1 2010-11-25 00:00:11 <bost_> hi all
  2 2010-11-25 00:00:38 <tcatm> hi
  3 2010-11-25 00:04:22 <ColonelPanic1> oh sweet, nice to see you made some progress jgarzik
  4 2010-11-25 00:05:21 <jgarzik> tcatm: my sha256 algorithm was byteswapping its input in the initial LOAD_OP stuff.  the data was already byteswapping, so that resulted in incorrect hash results.
  5 2010-11-25 00:05:50 <tcatm> ah interesting
  6 2010-11-25 00:06:42 <jgarzik> verified generation on testnet, so it + new getwork... works.
  7 2010-11-25 00:07:10 <tcatm> cool. how big is the network latency for getwork?
  8 2010-11-25 00:08:38 <jgarzik> small, when it's localhost :)
  9 2010-11-25 00:09:19 <jgarzik> I don't stop nonce'ing at 0xffff.  You're looking at a metric ton of 'getwork' requests at that point.  I continue incrementing nonce to 0xffffff.
 10 2010-11-25 00:09:43 <ColonelPanic1> ffff pidgin doesn't do SASL, I need another irc client
 11 2010-11-25 00:09:53 <tcatm> for opencl miners latency is an issue
 12 2010-11-25 00:09:58 <ColonelPanic1> or I need linux on here
 13 2010-11-25 00:16:07 <jgarzik> tcatm: do you mean time between new TX (and thus new block header) in bitcoin client, and time that new data is hashed in the miner?
 14 2010-11-25 00:17:22 <tcatm> no, time between two scanhash() calls including the time it takes to fetch a new workunit
 15 2010-11-25 00:17:51 <nanotube> jgarzik: congrats on getting it working :)
 16 2010-11-25 00:18:38 <jgarzik> tcatm: how does that become an issue?
 17 2010-11-25 00:19:20 <tcatm> during that time the miner is not mining
 18 2010-11-25 00:21:12 <jgarzik> tcatm: decrease the number of workunit fetches with a patch like
 19 2010-11-25 00:21:13 <jgarzik> -        if ((nNonce & 0xffff) == 0)
 20 2010-11-25 00:21:42 <jgarzik> and spend more overall time mining
 21 2010-11-25 00:23:51 <jgarzik> bitcoin really should accepting gzip encoding for JSON, too, IMO
 22 2010-11-25 00:23:56 <jgarzik> accept
 23 2010-11-25 00:24:54 <tcatm> is the address cache full of invalid addrs?
 24 2010-11-25 00:32:08 <MT`AwAy> [10:23:52] <jgarzik> bitcoin really should accepting gzip encoding for JSON, too, IMO <- you can enable layer level compression when using SSL
 25 2010-11-25 00:32:43 <brocktice> Is there some way to make the server listen to clients beside localhost?
 26 2010-11-25 00:32:59 <brocktice> I can do an ssh tunnel but it's not ideal
 27 2010-11-25 00:33:19 <MT`AwAy> brocktice: rpcallowip=* for example
 28 2010-11-25 00:33:28 <brocktice> ah, thanks
 29 2010-11-25 00:33:33 <MT`AwAy> if you do that it'll listen on any instead of 127.0.0.1
 30 2010-11-25 00:34:07 <brocktice> wildcards are ok?
 31 2010-11-25 00:34:12 <brocktice> Like 10.0.0.*
 32 2010-11-25 00:34:18 <MT`AwAy> I guess so
 33 2010-11-25 00:34:28 <brocktice> one way to find out :)
 34 2010-11-25 00:35:33 <brocktice> yep, seems to work
 35 2010-11-25 00:35:35 <brocktice> yay!
 36 2010-11-25 00:35:41 <brocktice> thanks! address?
 37 2010-11-25 00:35:59 <tcatm> mhm how do i work around a "file descriptor out of range"?
 38 2010-11-25 00:36:10 <doublec> brocktice, keep in mind if you're not using SSL that your rpcpassword is being sent in the clear
 39 2010-11-25 00:36:15 <brocktice> tcatm, what's the context?
 40 2010-11-25 00:36:17 <kermit> tcatm: are you using select()
 41 2010-11-25 00:36:19 <brocktice> doublec, It's just on the local network
 42 2010-11-25 00:36:22 <tcatm> kermit: yes
 43 2010-11-25 00:36:30 <doublec> ok, just thought I'd mention it just in case
 44 2010-11-25 00:36:31 <brocktice> doublec, I want to have multiple miners connect to one bitcoind
 45 2010-11-25 00:36:37 <brocktice> doublec, I appreciate it though
 46 2010-11-25 00:36:51 <kermit> tcatm: i think i heard you should use poll() now
 47 2010-11-25 00:37:24 <tcatm> mhm should actually work. i'm using seperate threads anyway
 48 2010-11-25 00:37:30 <kermit> tcatm: i think i heard one reason is exactly that, it can handle more descriptors than select(), but i may be wrong
 49 2010-11-25 00:37:32 <MT`AwAy> brocktice: address of?
 50 2010-11-25 00:37:40 <tcatm> (i'm trying to connect to as many bitcoin nodes as possible)
 51 2010-11-25 00:37:44 <kermit> tcatm: poll() isnt actually polling, its just like select()
 52 2010-11-25 00:38:49 <kermit> tcatm: all i remember for sure is that there's no reason to use select() anymore if you have poll()
 53 2010-11-25 00:39:03 <brocktice> MT`AwAy, a bitcoin address to which I can express my appreciation for your help
 54 2010-11-25 00:39:07 <MT`AwAy> oh :p
 55 2010-11-25 00:39:09 <tylergillies> whats the difference between multiple miners talking to a single daemon and just runnning multiple daemons?
 56 2010-11-25 00:39:24 <brocktice> tylergillies, if they're using multiple daemons they can compete
 57 2010-11-25 00:39:42 <MT`AwAy> brocktice: 1CSKtRCzoa1hJbpZ1TXdDwdVW8PX4hZatV for example :)
 58 2010-11-25 00:39:43 <brocktice> if they're sharing one then getwork will give them different stuff to work on, though I may be misunderstanding
 59 2010-11-25 00:40:08 <tylergillies> brocktice: oh gotcha, makes sense, even if you might be wrong ;)
 60 2010-11-25 00:40:24 <kermit> tcatm: also if you're using more than 1024 descriptors, maybe look into setrlimit
 61 2010-11-25 00:40:37 <brocktice> tylergillies, Also, it's one less bitcoin daemon to manage
 62 2010-11-25 00:44:12 <brocktice> I love it when I get these emails: http://img535.imageshack.us/img535/7010/mineremail.png
 63 2010-11-25 00:45:03 <MT`AwAy> :p
 64 2010-11-25 00:45:18 <tcatm> my spamfilter would trigger :)
 65 2010-11-25 00:47:03 <brocktice> tcatm, I have a special sieve filter that drops that right in my high-priority folder
 66 2010-11-25 00:47:11 <brocktice> my phone also beeps to let me know :)
 67 2010-11-25 00:47:50 <tcatm> i check once a day how many blocks my miners found
 68 2010-11-25 00:48:17 <brocktice> I've had a few breakdowns so I'm neurotic about it
 69 2010-11-25 00:48:31 <brocktice> I also have temperature, load, process, and pump speed monitors that'll email me if they go awry
 70 2010-11-25 00:48:34 <brocktice> brb
 71 2010-11-25 00:54:40 <tcatm> okay, polling is now working. let's see if it runs into other limits
 72 2010-11-25 00:56:54 <tcatm> seems to work. i can trace transactions
 73 2010-11-25 00:57:00 <brocktice> once the hardware is paid off I'll be less neurotic about it
 74 2010-11-25 00:57:31 <tcatm> i've got about 100% profit from mining now
 75 2010-11-25 00:57:50 <brocktice> yeah the price will have to go pretty low after the hardware is paid off for it not to be profitable
 76 2010-11-25 00:57:56 <brocktice> or the difficulty much much higher
 77 2010-11-25 00:58:51 <tcatm> yes. though current price/difficulty is not friendly to new miners
 78 2010-11-25 00:58:59 <brocktice> yep
 79 2010-11-25 00:59:04 <brocktice> I think I got in toward the end of that
 80 2010-11-25 00:59:14 <brocktice> so I hope it stabilizes :)
 81 2010-11-25 00:59:14 <tcatm> i wouldn't invest any further
 82 2010-11-25 00:59:35 <brocktice> tcatm, I got a couple of 5770s for my workstation for eyefinity, realized the first wouldn't work the day it arrived so I ordered a second
 83 2010-11-25 00:59:41 <brocktice> (with 5 ports)
 84 2010-11-25 00:59:54 <brocktice> I figured rather than return the first one I'd use it to mine too
 85 2010-11-25 00:59:57 <tcatm> i hope to rip one hd 5870 from my miners one day to use in my workstation
 86 2010-11-25 01:00:01 <brocktice> but it wasn't an intentional additional investment.
 87 2010-11-25 01:00:26 <brocktice> I had been driving 4 displays from 2 nvidia cards but I couldn't use compiz and I had to use fakeXinerema
 88 2010-11-25 01:00:38 <brocktice> I figured out an eyefinity 5 or 6 card would let me have it all
 89 2010-11-25 01:00:43 <brocktice> plus I could mine with it
 90 2010-11-25 01:00:45 <brocktice> so, bought
 91 2010-11-25 01:01:05 <tcatm> i bought them only for mining
 92 2010-11-25 01:01:16 <brocktice> tcatm, what're you running now?
 93 2010-11-25 01:01:25 <tcatm> 3x 5870 + 2x 5970
 94 2010-11-25 01:01:49 <brocktice> cool, 2 boxes I take it?
 95 2010-11-25 01:02:06 <tcatm> yup
 96 2010-11-25 01:02:15 <brocktice> prices have come down quite a bit in anticipation of the 6990 announcement
 97 2010-11-25 01:02:16 <tcatm> 2x 5970 has 23 days uptime
 98 2010-11-25 01:02:24 <brocktice> you can now get a 5970 for $480 with rebate.
 99 2010-11-25 01:02:34 <brocktice> I paid $550 each the prior week :(
100 2010-11-25 01:02:41 <tcatm> 5870 9 days (hmm i didn't even notice a reboot?!)
101 2010-11-25 01:03:12 <brocktice> so you have the 5970s in one and the 5870s in the other?
102 2010-11-25 01:03:20 <tcatm> yep
103 2010-11-25 01:03:36 <tcatm> 5970 need space for airflow
104 2010-11-25 01:03:47 <brocktice> yeah
105 2010-11-25 01:03:59 <brocktice> or waterflow :)
106 2010-11-25 01:04:17 <tcatm> i was considering watercooling, too
107 2010-11-25 01:04:34 <brocktice> If I upgrade my PSU I can probably fit in 3 with water
108 2010-11-25 01:04:35 <tcatm> then i tested with one 5970 and figured aircooling would work well
109 2010-11-25 01:04:52 <brocktice> my stupid motherboard has awful slot placement so one was always overheating
110 2010-11-25 01:05:03 <MT`AwAy> brocktice: thanks btw!
