about the same length as a 5970, maybe a tad shorter
836 2011-01-26 11:30:11 <molecular> ArtForz, I modded your safety-logic: "if ( (newfanspeed == 0 && curtemp > targettemp - 10.0) || curtemp > targettemp + 5.0 ) {" -> starts up earlier and still doesn't interfere with normal operation
837 2011-01-26 11:30:36 <ArtForz> neat
838 2011-01-26 11:31:11 <ArtForz> sorry, we need those for lisp
839 2011-01-26 11:31:23 <molecular> xelister, I also need some ":" and "-"
840 2011-01-26 11:31:38 <molecular> for lisp? that's not nearly enough for even the simplest task ;)
841 2011-01-26 11:32:12 <andrew12^mac> http://xkcd.com/297/
842 2011-01-26 11:32:35 <ArtForz> lol, yeah
843 2011-01-26 11:33:20 <noagendamarket> Im not a coder yet I laughed at that comic
844 2011-01-26 11:33:32 <noagendamarket> I dont know why lol
845 2011-01-26 11:34:04 <ArtForz> old joke, LISP - Lots of Irritating Stupid Parens
846 2011-01-26 11:34:41 <noagendamarket> yea I got that part :)
847 2011-01-26 11:34:53 <ArtForz> so, looks like no go for 4*6990 without having to fuck with risers to get some air space between em
848 2011-01-26 11:34:56 <andrew12^mac> but who was Scheme?
849 2011-01-26 11:36:18 <lfm> andrew12just a brand name dialect
850 2011-01-26 11:36:32 <andrew12^mac> heh
851 2011-01-26 11:36:51 <molecular> can I overclock higher when the chips are cooler? a lot, or just a little?
852 2011-01-26 11:36:52 <lfm> like basic is to gwbasic
853 2011-01-26 11:37:15 <ArtForz> usually yes, and usually 10-20 MHz
854 2011-01-26 11:37:51 <ArtForz> probably quite a bit more if you were close to 90???C before
855 2011-01-26 11:37:57 <andrew12> wut if i put my gpu in the snow? :p
856 2011-01-26 11:38:11 <andrew12> (i dont even have a gpu)
857 2011-01-26 11:38:29 <lfm> youd be checking your product return conditions
858 2011-01-26 11:39:55 <slush1> Diablo-D3: I have just another problem with one miner box. Miner started with all 4 devices, but after some time, one device after one device hangs in miner. Sometimes it run on 4 devices for few hours, but sometimes it hangs few minutes after restart
859 2011-01-26 11:40:25 <ArtForz> keep core temp low and most 5970s can hit >850MHz @ stock V
860 2011-01-26 11:40:25 <slush1> Diablo-D3: Ctrl+C does not work when there are stopped devices
861 2011-01-26 11:40:34 <ArtForz> yeah, they crashed
862 2011-01-26 11:41:03 <slush1> Diablo-D3: so I have to kill process directly. Then I see in top "java <defunct>" which take 100% of CPU...
863 2011-01-26 11:41:05 <sipa> ArtForz: mine locks up easily over 820MHz
864 2011-01-26 11:41:10 <ArtForz> probably overheating or OCed too far
865 2011-01-26 11:41:27 <ArtForz> the cheap sapphire 5970s seem to be bad OCers
866 2011-01-26 11:41:38 <ArtForz> well, not really bad, just a lot more variance
867 2011-01-26 11:41:49 <molecular> arforz: sound nice. I started running into trouble beyond 800Mhz @87-90???C
868 2011-01-26 11:41:57 <ArtForz> errr... yes
869 2011-01-26 11:42:05 <ArtForz> I try to keep GPU temp < 80???C
870 2011-01-26 11:42:07 <lfm> or power supply running outa juice
871 2011-01-26 11:42:08 <molecular> artforz: that's what I have ;(
872 2011-01-26 11:42:29 <ArtForz> most older 5970s usually had no problem getting 830+MHz on both cores
873 2011-01-26 11:42:43 <molecular> I'll have to check cores independently at some point
874 2011-01-26 11:42:53 <molecular> had to get temp under control first, though
875 2011-01-26 11:42:55 <ArtForz> one of my sapphires is maxing at 790/870
876 2011-01-26 11:43:05 <molecular> interesting
877 2011-01-26 11:43:33 <sipa> ArtForz: you mean 790 for one core and 870 for the other?