111 2010-11-25 01:05:08 <brocktice> I should have thought of that before buying it, but there are so many variables it's not surprising I missed it
112 2010-11-25 01:05:16 <brocktice> MT`AwAy, no problem, thanks for the help
113 2010-11-25 01:05:16 <tcatm> i'm around 70..80 ???C depending on weather conditions :)
114 2010-11-25 01:05:37 <brocktice> mine currently range from 45.5 to 52.5
115 2010-11-25 01:05:50 <brocktice> If I turn the fans down to 'normal' noise levels it's 52-58
116 2010-11-25 01:07:40 <ArtForz> anything below the glass point of the underfill should be fine for pretty much forever
117 2010-11-25 01:07:56 <brocktice> ArtForz, yeah I have to measure for those direct connects
118 2010-11-25 01:08:08 <brocktice> If I can pull off three cards in a row I'm sure the rad can handle it
119 2010-11-25 01:08:09 <tcatm> i don't expect them to survive mining for years
120 2010-11-25 01:08:22 <ArtForz> and afair AMD uses 95???C underfill
121 2010-11-25 01:09:29 <brocktice> Art, yours run at, what, around 70?
122 2010-11-25 01:09:53 <ArtForz> depends on air temp, usually somewhere between 75 and 80
123 2010-11-25 01:10:35 <brocktice> and 910 MHz?
124 2010-11-25 01:10:38 <ArtForz> yep
125 2010-11-25 01:13:00 <ArtForz> there seems to be something a bit wrong
126 2010-11-25 01:13:03 <brocktice> There's been sporadic tx spamming right?
127 2010-11-25 01:13:06 <brocktice> ArtForz, with what?
128 2010-11-25 01:13:16 <ArtForz> not sure yet, either my client or the overall network
129 2010-11-25 01:13:21 <ArtForz> getdatas take AGES to return
130 2010-11-25 01:13:50 <tcatm> i just hacked a client that connects to >500 nodes
131 2010-11-25 01:13:58 <tcatm> but that shouldn't cause problems
132 2010-11-25 01:14:11 <ArtForz> nope
133 2010-11-25 01:14:47 <ArtForz> you know, the usual got inv x, asking for x..., getdata x, got tx/block x
134 2010-11-25 01:15:35 <tcatm> is 76a914e3ebb3583e1904a814a07d349a61c245f0c5656a88ac a spam address?
135 2010-11-25 01:16:34 <ArtForz> 11/25/2010 02:15:10 received: inv (37 bytes)
136 2010-11-25 01:16:35 <ArtForz> got inventory: tx 654771262e994a95a84b  new
137 2010-11-25 01:16:44 <ArtForz> 11/25/2010 02:16:17 received: tx (258 bytes)
138 2010-11-25 01:17:12 <ArtForz> 77 seconds from getdata to response ...
139 2010-11-25 01:19:14 <theymos> BBE now has short links for transaction, block, and address pages. Example: https://blockexplorer.com/b/91385 . What do you think about the way I'm styling these links on the pages?
140 2010-11-25 01:19:58 <brocktice> theymos, with the hashes you mean?
141 2010-11-25 01:20:11 <tcatm> theymos: connection timeout :/
142 2010-11-25 01:20:36 <theymos> brocktice: I mean the "Short link: https://blockexplorer.com/b/91385" text.
143 2010-11-25 01:21:18 <brocktice> Why not just make the Block 91385 the link?
144 2010-11-25 01:21:25 <brocktice> Also, why does everything have a question mark after it?
145 2010-11-25 01:21:28 <tcatm> it looks bit lost
146 2010-11-25 01:21:49 <theymos> The question marks produce help text when you hover your mouse cursor over them.
147 2010-11-25 01:22:03 <brocktice> oh
148 2010-11-25 01:22:05 <brocktice> so they do
149 2010-11-25 01:22:05 <theymos> tcatm: Are you still getting that error?
150 2010-11-25 01:22:16 <brocktice> that's not super intuitive
151 2010-11-25 01:22:25 <theymos> I should probably underline them or something.
152 2010-11-25 01:22:46 <tcatm> theymos: sometimes. maybe my connection is a bit overloaded
153 2010-11-25 01:23:17 <tcatm> i'd make the blocknumber the link i.e. <h1>Block <a>1234</a></h1>
154 2010-11-25 01:23:43 <brocktice> I like that
155 2010-11-25 01:24:17 <theymos> Then people would have to 'right click->copy link' to get the short link. Hardly anyone would notice the feature.
156 2010-11-25 01:25:00 <tcatm> you could rewrite the url in the addressbar to be the short link
157 2010-11-25 01:26:35 <tcatm> especially as the url already looks pretty clean. i don't think many people would even look for a shorter link before copy&pasting the url from the addressbar
158 2010-11-25 01:28:08 <theymos> The real links are long to avoid any possible ambiguity, and also to make automatic processing easier.
159 2010-11-25 01:28:52 <tcatm> aren't blocknumbers unique?
160 2010-11-25 01:29:01 <theymos> They can be replaced.
161 2010-11-25 01:29:28 <ArtForz> tcatm: if you still have that well-connected client online, can you monitor inv (= tx/block) spread?
162 2010-11-25 01:30:15 <tcatm> ArtForz: i only monitor transactions, their first appearance and timestamp
163 2010-11-25 01:30:26 <ArtForz> as it looks like txs/blocks are spreading a lot slower than they should
164 2010-11-25 01:30:51 <tcatm> how would i monitor that tx/block spread?
165 2010-11-25 01:31:00 <ArtForz> incoming inv messages
166 2010-11-25 01:32:09 <tcatm> i'll try to hack it to do that (it's based on your half client)
167 2010-11-25 01:32:13 <ArtForz> normally when a tx/block spreads the time between the first peer saying inv(somehash) and the rest of em doing the same is pretty low
168 2010-11-25 01:32:25 <ArtForz> usually a few seconds at most
169 2010-11-25 01:32:50 <ArtForz> currently it looks more like 60-80 seconds
170 2010-11-25 01:32:56 <tcatm> ah so you want to measure the time between first and last broadcast of a tx?
171 2010-11-25 01:33:01 <ArtForz> kinda
172 2010-11-25 01:33:15 <ArtForz> rememebr peers broadcast a inv/txhash) when they get a tx
173 2010-11-25 01:33:27 <ArtForz> so when you look for incoming invs you can time when each peer got that tx
174 2010-11-25 01:33:35 <ArtForz> same for blocks
175 2010-11-25 01:33:42 <tcatm> yes that's what i'm doing to find the origin of TXs
176 2010-11-25 01:33:53 <ArtForz> yep
177 2010-11-25 01:34:13 <tcatm> found 3 IPs where most of them originate
178 2010-11-25 01:34:16 <ArtForz> and normally the time between the first peer getting it and the majority of em getting it is pretty damn small
179 2010-11-25 01:35:07 <ArtForz> and currently that time looks to be in the 1min+ range
180 2010-11-25 01:35:45 <ArtForz> I suspect something in inv or tx cache handling is scaling at O(n**2)
181 2010-11-25 01:36:01 <tcatm> i'll try to get that value
182 2010-11-25 01:38:58 <tcatm> http://pastebin.com/0RteirNq the top IPs might be the spammer or are close to him
183 2010-11-25 01:40:38 <brocktice> ArtForz, are you seeing a hit to your mining rate?
184 2010-11-25 01:41:20 <ArtForz> nope
185 2010-11-25 01:42:48 <xelister> how nice, my PSU doesnt fit my 5970.
186 2010-11-25 01:42:54 <ArtForz> huh?
187 2010-11-25 01:43:13 <brocktice> 'fit'?
188 2010-11-25 01:43:23 <xelister> well 5970 needs a 6 pin AND an 8 pit power, right?
189 2010-11-25 01:43:26 <ArtForz> yep
190 2010-11-25 01:43:50 <xelister> since 5970 is so incredibly long (taking every milimeter of space allowed by AT specs I presume)
191 2010-11-25 01:43:56 <ArtForz> nope
192 2010-11-25 01:43:58 <xelister> and has the power input on the end
193 2010-11-25 01:44:11 <ArtForz> it's actually about 0.5" shorter then what PCIe spec allows
194 2010-11-25 01:44:36 <xelister> and CPU power is on the other side of mainboard, and distance between this blackwhite cable (12 V?) 6 pin and 8 pin are like 8 cm instead of needed 12
195 2010-11-25 01:44:44 <xelister> I can connect either the GPU or the CPU, not both :]
196 2010-11-25 01:45:15 <ArtForz> imo they should've gone full length and added the standard holes for a PCI retention thingie
197 2010-11-25 01:45:22 <brocktice> you need an extension cord
198 2010-11-25 01:45:25 <ArtForz> yep
199 2010-11-25 01:45:37 <xelister> how this is named?
200 2010-11-25 01:45:38 <brocktice> didn't the 5970 come with a 4pin molex->6+8 or something?
201 2010-11-25 01:45:46 <xelister> even wikipedia doesnt have article on PEG connector
202 2010-11-25 01:45:50 <tcatm> average tx spread in seconds = 1754.25988874
203 2010-11-25 01:45:53 <ArtForz> PEG?
204 2010-11-25 01:46:10 <tcatm> not really sure if i calculated that correctly
205 2010-11-25 01:46:14 <xelister> the blackwhite cables from PSU, 6 and 8 pins, they are 12 V, right?
206 2010-11-25 01:46:37 <xelister> what is technical name for this connector type, that 6 and 8 pin power
207 2010-11-25 01:46:52 <ArtForz> theres 2 kinds of 8 pin power
208 2010-11-25 01:46:58 <brocktice> pci-e?
209 2010-11-25 01:47:05 <ArtForz> EPS12V (the extended ATX 12V) and PCIe 8-pin
210 2010-11-25 01:47:15 <xelister> one cable has 6 and 8 and is black-and-white
211 2010-11-25 01:47:32 <xelister> no, wrong
212 2010-11-25 01:47:36 <brocktice> if it has 6 and 8 it's pci-e
213 2010-11-25 01:47:38 <tcatm> 122.794999997 seconds
214 2010-11-25 01:47:41 <ArtForz> yep
215 2010-11-25 01:47:44 <ArtForz> or 6+2
216 2010-11-25 01:47:51 <xelister> one is black and yellow, it powers CPU by 4 pin and has 8 pin also - this one was too short
217 2010-11-25 01:48:02 <xelister> ^--- that one is named..?  ESP12V ?
218 2010-11-25 01:48:10 <ArtForz> probbaly
219 2010-11-25 01:48:42 <ArtForz> if it's on a cable with the ATX12V (4-pin), it's usually a EPS12V
220 2010-11-25 01:49:18 <ArtForz> http://www.playtool.com/pages/psuconnectors/connectors.html
221 2010-11-25 01:49:53 <brocktice> does anyone know offhand the command-line utility in RHEL or CentOS that lets you pick package groups?
222 2010-11-25 01:50:48 <ArtForz> notice which "pins" are rounded on EPS12V and PCIe 8-pin
223 2010-11-25 01:51:34 <brocktice> ah, grouplist/groupinstall
224 2010-11-25 01:52:24 <ArtForz> of course if you have black+yellow gnd/12v it's easy as hell to tell em apart, eps12v has 12v on clip side row, PCIe has GND on clip side
225 2010-11-25 01:52:43 <brocktice> ArtForz, my motherboard had a cap on half the pins on the eps12v connector
226 2010-11-25 01:52:49 <brocktice> but I took it off and plugged in the whole thing
227 2010-11-25 01:52:52 <brocktice> do you know what that's about?
228 2010-11-25 01:52:54 <ArtForz> yeah, so did mine
229 2010-11-25 01:53:05 <ArtForz> I have no clue whats the point of that
230 2010-11-25 01:53:09 <brocktice> nothing at all about it in the manual
231 2010-11-25 01:53:37 <ArtForz> maybe so it doesn't look like a unused receptable when you only have a ATX12V
232 2010-11-25 01:53:38 <brocktice> hm this is odd
233 2010-11-25 01:53:50 <brocktice> I have transactions that have not been confirmed since 30 minutes ago
234 2010-11-25 01:53:54 <brocktice> no block?
235 2010-11-25 01:54:16 <brocktice> hurdd noy
236 2010-11-25 01:54:21 <brocktice> whoa misalignment
237 2010-11-25 01:54:22 <brocktice> guess not
238 2010-11-25 01:55:23 <ArtForz> bbiab, seeing if I can resurrect my dead 850W PSU
239 2010-11-25 01:55:29 <brocktice> good luck!
240 2010-11-25 01:57:13 <tcatm> i wonder why i can't connect to more than ~600 nodes
241 2010-11-25 01:57:26 <kermit> tcatm: did you see setrlimit?
242 2010-11-25 01:57:35 <kermit> tcatm: the default limit of file descriptors is 1024
243 2010-11-25 01:57:37 <tcatm> yes
244 2010-11-25 01:57:50 <tcatm> i would get an error if i opened more than 1024 FDs
245 2010-11-25 01:57:52 <anarchy_> ;;bc,stats
246 2010-11-25 01:57:54 <gribble> Current Blocks: 93681 | Current Difficulty: 6866.89864897 | Next Difficulty At Block: 94752 | Next Difficulty In: 1071 blocks | Next Difficulty In About: 1 week, 0 days, 4 hours, 20 minutes, and 54 seconds | Next Difficulty Estimate: 7111.99690744
247 2010-11-25 01:58:05 <brocktice> up up
248 2010-11-25 01:58:09 <brocktice> just a little this time
249 2010-11-25 01:58:17 <kermit> tcatm: if you're behind NAT, your router might not be able to keep track of that many connections.
250 2010-11-25 01:58:28 <brocktice> oh yeah, check the limit
251 2010-11-25 01:58:35 <brocktice> if you have the option
252 2010-11-25 02:00:05 <brocktice> Merchants should really be all over bitcoins
253 2010-11-25 02:00:07 <tcatm> mhm don't know if i can even check that
254 2010-11-25 02:00:19 <brocktice> Instant, no tx fees (or very small), no chargebacks
255 2010-11-25 02:00:40 <brocktice> tcatm, I know DD-WRT lets you set it
256 2010-11-25 02:01:26 <brocktice> ah finally someone found 93681
257 2010-11-25 02:01:37 <tcatm> it's not running dd-wrt
258 2010-11-25 02:01:57 <brocktice> sadly :)
259 2010-11-25 02:06:41 <ColonelPanic1> 1) upgrade kernel 2) get 1/2 the Mhash/s as before
260 2010-11-25 02:06:43 <tcatm> do nodes broadcast TX multiple times?
261 2010-11-25 02:08:08 <brocktice> ColonelPanic1, which distro/colonel?
262 2010-11-25 02:08:18 <ColonelPanic1> ubuntu 10.10
263 2010-11-25 02:08:37 <brocktice> that's pretty severe
264 2010-11-25 02:09:04 <ColonelPanic1> last night I had 100Mhash/s now I have 30-50Mhash/s and that is the only change I've made
265 2010-11-25 02:09:26 <MT`AwAy> 30~50Mhash/s that's what I do with my cpu
266 2010-11-25 02:09:39 <AAA_awright> Kill the nice level
267 2010-11-25 02:10:19 <AAA_awright> ;;bc,stats
268 2010-11-25 02:10:21 <gribble> Current Blocks: 93682 | Current Difficulty: 6866.89864897 | Next Difficulty At Block: 94752 | Next Difficulty In: 1070 blocks | Next Difficulty In About: 1 week, 0 days, 4 hours, 22 minutes, and 9 seconds | Next Difficulty Estimate: 7104.49020634
269 2010-11-25 02:10:36 <AAA_awright> "1 week, 0 days, 4 hours, 22 minutes, and 9 seconds" That's what we call "comic precision"
270 2010-11-25 02:11:24 <ColonelPanic1> AAA_awright: it's nice was at 0
271 2010-11-25 02:11:38 <brocktice> ColonelPanic1, what hardware?
272 2010-11-25 02:11:40 <AAA_awright> ColonelPanic1: NEGATIVE 20!
273 2010-11-25 02:11:49 <brocktice> yah -20 if you're serious, yo
274 2010-11-25 02:12:05 <AAA_awright> Just don't be suprised when you can't even type in your IRC client
275 2010-11-25 02:12:07 <ColonelPanic1> it doesn't make a bit of difference
276 2010-11-25 02:12:09 <ColonelPanic1> the balls are inert
277 2010-11-25 02:12:16 <ColonelPanic1> irc client is on another computer
278 2010-11-25 02:12:29 <ColonelPanic1> brocktice: radeon 5850 mobile
279 2010-11-25 02:12:41 <brocktice> ColonelPanic1, hm, and you're sure it was just the kernel?
280 2010-11-25 02:12:54 <ColonelPanic1> I didn't change anything else between then and now
281 2010-11-25 02:13:02 <ColonelPanic1> I just did homework on it
282 2010-11-25 02:15:08 <ColonelPanic1> im gonna switch back to the old kernel and see how it does
283 2010-11-25 02:22:19 <ColonelPanic2> MT`AwAy: what miner are you using for you CPU?
284 2010-11-25 02:37:02 <jgarzik> tcatm: yes, nodes broadcast TX's multiple times
285 2010-11-25 02:47:08 <ColonelPanic1> still mines just as slow on the older kernel
286 2010-11-25 02:47:09 <ColonelPanic1> wtf happened
287 2010-11-25 02:56:25 <ColonelPanic1> the fans don't even turn on, now
288 2010-11-25 02:59:23 <Kiba> http://ubitio.us/file/download/19
289 2010-11-25 02:59:29 <Kiba> Download!
290 2010-11-25 02:59:58 <ColonelPanic1> pass
291 2010-11-25 03:02:20 <nanotube> nameless|: while you're at it... update the topic for v .15 ? :)
292 2010-11-25 03:02:46 <ColonelPanic1> I really wish I could figure out why my hashing has been cut in half
293 2010-11-25 03:03:13 <nameless> |I suppose the link needs updated too doesn't it?
294 2010-11-25 03:03:14 <davex__> run on an encrypted partition?
295 2010-11-25 03:03:27 <nanotube> nameless|: i suppose so. :)
296 2010-11-25 03:04:36 <nameless> |nanotube: Just PM me if you need the topic updated. I notice it faster than checking for myself :p
297 2010-11-25 03:05:32 <nanotube> nameless|: ok, sounds good. will do. :) or you could give me op and i can do it myself :P
298 2010-11-25 03:05:58 <nameless> |oh right
299 2010-11-25 03:06:04 <nameless> |freenode doesn't support halfop
300 2010-11-25 03:06:08 <nameless> |as +h
301 2010-11-25 03:06:42 <nanotube> yea guess not. well you don't have to give me op if you don't trust me ;)
302 2010-11-25 03:07:05 <nameless> |I don't know you
303 2010-11-25 03:07:16 <nameless> |So, I don't really trust you :p
304 2010-11-25 03:07:28 <gribble> OTC web of trust rating system: http://trust.bitcoin-otc.com/
305 2010-11-25 03:07:28 <nanotube> mmm... you can check out my trust rating on the wot. see ,,(whatis #bitcoin-otc wot)
306 2010-11-25 03:08:23 <Kiba> nanotube: you get paid in bitcoins for ads?
307 2010-11-25 03:08:50 <nanotube> Kiba: yes
308 2010-11-25 03:11:43 <ColonelPanic1> oh fucking alienware, the "stealth mode" feature was enabled
309 2010-11-25 03:15:13 <nanotube> ColonelPanic1: what is that?
310 2010-11-25 03:18:50 <Kiba> nanotube: check out my art yet?
311 2010-11-25 03:19:32 <nanotube> Kiba: nope, i'm not really into art. ...
312 2010-11-25 03:22:55 <Kiba> :*(
313 2010-11-25 03:23:14 <Kiba> look like, I'll just break even with my investment
314 2010-11-25 03:36:22 <ArtForz> hehehe
315 2010-11-25 03:36:26 <ArtForz> that PSU works again
316 2010-11-25 03:37:20 <ArtForz> havent yet done a full load test with scope traces, but input/output power looks ok
317 2010-11-25 03:38:15 <ArtForz> +12V drops from 12.2 @ idle to 11.9 @ 750W load
318 2010-11-25 03:38:48 <ArtForz> so unless I missed a few blown caps it should be ok
319 2010-11-25 03:40:25 <Kiba> finally, my transaction is getting through...
320 2010-11-25 03:43:33 <Diablo-D3> ArtForz: lolwhat
321 2010-11-25 03:44:04 <ArtForz> rememeber theat 850W PSU where I blew up the 12V rectifiers?
322 2010-11-25 03:44:23 <Kiba> tommorow...
323 2010-11-25 03:44:29 <Kiba> I am going to upload a special art piece
324 2010-11-25 03:44:34 <Diablo-D3> ArtForz: yes, but why fix it?
325 2010-11-25 03:44:42 <Kiba> a protrait of F.A Hayek.
326 2010-11-25 03:44:45 <Diablo-D3> ArtForz: a week of mining for you can fix it
327 2010-11-25 03:44:48 <ArtForz> parts total = $1.80
328 2010-11-25 03:46:00 <ArtForz> not to mention fixing stuff is fun
329 2010-11-25 04:37:57 <Diablo-D3> ArtForz: erm, how fast is a 4870?
330 2010-11-25 04:38:20 <ArtForz> 1.2x 4850
331 2010-11-25 04:38:33 <Diablo-D3> http://www.bitcoin.org/smf/index.php?topic=1721.msg24138#msg24138
332 2010-11-25 04:39:07 <ArtForz> soudns like something is going wrong
333 2010-11-25 04:39:20 <Diablo-D3> I agree.