878 2011-01-26 11:43:36 <ArtForz> yep
879 2011-01-26 11:43:55 <sipa> quite a difference
880 2011-01-26 11:43:56 <ArtForz> with the faster core running ~5???C hotter at same clocks
881 2011-01-26 11:44:05 <Keefe> and i have one at 800/865
882 2011-01-26 11:44:27 <ArtForz> my older cards seem better matched speed and temp-wise
883 2011-01-26 11:44:34 <molecular> aticonfig is reporting different temp than adl_sdk
884 2011-01-26 11:44:40 <Diablo-D3> slush1: thats clearly a driver bug
885 2011-01-26 11:44:42 <Keefe> with the slower one 4 C hotter
886 2011-01-26 11:44:43 <Diablo-D3> slush1: I cant fix that.
887 2011-01-26 11:44:47 <Diablo-D3> it seems to be related to 10.12
888 2011-01-26 11:44:57 <lfm> ever heard of 5870 x2?
889 2011-01-26 11:45:03 <ArtForz> yep
890 2011-01-26 11:45:13 <Diablo-D3> lfm: those are stock full mhz 5970s
891 2011-01-26 11:45:24 <Diablo-D3> as in, the 5870s are at 5870 speeds
892 2011-01-26 11:45:27 <Diablo-D3> instead of underclocked
893 2011-01-26 11:45:30 <Diablo-D3> its a marketing gimmic
894 2011-01-26 11:45:43 <ArtForz> my guess, the sapphire 5970s were made from remaining cypresses after 58xx got phased out
895 2011-01-26 11:46:45 <ArtForz> which makes sense
896 2011-01-26 11:46:52 <slush1> Diablo-D3: what's the preferred driver version?
897 2011-01-26 11:46:56 <ArtForz> 5970 had tiny volume compared to 58xx
898 2011-01-26 11:47:40 <ArtForz> = chips that would still make 5970 spec but ran a bit slow/hot/... went to 5850 or 5830 instead
899 2011-01-26 11:47:51 <slush1> Diablo-D3: It's pretty interesting, that one box with two cards is working stable. This second box is using the same hardware (MB+CPU) and HDD is the exact clone of first box
900 2011-01-26 11:47:54 <Diablo-D3> slush1: well, Im using 10.12, and Ive only seen it once
901 2011-01-26 11:48:01 <slush1> Diablo-D3: So only one difference is in card manufacturer
902 2011-01-26 11:48:55 <ArtForz> obviously if you just use all chips that make 5970 spec, you get a lot more variance
903 2011-01-26 11:50:35 <slush1> Diablo-D3: I'm using also 10.12
904 2011-01-26 11:51:46 <Diablo-D3> try 10.11 and see if it goes away
905 2011-01-26 11:54:56 <slush1> ok
906 2011-01-26 12:19:42 <slush1> One of my core is quite hot on 735MHz - 82.5 oC. Is good idea to remove cooler and add thermal paste?
907 2011-01-26 12:21:00 <slush1> It's not a new card, I suspect that there was water cooling before, so maybe cooler is badly mounted
908 2011-01-26 12:22:47 <gribble> ...
909 2011-01-26 12:22:47 <xelister> ;;calc 165000
910 2011-01-26 12:22:51 <gribble> ...
911 2011-01-26 12:22:51 <xelister> ;; calc 165000
912 2011-01-26 12:24:22 <Diablo-D3> slush1: ADD?