334 2010-11-25 04:39:32 <ArtForz> theoretical peak is 94.4
335 2010-11-25 04:45:13 <Diablo-D3> so I wonder what broke
336 2010-11-25 05:38:05 <LobsterMan> is there any kind of transaction spam going on at all?
337 2010-11-25 05:38:09 <LobsterMan> my shit is spiking again
338 2010-11-25 05:38:25 <LobsterMan> and hashrate drops
339 2010-11-25 05:39:06 <ByteCoin> Yeah it looks like it's continuing.
340 2010-11-25 05:39:20 <ByteCoin> Don't see why it would affect your hashrate though...
341 2010-11-25 05:39:25 <LobsterMan> me either.....
342 2010-11-25 05:39:35 <LobsterMan> i'm using a client built from r...186?
343 2010-11-25 05:39:37 <jgarzik> LobsterMan: according to http://blockexplorer.com/ there is spam
344 2010-11-25 05:40:20 <jgarzik> ByteCoin: well, if your algorithm interrupts itself upon arrival of a new TX, to obtain more work, that would affect overall hashrate due to increased 'getwork' interruptions
345 2010-11-25 05:40:41 <ByteCoin> Shouldn't really be significant
346 2010-11-25 05:40:54 <LobsterMan> the dips in temperature coincide with a drop in gpu hashrate
347 2010-11-25 05:40:58 <LobsterMan> http://imgur.com/evL48.png
348 2010-11-25 05:41:46 <ByteCoin> Ok. If the getwork interruption takes a significant amount of time and the cards go idle then yes
349 2010-11-25 05:42:15 <ByteCoin> You really only need to update the jobs the card is working on when the new getwork has completed
350 2010-11-25 05:42:33 <ByteCoin> If you find a winning hash based on slightly old work it's still fine
351 2010-11-25 05:42:42 <ByteCoin> You have to check your implementation
352 2010-11-25 05:43:06 <LobsterMan> i'm still just using m0mchil's miner
353 2010-11-25 05:43:07 <jgarzik> correct.  but it is easier for single-threaded code to "while (1) { getwork; work for a while; }"
354 2010-11-25 05:43:18 <Kiba> BMAA members, time to vote...
355 2010-11-25 05:43:20 <Kiba> http://fivegrinder.com/group/Bitcoin%20Mutual%20Aid%20Association/Bounty%20for%20Association%20Site
356 2010-11-25 05:43:25 <jgarzik> I think that's what momchil's miner does; that's what my stupid-simple CPU miner does.
357 2010-11-25 05:43:46 <jgarzik> tcatm was complaining about the work interrupt, idling cards while getwork'ing
358 2010-11-25 05:44:10 <ByteCoin> Ok. Thx for that. Not familiar with implementations....
359 2010-11-25 05:44:54 <ByteCoin> Can someone zip up their blk001.dat file and make it available to me. I have a theory I want to test...
360 2010-11-25 05:44:58 <ByteCoin> please?
361 2010-11-25 05:45:39 <LobsterMan> what's contained in that file?
362 2010-11-25 05:45:47 <ByteCoin> It's the block list
363 2010-11-25 05:45:47 <nanotube> that's the blockchain
364 2010-11-25 05:46:40 <LobsterMan> would sharing my copy be compromising my identity or security or anything?
365 2010-11-25 05:47:10 <ByteCoin> Well perhaps you guys would know the answer to the the basic question. Does everyone have the same blk001.dat?
366 2010-11-25 05:47:29 <ByteCoin> LobsterMan: Not really.
367 2010-11-25 05:47:32 <nanotube> afaik, the only thing that you should keep secure is the wallet.dat, LobsterMan . but best see if jgarzik agrees with that assessment. :)
368 2010-11-25 05:47:52 <nanotube> ByteCoin: besides possibility of network lag in latest block distribution... i think it should be the same.
369 2010-11-25 05:48:00 <ByteCoin> I bet people have different orphan blocks.
370 2010-11-25 05:49:42 <nanotube> are orphan blocks even kept in the .blk file? i'd have thought they get discarded once the branches are clearly orphan
371 2010-11-25 05:49:56 <ByteCoin> Nope
372 2010-11-25 05:50:46 <ByteCoin> There's no provision AFAIK for removing stuff from the blk001.dat file
373 2010-11-25 05:51:35 <LobsterMan> mine's about 50mb
374 2010-11-25 05:51:46 <LobsterMan> 51.6 MB (54,142,241 bytes)
375 2010-11-25 05:54:36 <jgarzik> the more interesting question is whether or not all platforms (windows, linux, 32-bit, 64-bit) have compatible blk0001.dat files?
376 2010-11-25 05:54:44 <LobsterMan> heh
377 2010-11-25 05:55:22 <yebyen> pffffft
378 2010-11-25 05:55:25 <yebyen> what about -testnet?
379 2010-11-25 05:55:33 <yebyen> there are plenty of reasons to want a bare client
380 2010-11-25 05:55:47 <yebyen> i've had 4 nodes generating for days and i haven't got a single solution
381 2010-11-25 05:56:17 <yebyen> your blockchain is invalid :D
382 2010-11-25 05:56:28 <jgarzik> 'rm blk0001.dat' is not difficult, if you want a bare client.  Novice users suffer through pointless slow download upon first usage.
383 2010-11-25 05:56:46 <LobsterMan> it should maybe be available if people want it
384 2010-11-25 05:56:49 <ByteCoin> jgarzik: They appear to be compatible.
385 2010-11-25 05:56:52 <LobsterMan> but not dstributed by default
386 2010-11-25 05:56:56 <jgarzik> it should not take 30 minutes to use shiny new software, in year 2010.
387 2010-11-25 05:57:21 <ByteCoin> jgarzik: Balance sheets for the win!
388 2010-11-25 05:57:40 <jgarzik> ByteCoin: the serialization code for CBlock should be the same on all platforms...
389 2010-11-25 05:57:44 <jgarzik> otherwise network wouldn't work
390 2010-11-25 05:57:59 <ArtForz> blk0001 seems to be identical between 32/64 bit and windows/linux
391 2010-11-25 05:58:14 <ArtForz> blkindex.dat ... not so much
392 2010-11-25 05:58:14 <ByteCoin> Not sure what you're driving at in your recent remark jgarzik....
393 2010-11-25 05:59:10 <ArtForz> we'd need to add code to reinit blkindex from a existing blk0001.dat
394 2010-11-25 05:59:28 <jgarzik> ByteCoin: bitcoin encodes a C++ data structure, CBlock, into a serialized byte stream for either network or disk storage.  it stands to reason that the blk0001.dat would be the same on all platforms for that reason, but I've never personally verified this.
395 2010-11-25 05:59:44 <ArtForz> I did, it is identical
396 2010-11-25 06:00:14 <ByteCoin> As I mentioned jgarzik, it appears to be compatible at least between a 32bit ubuntu install and 32 bit win32
397 2010-11-25 06:00:28 <jgarzik> sounds confirmed, then :)
398 2010-11-25 06:00:34 <ArtForz> I compared 64 bit linux, 32 bit linux and 32 bit windows
399 2010-11-25 06:00:54 <ArtForz> blk0001 is just serialized blocks and perfectly portable
400 2010-11-25 06:01:06 <jgarzik> next question:  given a blk0001.dat and no index, will bitcoin properly rebuild index & verify block chain?
401 2010-11-25 06:01:07 <ArtForz> blkindex is a bdb and ... not so portable
402 2010-11-25 06:01:10 <ArtForz> nope
403 2010-11-25 06:01:16 <ByteCoin> Artforz: Could I have a zipped blk001.dat from your longest running fast miner? I'm studying orphan blocks.
404 2010-11-25 06:01:36 <sneak> hi guys
405 2010-11-25 06:01:57 <sneak> ArtForz: how goes the farm
406 2010-11-25 06:02:16 <ArtForz> it's happily farming away
407 2010-11-25 06:02:30 <sneak> so's my 1x 5970 :)
408 2010-11-25 06:03:33 <ArtForz> ewww
409 2010-11-25 06:03:41 <jgarzik> with the right setting, bdb databases are portable.
410 2010-11-25 06:03:43 <ArtForz> even 7zipped blk0001 is 34MB
411 2010-11-25 06:04:26 <jgarzik> default bdb is non-portable, native endian.  but it does have a fixed endian, portable setting.
412 2010-11-25 06:04:53 <ArtForz> its not a endianess issue, looks more like a 32/64 issue
413 2010-11-25 06:05:05 <ByteCoin> Artforz: Well all i'd need is the orphan blocks... Can you strip them out?
414 2010-11-25 06:05:10 <jgarzik> inherent to bdb?  or bitcoin?
415 2010-11-25 06:05:24 <ArtForz> ByteCoin: should be able to, sec
416 2010-11-25 06:05:30 <ByteCoin> I suppose you could start a new client and download the blockchain then do a diff..
417 2010-11-25 06:05:40 <ArtForz> probably to the bdb settings bitcoin uses
418 2010-11-25 06:05:57 <ArtForz> I already have python code to parse the block chain and find the main chain
419 2010-11-25 06:06:32 <ArtForz> just need to add code so it writes any block not in the main chain to a new file in the order they appear in blk0001
420 2010-11-25 06:06:33 <ByteCoin> Artforz: Be carefull as I have instances where the branches are a couple of nodes long
421 2010-11-25 06:06:44 <ArtForz> so?
422 2010-11-25 06:06:46 <ByteCoin> The naive method is not reliable
423 2010-11-25 06:07:01 <ArtForz> I do the same sum-of-difficulty calc as mainline does
424 2010-11-25 06:07:36 <ByteCoin> Naive method being orphan blocks=blocks whose hashes don'e appear in the list of previous hashes
425 2010-11-25 06:07:43 <ArtForz> huh?
426 2010-11-25 06:08:01 <ByteCoin> Nevermind...
427 2010-11-25 06:08:44 <ByteCoin> I was just saying that if the orphan blocks never had any children then you could find them on that basis
428 2010-11-25 06:09:08 <ByteCoin> It's not always true but it is the vast majority of the time
429 2010-11-25 06:09:22 <ArtForz> no
430 2010-11-25 06:09:30 <ArtForz> orphan blocks = blocks not in main chain
431 2010-11-25 06:09:57 <ArtForz> main chain = chain from genesis to block with highest cumulative difficulty
432 2010-11-25 06:18:17 <Diablo-D3> ArtForz: so, I wonder why that guy is having issues
433 2010-11-25 06:18:32 <Diablo-D3> ArtForz: whats the maximum global item size on 4xxx? 2^32?
434 2010-11-25 06:19:40 <ArtForz> less
435 2010-11-25 06:20:13 <Diablo-D3> rutrow
436 2010-11-25 06:20:15 <Diablo-D3> how less?
437 2010-11-25 06:20:30 <ArtForz> isnt there a opencl get_info something for that?
438 2010-11-25 06:20:41 <Diablo-D3> yes but afiak it just returns 2^32
439 2010-11-25 06:22:19 <ArtForz> http://www.sendspace.com/file/z3rcmh
440 2010-11-25 06:23:06 <Diablo-D3> jesus christ my miner rapes X
441 2010-11-25 06:23:22 <ByteCoin> You are a star ArtForz
442 2010-11-25 06:23:36 <Diablo-D3> ArtForz: my hack has a side effect
443 2010-11-25 06:23:43 <Diablo-D3> ArtForz: I now need to double the attempted fps
444 2010-11-25 06:24:01 <ArtForz> Diablo-D3: more workitems and less work per item = more X rapeage
445 2010-11-25 06:24:15 <Diablo-D3> yup
446 2010-11-25 06:24:23 <ArtForz> thats the main reason why I'm looping over nonces inside the kernel
447 2010-11-25 06:24:26 <Diablo-D3> but its now Im queuing two kernels in a row
448 2010-11-25 06:24:38 <Diablo-D3> so if I need -f 120, I now need -f 240
449 2010-11-25 06:24:44 <Diablo-D3> and its probably closer to -f 360 >_>
450 2010-11-25 06:24:56 <Diablo-D3> ArtForz: also, I can't seem to get good ... per... fo..
451 2010-11-25 06:24:58 <Diablo-D3> oh hey
452 2010-11-25 06:25:08 <Diablo-D3> I never tested looping with the split kernel
453 2010-11-25 06:25:56 <Diablo-D3> I mean
454 2010-11-25 06:26:03 <Diablo-D3> I'm already getting 99.1% efficiency
455 2010-11-25 06:26:13 <ArtForz> which I still kinda doubt
456 2010-11-25 06:26:32 <Diablo-D3> so do I
457 2010-11-25 06:26:36 <Diablo-D3> but its _there_
458 2010-11-25 06:27:05 <OneFixt> is it possible to check temperature without having a monitor connected?
459 2010-11-25 06:27:19 <ArtForz> errr... yes?
460 2010-11-25 06:27:28 <ByteCoin> OneFixt: Stick your finger on it...
461 2010-11-25 06:28:18 <ArtForz> set DISPLAY and XAUTHORITY, aticonfig --adapter=all --odgt
462 2010-11-25 06:28:19 <Diablo-D3> OneFixt: aticonfig, duh
463 2010-11-25 06:28:35 <Diablo-D3> speaking of which
464 2010-11-25 06:28:37 <Diablo-D3> Adapter 0 - ATI Radeon HD 4800 Series
465 2010-11-25 06:28:38 <Diablo-D3> ffff
466 2010-11-25 06:28:49 <Diablo-D3> Im going to cook my gpu at this rate
467 2010-11-25 06:28:54 <OneFixt> aticonfig needs X
468 2010-11-25 06:29:04 <ArtForz> errr, so?
469 2010-11-25 06:29:07 <ArtForz> opencl needs X
470 2010-11-25 06:29:12 <Diablo-D3> OneFixt: this is why you export DISPLAY=:0 first
471 2010-11-25 06:29:19 <OneFixt> sorry let me rephrase = can i do it through ssh?
472 2010-11-25 06:29:25 <Diablo-D3> OneFixt: it tells all further X using commands to use your local display server
473 2010-11-25 06:29:30 <Diablo-D3> instead of, say, ssh passthrough
474 2010-11-25 06:29:32 <ArtForz> <artforz> set DISPLAY and XAUTHORITY, aticonfig --adapter=all --odgt
475 2010-11-25 06:29:52 <OneFixt> hmm, let me try that
476 2010-11-25 06:30:04 <ArtForz> = EXACTLY THE SAME THING you do to run the miner via ssh
477 2010-11-25 06:30:06 <Diablo-D3> people sometimes >_>
478 2010-11-25 06:30:36 <Diablo-D3> ArtForz: btw I dont think I have a problem
479 2010-11-25 06:30:53 <Diablo-D3> ArtForz: max global work size on split mode is only 4194304
480 2010-11-25 06:31:35 <ArtForz> that shouldnt be a problem
481 2010-11-25 06:32:01 <Diablo-D3> yeah it shouldnt
482 2010-11-25 06:32:04 <Diablo-D3> so wtf is going on
483 2010-11-25 06:32:22 <Diablo-D3> hrm, I wonder if the max should be that minus 1
484 2010-11-25 06:32:39 <ArtForz> you ALWAYS want global work size a multiple of local work size
485 2010-11-25 06:32:48 <Diablo-D3> it is
486 2010-11-25 06:32:53 <Diablo-D3> unless
487 2010-11-25 06:32:54 <Diablo-D3> rutrow
488 2010-11-25 06:33:10 <Diablo-D3> wait no
489 2010-11-25 06:34:22 <OneFixt> ArtForz: Diablo-D3: it works fine when i have a monitor connected
490 2010-11-25 06:34:28 <OneFixt> but when i don't, it doesn't show the temps
491 2010-11-25 06:34:36 <ArtForz> ahhh... yeah
492 2010-11-25 06:34:41 <Diablo-D3> durf what
493 2010-11-25 06:34:45 <ArtForz> start X with no monitor connected
494 2010-11-25 06:35:03 <kupo> I'm many versions behind now
495 2010-11-25 06:35:04 <Diablo-D3> ArtForz: nooe, no bug
496 2010-11-25 06:35:04 <OneFixt> i think i did that.. but let me restart it
497 2010-11-25 06:35:08 <ArtForz> driver goes haywire if you disconnect the monitor while X is running
498 2010-11-25 06:35:11 <OneFixt> ah
499 2010-11-25 06:35:13 <OneFixt> hm
500 2010-11-25 06:35:16 <Diablo-D3> ArtForz: I only + or - workSizes.
501 2010-11-25 06:35:33 <ArtForz> Diablo: hurrm, weird
502 2010-11-25 06:35:51 <OneFixt> gonna restart the machine and try all over w/out monitor
503 2010-11-25 06:36:01 <ArtForz> do you use events to serialize the kernels?
504 2010-11-25 06:36:40 <Diablo-D3> ArtForz: oh and 4194304 is / 256 anyhow
505 2010-11-25 06:36:49 <ArtForz> yep
506 2010-11-25 06:36:55 <ArtForz> it's a power of 2
507 2010-11-25 06:36:56 <Diablo-D3> I dont use events
508 2010-11-25 06:37:06 <ArtForz> so how do you serialize your 2 kernels?
509 2010-11-25 06:37:08 <OneFixt> ArtForz: Success!! thank you =)
510 2010-11-25 06:37:19 <Diablo-D3> ArtForz: Im not using a queue that can execute out of order
511 2010-11-25 06:37:35 <ArtForz> I dont think opencl allows that
512 2010-11-25 06:37:41 <Diablo-D3> erm, thats the default mode
513 2010-11-25 06:37:45 <ArtForz> no
514 2010-11-25 06:37:53 <Diablo-D3> you have to specifically ask for a queue that does OoOE
515 2010-11-25 06:37:58 <ArtForz> I mean just queueing 2 kernels where #2 depends on #1 results
516 2010-11-25 06:38:14 <Diablo-D3> by default, it always runs them in that order
517 2010-11-25 06:38:15 <Diablo-D3> ArtForz: and btw
518 2010-11-25 06:38:19 <Diablo-D3> even if it DID run them out of order
519 2010-11-25 06:38:23 <Diablo-D3> it still has to execute them
520 2010-11-25 06:38:26 <ArtForz> I think even with OOE disabled that violates opencl specs
521 2010-11-25 06:38:29 <Diablo-D3> you just get useless results
522 2010-11-25 06:38:33 <ArtForz> yep
523 2010-11-25 06:38:39 <Diablo-D3> so thats not the source of the bug
524 2010-11-25 06:39:18 <Diablo-D3> ArtForz: read the specification btw
525 2010-11-25 06:39:29 <Diablo-D3> the default queue style is always execute in order
526 2010-11-25 06:39:38 <ArtForz> yes
527 2010-11-25 06:40:01 <Diablo-D3> so if I queue kernel A, and then kernel B, it will always be A and then B
528 2010-11-25 06:40:11 <Diablo-D3> never B then A, never A and B concurrently
529 2010-11-25 06:40:14 <ArtForz> and afair it's still illegal to queue a kernel that depends on a prev kernels results without either using that ones completion event as a wait for #2 or expliciltyl waiting berfore submitting #2
530 2010-11-25 06:40:49 <Diablo-D3> thats not what the spec says.
531 2010-11-25 06:41:27 <ArtForz> weird
532 2010-11-25 06:41:28 <Diablo-D3> read section 5.8 in the 1.0 spec
533 2010-11-25 06:42:23 <Diablo-D3> If the CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE property of a command-queue is
534 2010-11-25 06:42:24 <Diablo-D3> to kernel B then kernel B will see the correct data in memory objects produced by execution of
535 2010-11-25 06:42:26 <Diablo-D3> kernel A.
536 2010-11-25 06:42:33 <Diablo-D3> also, fuck you evince
537 2010-11-25 06:42:39 <Diablo-D3> shitty paste quality is shitty
538 2010-11-25 06:42:50 <ArtForz> guess it's ok then
539 2010-11-25 06:42:58 <Diablo-D3> ArtForz: now, what ISNT guaranteed is differnet queues
540 2010-11-25 06:42:59 <ArtForz> I thought I read the exact opposite.. weird
541 2010-11-25 06:43:12 <ArtForz> hrrrm.. that might've been it
542 2010-11-25 06:43:12 <Diablo-D3> it can do whatever it wants across multiple queues
543 2010-11-25 06:43:24 <ArtForz> so techincally it should work
544 2010-11-25 06:43:24 <Diablo-D3> thats why I took events out
545 2010-11-25 06:43:35 <Diablo-D3> I was just chaining sub-queues internally
546 2010-11-25 06:43:47 <Diablo-D3> when I went to thread happy shit, I just make CL do it
547 2010-11-25 06:48:11 <jgarzik> is there an ISA newer than ATI r600?