913 2011-01-26 12:24:23 <Diablo-D3> dude no
914 2011-01-26 12:25:34 <Diablo-D3> you use something like arctic silver's two part cleaner
915 2011-01-26 12:27:31 <slush1> Diablo-D3: well, remove old and add new one
916 2011-01-26 12:27:39 <Diablo-D3> yeah, but you have to remove it correctly
917 2011-01-26 12:27:43 <Diablo-D3> and it also depends what it used
918 2011-01-26 12:27:58 <Diablo-D3> the heatsink on my 4850 was a sheet of thermal pad
919 2011-01-26 12:28:12 <Diablo-D3> with thermal goop on the gpu onky
920 2011-01-26 12:28:22 <Diablo-D3> so if I ever plan on remounting that, I'll have to redo all the thermal pad work too
921 2011-01-26 12:29:25 <slush1> hmm
922 2011-01-26 12:29:51 <slush1> Core can go >860 oC, but temperature is insane
923 2011-01-26 12:30:29 <slush1> 860 MHz :_D
924 2011-01-26 12:30:35 <UukGoblin> ;;bc,calc 165000
925 2011-01-26 12:30:37 <gribble> The average time to generate a block at 165000 Khps, given current difficulty of 18437.64439217 , is 5 days, 13 hours, 18 minutes, and 53 seconds
926 2011-01-26 12:30:58 <slush1> not oC. So I don't know if is better to let it be on basic frequency or try to fix cooling (with the risk that it will be even worse)
927 2011-01-26 12:35:09 <Diablo-D3> heh
928 2011-01-26 12:49:26 <tcatm> slush1: are talking about a 5970?
929 2011-01-26 12:50:34 <slush1> yes
930 2011-01-26 12:51:46 <tcatm> Is the air intake blocked in some way?
931 2011-01-26 12:52:34 <tcatm> What does DISPLAY=:0.0 aticonfig --pplib-cmd 'get fanspeed 0' output?
932 2011-01-26 12:53:27 <slush1> tcatm: fan is on 90%, second core is on 72 oC
933 2011-01-26 12:53:42 <slush1> (I set up fan to 90% by self, it's not an auto)
934 2011-01-26 12:53:44 <tcatm> RMA the card
935 2011-01-26 12:55:07 <UukGoblin> huh
936 2011-01-26 12:55:11 <UukGoblin> nice command
937 2011-01-26 12:55:22 <cosurgi> slush1: did you reduce the problem to just a single card?
938 2011-01-26 12:55:28 <UukGoblin> but reports 29% on my card where fan is totally disconnected from the card :-]
939 2011-01-26 12:55:51 <slush1> cosurgi: ?
940 2011-01-26 12:56:14 <slush1> tcatm: RMA is not possible, it's from second hand
941 2011-01-26 12:56:25 <slush1> tcatm: I'll consider this card as 5870 :)
942 2011-01-26 12:56:40 <tcatm> slush1: ebay?
943 2011-01-26 12:56:46 <slush1> tcatm: something similar
944 2011-01-26 12:57:00 <tcatm> maybe it was sold because of that defect
945 2011-01-26 12:57:06 <slush1> probably
946 2011-01-26 12:57:18 <slush1> I wrote email to seller, so I'll see
947 2011-01-26 12:57:38 <slush1> card is working on default clocks, so I don't think he will accept it back
948 2011-01-26 12:58:11 <tcatm> well, re-applying thermal paste *should* fix it but it's a little bit more complicated to do than on a CPU (two cores, bridge, memory and VRM)
949 2011-01-26 12:59:02 <slush1> any nice howto on the web?
950 2011-01-26 12:59:23 <slush1> Cores can go pretty high, so I'll probably try it somedays
951 2011-01-26 12:59:35 <tcatm> no idea. Do you know how to apply thermal paste in very tiny amounts?
952 2011-01-26 13:00:14 <slush1> If it is similar as with cleaning my teeth, then yes
953 2011-01-26 13:06:32 <tcatm> http://www.overclockers.com/applying-thermal-grease/ depending on size of DIE use either 3 or 4
954 2011-01-26 13:08:35 <slush1> nice. On CPUs I always used something like 3
955 2011-01-26 13:09:19 <tcatm> http://images.bit-tech.net/content_images/2010/05/sapphire/hd5970-2b.jpg