548 2010-11-25 06:48:19 <ArtForz> yep
549 2010-11-25 06:48:19 <jgarzik> http://developer.amd.com/gpu_assets/r600isa.pdf
550 2010-11-25 06:48:23 <ArtForz> r700
551 2010-11-25 06:48:30 <ArtForz> aka 5xxx series
552 2010-11-25 06:48:47 <ArtForz> the docs are strewn pretty much all over AMDs site
553 2010-11-25 06:48:55 <jgarzik> ah.  http://developer.amd.com/gpu_assets/R700-family_instruction_set_architecture.pdf
554 2010-11-25 06:49:50 <ArtForz> sec... yep, that's it
555 2010-11-25 06:50:41 <ArtForz> read 4.7.3 through 4.7.9
556 2010-11-25 06:50:46 <ArtForz> read it a few more times
557 2010-11-25 06:51:27 <Diablo-D3> whats in those sections? =P
558 2010-11-25 06:51:34 <ArtForz> realize why the ocl compiler is having such a hard time making good use of all VLIW units
559 2010-11-25 06:52:16 <ArtForz> GPR read/write, constant read, ... restrictions for V and T units
560 2010-11-25 06:53:33 <ArtForz> my guess is with VLIW4 we should see well >99% ALU usage
561 2010-11-25 06:53:52 <ArtForz> except for that small niggling fact that AMD crippled 32 bit adds
562 2010-11-25 06:54:06 <Diablo-D3> who needs the adds anyhow >_>
563 2010-11-25 06:54:16 <Diablo-D3> ArtForz: actually that might not even matter
564 2010-11-25 06:54:19 <jgarzik> the word I get is "harder for a compiler than ia64"
565 2010-11-25 06:54:20 <ArtForz> = in the end VLIW4 might end up slower per-ALU than VLIW5
566 2010-11-25 06:54:37 <ArtForz> jgarzik: yep
567 2010-11-25 06:54:48 <Diablo-D3> ArtForz: we have more pipes though
568 2010-11-25 06:55:11 <ArtForz> compared to r600/r700 the IA64 arch looks positively easy to optimize for
569 2010-11-25 06:56:44 <ArtForz> well, sah256 kinda has quite a few adds
570 2010-11-25 06:59:58 <Diablo-D3> http://www.khronos.org/registry/cl/sdk/1.0/docs/man/xhtml/clEnqueueNDRangeKernel.html
571 2010-11-25 07:00:10 <Diablo-D3> ArtForz: it seems for global work size the maximum is 32 bit.
572 2010-11-25 07:00:15 <Diablo-D3> there isnt any useful way to query it
573 2010-11-25 07:02:57 <ArtForz> btw, the add hit is quite a bit
574 2010-11-25 07:03:24 <ArtForz> 5xxx theoretical min is 4148 ops per bchash
575 2010-11-25 07:03:45 <ArtForz> 69xx is 5290
576 2010-11-25 07:04:05 <Diablo-D3> I assume 68xx is the same?
577 2010-11-25 07:04:16 <ArtForz> yep, 68xx = 5xxx
578 2010-11-25 07:05:01 <ArtForz> 68xx has 57xx style shaders
579 2010-11-25 07:06:13 <ArtForz> so at the same num_alus * clock 69xx should end up 27% slower than 58xx
580 2010-11-25 07:06:34 <ArtForz> 27.5% actually
581 2010-11-25 07:06:50 <Diablo-D3> ArtForz: what about 4xxx?
582 2010-11-25 07:07:02 <ArtForz> 6356 ops/clock
583 2010-11-25 07:07:07 <Diablo-D3> blergh
584 2010-11-25 07:07:14 <ArtForz> no bitalign = fail
585 2010-11-25 07:07:36 <Diablo-D3> I may buy a 68xx after all >_>
586 2010-11-25 07:07:51 <ArtForz> so let's see
587 2010-11-25 07:08:06 <ArtForz> 6990 is rumored to be 3840 ALUs @ 775MHz
588 2010-11-25 07:08:31 <ArtForz> = 2976Gops/s
589 2010-11-25 07:08:46 <Diablo-D3> what about a 6970?
590 2010-11-25 07:08:51 <Diablo-D3> Im not sure if I really want a dual gpu card
591 2010-11-25 07:09:01 <ArtForz> giving ... 562.5 Mh/s theoretical peak
592 2010-11-25 07:09:09 <Diablo-D3> thats...
593 2010-11-25 07:09:11 <Diablo-D3> not very high at all
594 2010-11-25 07:09:27 <ArtForz> yep
595 2010-11-25 07:09:41 <ArtForz> 5970 is 559.3 theoretical peak ...
596 2010-11-25 07:09:59 <Diablo-D3> those 5970 are looking mighty nice
597 2010-11-25 07:10:22 <ArtForz> 6970 is supposed to be 1920 @ 900MHz
598 2010-11-25 07:10:40 <ArtForz> ... 326.6Mh/s
599 2010-11-25 07:11:10 <Diablo-D3> and what about the 5870 and the 6870?
600 2010-11-25 07:11:12 <ArtForz> 5870 is 327.8 ...
601 2010-11-25 07:11:29 <ArtForz> 6870 is 243.0
602 2010-11-25 07:11:37 <Diablo-D3> huh.
603 2010-11-25 07:11:43 <ArtForz> ... yeah
604 2010-11-25 07:12:42 <ArtForz> 5850 is 251.7
605 2010-11-25 07:13:02 <Diablo-D3> lowest price not shit 5870 == $280
606 2010-11-25 07:13:38 <ArtForz> so 6870 slightly slower than 5850, 6970 about equal to 5870, 6990 about equal to 5970
607 2010-11-25 07:13:47 <Diablo-D3> lowest price not shit 6870 == $250
608 2010-11-25 07:14:04 <Diablo-D3> wait no
609 2010-11-25 07:14:10 <Diablo-D3> $270
610 2010-11-25 07:14:11 <ArtForz> 6870 is a bit slower than a 5850
611 2010-11-25 07:14:27 <Diablo-D3> argh newegg
612 2010-11-25 07:14:37 <Diablo-D3> lowest price not shit 6870 == $270
613 2010-11-25 07:14:42 <ArtForz> 58 *50*, not 70
614 2010-11-25 07:14:57 <Diablo-D3> lowest price not shit 5970 == $470
615 2010-11-25 07:15:10 <ArtForz> 6870 also happens to have exactly the same TDP as 5850
616 2010-11-25 07:15:32 <ArtForz> 5970s are just fucking awesome
617 2010-11-25 07:15:52 <ArtForz> especially considering the can be OCed to 5870 clocks without much of a problem
618 2010-11-25 07:16:15 <Diablo-D3> 5870 == 1.17 mhash per dollar
619 2010-11-25 07:16:26 <ArtForz> and then a 5970 is ~655.7 Mh/s peak
620 2010-11-25 07:16:53 <ArtForz> at 910 core it breaks 700Mh peak
621 2010-11-25 07:17:06 <Diablo-D3> 6870 == 0.9 mhash per dollar
622 2010-11-25 07:17:14 <ArtForz> yep
623 2010-11-25 07:17:28 <ArtForz> so far 6xxx looks like total *yawn*
624 2010-11-25 07:17:41 <Diablo-D3> 5970 == 1.39 mhash per dollar
625 2010-11-25 07:18:05 <ArtForz> OC that 5970 and it's cheaper per Mh than a 5770 ...
626 2010-11-25 07:18:18 <Diablo-D3> did you say what a 6850 does?
627 2010-11-25 07:18:32 <ArtForz> not very much
628 2010-11-25 07:18:41 <Diablo-D3> they dont cost very much
629 2010-11-25 07:18:47 <ArtForz> 10% faster than a 5770
630 2010-11-25 07:18:58 <Diablo-D3> so thats... what?
631 2010-11-25 07:19:09 <ArtForz> 179.3 Mh/s
632 2010-11-25 07:19:34 <ArtForz> 5770 is 163.9
633 2010-11-25 07:19:37 <Diablo-D3> 6850 = 0.94 mhash/dollar at 190
634 2010-11-25 07:19:57 <Diablo-D3> what about a 5850?
635 2010-11-25 07:20:33 <ArtForz> ArtForz> 5850 is 251.7
636 2010-11-25 07:20:56 <Diablo-D3> 5850 == 1.40 mhash/dollar at $180
637 2010-11-25 07:21:55 <Diablo-D3> woah
638 2010-11-25 07:22:05 <Diablo-D3> 5770 == 1.64 mhash/dollar at $100
639 2010-11-25 07:22:14 <ArtForz> factor in the rest of the system and 5970 wins hands down
640 2010-11-25 07:22:25 <Diablo-D3> ArtForz: well except
641 2010-11-25 07:22:29 <Diablo-D3> Im buying one card
642 2010-11-25 07:22:31 <Diablo-D3> not 24
643 2010-11-25 07:22:35 <ArtForz> so?
644 2010-11-25 07:22:37 <Diablo-D3> or whatever it is you're up to
645 2010-11-25 07:22:49 <Diablo-D3> so did I forget a card worth mentioning?
646 2010-11-25 07:22:57 <ArtForz> not really
647 2010-11-25 07:22:59 <ArtForz> still at 12, damn things are on backorder everywhere
648 2010-11-25 07:23:19 <Diablo-D3> 5850, 5970, and 5770 seem to be the best cards here
649 2010-11-25 07:23:25 <ArtForz> yep
650 2010-11-25 07:23:28 <ArtForz> cheapest 5970 I could get right now is ~$700 after tax
651 2010-11-25 07:23:59 <Kiba> I frequent the forum far too much
652 2010-11-25 07:24:15 <ArtForz> the cheap ones are ~$550 after tax here, and they're freaking near impossible to get
653 2010-11-25 07:24:24 <Diablo-D3> so hrm
654 2010-11-25 07:25:06 <Diablo-D3> I did 6870, I did 6850, I did 5870, I did 5850, I did 5770
655 2010-11-25 07:25:32 <Diablo-D3> is there a 6770 yet?
656 2010-11-25 07:25:55 <Diablo-D3> no
657 2010-11-25 07:26:03 <ArtForz> nope ,and there probly wont be one for > 6 months
658 2010-11-25 07:26:35 <ArtForz> 67xx is supposed to be first GPU on 28nm, Q2-Q3 '11
659 2010-11-25 07:26:50 <Diablo-D3> my 4850 is 1000 gflops, 110 watts
660 2010-11-25 07:27:39 <Diablo-D3> a 5850 is 2088 gflops, and 151 watts
661 2010-11-25 07:27:44 <ArtForz> 4850 = 78.7 Mh/s peak
662 2010-11-25 07:27:57 <Diablo-D3> ArtForz: Im just considering here my future upgrade
663 2010-11-25 07:28:01 <Diablo-D3> I want it to be a good fit as replacement
664 2010-11-25 07:28:10 <ArtForz> just saying, 4xxx needs a lot more ops to do one hash
665 2010-11-25 07:28:26 <jgarzik> http://cgi.ebay.com/VisionTek-ATI-Radeon-HD-5970-2GB-Video-Card-BRAND-NEW-/170569700026?pt=PCC_Video_TV_Cards&hash=item27b6bf12ba
666 2010-11-25 07:28:28 <jgarzik> $400
667 2010-11-25 07:28:37 <Diablo-D3> jgarzik: visiontek sucks
668 2010-11-25 07:28:43 <ArtForz> 5770 is EXACTLY the same Gops/s as a 4890
669 2010-11-25 07:28:56 <Diablo-D3> 5770 is 1360 gflops at 108 watts
670 2010-11-25 07:29:00 <Diablo-D3> so its not really worth grabbing a 5770
671 2010-11-25 07:29:19 <Diablo-D3> [03:22:05] <Diablo-D3> 5770 == 1.64 mhash/dollar at $100
672 2010-11-25 07:29:22 <ArtForz> welll
673 2010-11-25 07:29:23 <Diablo-D3> [03:20:56] <Diablo-D3> 5850 == 1.40 mhash/dollar at $180
674 2010-11-25 07:29:28 <Diablo-D3> [03:17:41] <Diablo-D3> 5970 == 1.39 mhash per dollar
675 2010-11-25 07:29:34 <Diablo-D3> [03:16:15] <Diablo-D3> 5870 == 1.17 mhash per dollar
676 2010-11-25 07:29:46 <Diablo-D3> so hrm
677 2010-11-25 07:29:57 <ArtForz> a 5770 is still over 2x as fast for bchash than a 4850 ...
678 2010-11-25 07:30:09 <Diablo-D3> yeah but its the same 3d performance
679 2010-11-25 07:30:12 <Diablo-D3> so its not really worth upgrading to
680 2010-11-25 07:30:27 <ArtForz> yep
681 2010-11-25 07:30:49 <Diablo-D3> so 5850 seems to be the bottom contender
682 2010-11-25 07:30:55 <ArtForz> well, if you go for 3D perf 68xx makes sense
683 2010-11-25 07:31:13 <Diablo-D3> 5870 is 2720/188
684 2010-11-25 07:31:17 <ArtForz> a 6870 is about == 5870 for 3d
685 2010-11-25 07:31:18 <Diablo-D3> that doesnt seem too horriffic
686 2010-11-25 07:31:28 <Diablo-D3> yeah, but its lower in bitcoin performance
687 2010-11-25 07:32:07 <ArtForz> hurrm, why would someone sell a NEW 5970 for $400 ?
688 2010-11-25 07:33:23 <Diablo-D3> I have $116.19 in my btc fund
689 2010-11-25 07:33:35 <ArtForz> oh, and OEM cards come with NO accessories
690 2010-11-25 07:33:45 <Diablo-D3> yeah I dont really need the fucking accessories
691 2010-11-25 07:33:54 <Diablo-D3> I wish I could buy OEM cards through newegg
692 2010-11-25 07:34:33 <OneFixt> ;;bc,stats
693 2010-11-25 07:34:36 <gribble> Current Blocks: 93709 | Current Difficulty: 6866.89864897 | Next Difficulty At Block: 94752 | Next Difficulty In: 1043 blocks | Next Difficulty In About: 1 week, 0 days, 1 hour, and 8 seconds | Next Difficulty Estimate: 7063.19106176
694 2010-11-25 07:34:46 <Diablo-D3> ArtForz: so hrm
695 2010-11-25 07:35:04 <Diablo-D3> I think it may be a toss up of either 5850 or 5870
696 2010-11-25 07:35:14 <ArtForz> dual 5770s :P
697 2010-11-25 07:35:19 <Diablo-D3> single slot, fool
698 2010-11-25 07:35:53 <Diablo-D3> 5870 gives me 2720, 188 watts, thats like almost 3x performance for 2x watts
699 2010-11-25 07:35:59 <Diablo-D3> and probably not really add to the heat output any
700 2010-11-25 07:36:19 <ArtForz> 5870 is 188W TDP, 4850 is 110W
701 2010-11-25 07:36:27 <Diablo-D3> I said almost 2x
702 2010-11-25 07:36:34 <ArtForz> same power bracket would be a 5770
703 2010-11-25 07:36:39 <ArtForz> at 108W
704 2010-11-25 07:36:58 <Diablo-D3> so hrm
705 2010-11-25 07:37:07 <Diablo-D3> but they're fucking $250
706 2010-11-25 07:39:39 <Diablo-D3> and $5850s are $180
707 2010-11-25 07:39:41 <Diablo-D3> meh
708 2010-11-25 07:40:15 <ArtForz> GPU upgrades cost money, news at 11
709 2010-11-25 07:41:23 <Diablo-D3> well, GPU production for me is basically dead
710 2010-11-25 07:42:18 <ArtForz> huh?
711 2010-11-25 07:42:36 <Kiba> seem like Diablo-D3 have a lot of problem with bitcoin mining
712 2010-11-25 07:47:41 <Diablo-D3> ArtForz: not enough mhash
713 2010-11-25 07:52:46 <ArtForz> needs more cowb... mhash
714 2010-11-25 08:02:38 <Diablo-D3> lol
715 2010-11-25 08:07:12 <tylergillies> ive heard i should backup wallet after every transaction. is there a point to that? or is it just after every new keygen?
716 2010-11-25 08:07:28 <ArtForz> thanks to keypool it's now every 100 tx
717 2010-11-25 08:07:39 <ArtForz> or whatever you set as keypool size
718 2010-11-25 08:07:39 <tylergillies> keypool?
719 2010-11-25 08:07:43 <ArtForz> yep
720 2010-11-25 08:07:53 <tylergillies> where are docs for that?
721 2010-11-25 08:08:03 <ArtForz> somewhere on the forum
722 2010-11-25 08:08:15 <ArtForz> = it pre-generates "random" keys and stores em in wallet
723 2010-11-25 08:08:27 <ArtForz> and you can set how many it keeps around
724 2010-11-25 08:08:50 <tylergillies> oh cool
725 2010-11-25 08:08:56 <tylergillies> so default value is 100?
726 2010-11-25 08:09:02 <ArtForz> yep
727 2010-11-25 08:09:05 <tylergillies> thnx
728 2010-11-25 08:12:27 <Diablo-D3> Im sad
729 2010-11-25 08:12:36 <Diablo-D3> I havent produced any blocks yet
730 2010-11-25 08:14:25 <ArtForz> wow, that sucks
731 2010-11-25 09:17:41 <Diablo-D3> ArtForz: well
732 2010-11-25 09:17:46 <Diablo-D3> at least I know my miner works
733 2010-11-25 09:17:51 <Diablo-D3> it just doesnt work fast enough
734 2010-11-25 09:17:55 <joe_1> y did u abandon me last night
735 2010-11-25 09:17:59 <ArtForz> ?
736 2010-11-25 09:18:23 <Diablo-D3> joe_1: your mom is hotter
737 2010-11-25 09:18:43 <joe_1> fu
738 2010-11-25 09:18:45 <ArtForz> oh, right, I think I still have him on ignore
739 2010-11-25 09:18:50 <xelister> ArtForz: hi
740 2010-11-25 09:18:53 <Diablo-D3> so
741 2010-11-25 09:18:54 <Diablo-D3> meh
742 2010-11-25 09:19:00 <Diablo-D3> ArtForz: I dunno
743 2010-11-25 09:19:03 <Diablo-D3> ArtForz: I think I wasted my time
744 2010-11-25 09:19:09 <Diablo-D3> I mean, sure, I now have the best miner ever
745 2010-11-25 09:19:10 <ArtForz> xelister: hmm?
746 2010-11-25 09:19:14 <Diablo-D3> but
747 2010-11-25 09:19:16 <Diablo-D3> whats the point
748 2010-11-25 09:19:19 <ArtForz> on a completel??y hopeless arch :P
749 2010-11-25 09:19:26 <Diablo-D3> no
750 2010-11-25 09:19:32 <Diablo-D3> it was written for all archs
751 2010-11-25 09:19:44 <ArtForz> 4xxx is like half the hash/W compared to 5xxx
752 2010-11-25 09:19:48 <Diablo-D3> I suspect the -s hack will work wonders on geforces too
753 2010-11-25 09:19:50 <ArtForz> and you're optimizing for 4xxx ...
754 2010-11-25 09:19:52 <Diablo-D3> but I dont know anyone that has one
755 2010-11-25 09:19:53 <Diablo-D3> ArtForz: no Im not
756 2010-11-25 09:20:03 <Diablo-D3> what optimizations besides the split was for 4xxx?
757 2010-11-25 09:20:25 <ArtForz> no clue
758 2010-11-25 09:20:29 <Diablo-D3> none
759 2010-11-25 09:20:43 <Diablo-D3> the rest was just optimizing the host code to work optimally period.
760 2010-11-25 09:20:49 <ArtForz> btw, did you update to work with r189 getwork yet?
761 2010-11-25 09:20:52 <Diablo-D3> no
762 2010-11-25 09:20:57 <Diablo-D3> r189's getwork is broken
763 2010-11-25 09:21:00 <Diablo-D3> it needs to be fixed
764 2010-11-25 09:21:04 <ArtForz> whats broken?
765 2010-11-25 09:21:07 <xelister> ArtForz: so 5970 needs what 8 bit cable? EPS12V ?
766 2010-11-25 09:21:19 <ArtForz> xelister: PCIe 8pin and 6pin
767 2010-11-25 09:21:31 <Diablo-D3> ArtForz: the output is completely incompatible with m0's
768 2010-11-25 09:21:49 <ArtForz> and there is no way to tell em apart?
769 2010-11-25 09:21:54 <Diablo-D3> oh, there is
770 2010-11-25 09:21:56 <Diablo-D3> one works, the other doesnt
771 2010-11-25 09:22:05 <Diablo-D3> but Im not going to write code that Im going to just remove later
772 2010-11-25 09:22:08 <Diablo-D3> satoshi WILL fix is
773 2010-11-25 09:22:09 <Diablo-D3> (his
774 2010-11-25 09:22:11 <Diablo-D3> period
775 2010-11-25 09:22:17 <ArtForz> satoshis is imo coded a lot cleaner than m0s getwork
776 2010-11-25 09:22:18 <Diablo-D3> if he doesnt, then say goodbye to mining
777 2010-11-25 09:22:18 <xelister> ArtForz: so PCIe power cables are end with black plugs right? I have only 2 x 6 pin black-socket-ended cable... dont have 8 pin at all.
778 2010-11-25 09:22:28 <Diablo-D3> ArtForz: I dont give a shit if he coded it cleaner
779 2010-11-25 09:22:52 <Diablo-D3> xelister: you can convert molex plugs into 8 pin
780 2010-11-25 09:22:58 <Diablo-D3> xelister: just buy the converters
781 2010-11-25 09:23:14 <ArtForz> just make sure your PSU rails are beefy enough
782 2010-11-25 09:23:22 <xelister> Diablo-D3: I wonder if I do not need a nowe PSU. Anyway this one is just 550 W
783 2010-11-25 09:23:28 <ArtForz> 5970 seems to like to pull most of its power through the 8pin
784 2010-11-25 09:23:29 <Diablo-D3> xelister: 550 isnt enough
785 2010-11-25 09:23:42 <xelister> Diablo-D3: well not neseserly, it depends on the rails
786 2010-11-25 09:23:49 <ArtForz> 550W for a 5970 is asking for trouble
787 2010-11-25 09:23:55 <xelister> good 550 W > 700 W psu afaik
788 2010-11-25 09:23:56 <ArtForz> stock 5970 = 300W
789 2010-11-25 09:24:07 <ArtForz> decent CPU = 110W
790 2010-11-25 09:24:29 <xelister> so PCI-e 8 pin power is 5 V ? and takes from the 5 V rail?
791 2010-11-25 09:24:47 <xelister> and CPU take from the 12 V rail, right?
792 2010-11-25 09:24:48 <ArtForz> thats 410W on +12 alone, without chipset, ram, Hd or DVD drive
793 2010-11-25 09:25:18 <Diablo-D3> xelister: upgrade to that one corsair I found
794 2010-11-25 09:25:37 <Diablo-D3> http://www.newegg.com/Product/Product.aspx?Item=N82E16817139015
795 2010-11-25 09:25:40 <ArtForz> some DVD drives pull >30W on +12 when spinning up
796 2010-11-25 09:25:40 <Diablo-D3> vrrrrrm
797 2010-11-25 09:25:45 <xelister> but a decent PSU will cost like 200 USD it seems
798 2010-11-25 09:25:49 <ArtForz> nope
799 2010-11-25 09:25:54 <Diablo-D3> has 4x 6+2 plugs
800 2010-11-25 09:25:56 <Diablo-D3> and 850 watts
801 2010-11-25 09:26:03 <Diablo-D3> and is modular
802 2010-11-25 09:26:06 <xelister> THIS ALL SUCK DONKEY BALLS
803 2010-11-25 09:26:08 <ArtForz> a decent PSU for one or two 5970s is < $100
804 2010-11-25 09:26:13 <xelister> no wait, this is real life.
805 2010-11-25 09:26:20 <ArtForz> decent as in "wont blow up after 10 seconds"
806 2010-11-25 09:26:26 <Diablo-D3> ArtForz: meh
807 2010-11-25 09:26:42 <Diablo-D3> I wouldnt mind that corsair
808 2010-11-25 09:26:47 <xelister> so PCI-E, both 6 and 8 pin, takes from 12 V rail?
809 2010-11-25 09:26:51 <Diablo-D3> xelister: yes
810 2010-11-25 09:26:52 <ArtForz> yup
811 2010-11-25 09:26:58 <Diablo-D3> xelister: oh, and thats a single rail psu there
812 2010-11-25 09:27:02 <Diablo-D3> so no needing to deal with fucking bullshit
813 2010-11-25 09:27:12 <xelister> are there PSUs that have 2 rails for 12 V ?
814 2010-11-25 09:27:20 <Diablo-D3> xelister: BAD
815 2010-11-25 09:27:23 <Diablo-D3> >1 rail is shit psu
816 2010-11-25 09:27:24 <ArtForz> 2, 3, 4, 6, 8
817 2010-11-25 09:27:27 <ArtForz> nope
818 2010-11-25 09:27:30 <Diablo-D3> more rails == more fail
819 2010-11-25 09:27:34 <ArtForz> multirail is actually noit a BAD idea
820 2010-11-25 09:27:47 <ArtForz> take a single rail PSU, pull 50A through a single molex
821 2010-11-25 09:27:58 <ArtForz> oh look, glowing wires
822 2010-11-25 09:28:03 <Diablo-D3> yes, but you have to intelligently load balance your rails
823 2010-11-25 09:28:04 <xelister> which is better then? I have 1... say 2 x 5970, up to 6 hdd, strong intell quadcore, 4 ram banks, dvd recorder
824 2010-11-25 09:28:18 <xelister> what exactly PSU config I need, in rails power?
825 2010-11-25 09:28:36 <Diablo-D3> xelister: I have that, but single 4850 and a intel dual core
826 2010-11-25 09:28:37 <xelister> 12V: 350(*2) W for radeons,  plus... 30 W per hard-drive ?
827 2010-11-25 09:28:40 <Diablo-D3> xelister: and I use 350 real
828 2010-11-25 09:28:46 <ArtForz> on any decent multirail design you can load any plug to ... well.. whatever the rail limit is, and nothing will catch fire
829 2010-11-25 09:28:55 <Diablo-D3> xelister: so you need like 700 real
830 2010-11-25 09:29:19 <xelister> well 700 W real on 12 V single-rail is just the radeons, the CPU we said is like 100
831 2010-11-25 09:29:24 <xelister> so 850 W on real 12 V ?
832 2010-11-25 09:29:25 <ArtForz> yep
833 2010-11-25 09:29:57 <ArtForz> = for dual 5970s a 1kW PSU is a pretty good idea
834 2010-11-25 09:30:04 <xelister> sata hard drives use what rail?  it says sata power use all 3V 5V 12V
835 2010-11-25 09:30:16 <ArtForz> all 3.5" HDs use 12V for the motor
836 2010-11-25 09:30:20 <ArtForz> so do DVD drives
837 2010-11-25 09:30:32 <xelister> so 12 V mostly
838 2010-11-25 09:30:36 <ArtForz> yep
839 2010-11-25 09:30:37 <xelister> one hdd... 20 W ?
840 2010-11-25 09:30:41 <Diablo-D3> xelister: no
841 2010-11-25 09:30:43 <ArtForz> ~ 10W normally
842 2010-11-25 09:30:47 <ArtForz> 30W+ when spinning up
843 2010-11-25 09:30:47 <Diablo-D3> around 10
844 2010-11-25 09:30:57 <Diablo-D3> yeah, but staggered spin up fixes that
845 2010-11-25 09:31:07 <ArtForz> staggered spinup? in ma consumer box? on what planet?
846 2010-11-25 09:31:15 <Diablo-D3> sata does it
847 2010-11-25 09:31:33 <xelister> also what if all hard drives wake up from sleep in same time. yeah better be safe then sorry with hdds
848 2010-11-25 09:31:35 <ArtForz> did you test it with onboard and cheap fakeraid controllers?
849 2010-11-25 09:31:42 <Diablo-D3> ArtForz: yes
850 2010-11-25 09:31:53 <xelister> btw what hapens on overload? someone doesnt get power and GPU / hdd powers off, or something catches fire
851 2010-11-25 09:32:01 <Diablo-D3> xelister: on a good psu? it shuts down
852 2010-11-25 09:32:13 <Diablo-D3> on a bad psu? shit catches fire.
853 2010-11-25 09:32:23 <xelister> lol
854 2010-11-25 09:32:25 <ArtForz> on a good psu: it blows up
855 2010-11-25 09:32:26 <xelister> for real? cool
856 2010-11-25 09:32:30 <Diablo-D3> ArtForz: not true
857 2010-11-25 09:32:37 <Diablo-D3> unless it arcs for no fucking reason
858 2010-11-25 09:32:40 <ArtForz> on a bad psu: it blows up and takes out everything
859 2010-11-25 09:32:56 <xelister> I should buy halogen one.
860 2010-11-25 09:32:57 <Diablo-D3> unless you're doing like 4x what the PSU can do
861 2010-11-25 09:33:06 <ArtForz> nope, "only" 110%
862 2010-11-25 09:33:17 <Diablo-D3> ArtForz: I've tripped PSUs before
863 2010-11-25 09:33:22 <ArtForz> ... for a few hours ... at 45???C ambient
864 2010-11-25 09:33:24 <Diablo-D3> they just shut off with no damage
865 2010-11-25 09:33:24 <xelister> ArtForz: you use halogen extinguishers?  really should ;)
866 2010-11-25 09:33:27 <Diablo-D3> ArtForz: lol
867 2010-11-25 09:33:40 <Diablo-D3> xelister: dude, you cant buy a halogen extinguisher
868 2010-11-25 09:33:55 <ArtForz> and BLAM said the switching transistor in the active PFC
869 2010-11-25 09:34:09 <Diablo-D3> xelister: just use a fucking normal C02
870 2010-11-25 09:34:16 <Diablo-D3> er CO2
871 2010-11-25 09:34:21 <ArtForz> yeah, just get co2
872 2010-11-25 09:34:22 <xelister> Diablo-D3: and all electornics is ruined isnt it
873 2010-11-25 09:34:27 <ArtForz> huh? wtf?
874 2010-11-25 09:34:29 <xelister> well sure you CAN buy
875 2010-11-25 09:34:35 <ArtForz> it's fucking carbon dioxide
876 2010-11-25 09:34:37 <Diablo-D3> xelister: you need a license for halogen extinguishers iirc
877 2010-11-25 09:34:44 <xelister> you can even buy RPG if you really want to
878 2010-11-25 09:35:10 <ArtForz> overtemp didnt trip, overload didnt trip, nothing, it just went boom
879 2010-11-25 09:35:17 <Diablo-D3> ArtForz: wtf, shit psu
880 2010-11-25 09:35:17 <xelister> or buy children if you are with the government. or rights to grope children - just work with TSA. or to kill peopolel ,just join the army ;) </offtopic>
881 2010-11-25 09:35:27 <ArtForz> thermaltake toughpower 1kW
882 2010-11-25 09:35:27 <Diablo-D3> oh shut up xelister
883 2010-11-25 09:35:33 <Diablo-D3> ArtForz: thermaltake IS SHIT
884 2010-11-25 09:35:35 <Diablo-D3> SHIT SHIT SHIT
885 2010-11-25 09:35:36 <Diablo-D3> SHIT
886 2010-11-25 09:35:37 <Diablo-D3> _SHIT_
887 2010-11-25 09:35:39 <xelister> Diablo-D3: lol "you need a license"
888 2010-11-25 09:35:40 <Diablo-D3> all across the board
889 2010-11-25 09:35:44 <Diablo-D3> every single product
890 2010-11-25 09:35:45 <Diablo-D3> shit
891 2010-11-25 09:35:49 <Diablo-D3> psus, hsfs, cases
892 2010-11-25 09:35:51 <Diablo-D3> all shit
893 2010-11-25 09:36:13 <ArtForz> everythign is shit nowadays
894 2010-11-25 09:36:18 <Diablo-D3> true
895 2010-11-25 09:36:21 <xelister> except shit
896 2010-11-25 09:36:22 <Diablo-D3> I miss my silencers :<
897 2010-11-25 09:36:26 <Diablo-D3> best psu ever
898 2010-11-25 09:36:31 <Diablo-D3> fucking whores
899 2010-11-25 09:36:33 <Diablo-D3> they sold out to ocz
900 2010-11-25 09:36:35 <ArtForz> unless you get a $300+ industrial PSU RATED for 24/7 at full load
901 2010-11-25 09:36:37 <Diablo-D3> nuke china
902 2010-11-25 09:36:45 <Diablo-D3> ArtForz: yeah but they never have the good shit
903 2010-11-25 09:37:00 <xelister> so I need 650 W on 12V single-rail for single radeon,  and 1000 W for same for double, correct?
904 2010-11-25 09:37:06 <ArtForz> what?
905 2010-11-25 09:37:10 <xelister> anything other to pay attention to when buying PSU?
906 2010-11-25 09:37:11 <Diablo-D3> xelister: no
907 2010-11-25 09:37:25 <Diablo-D3> xelister: you need ~600 watts for 2x 5970
908 2010-11-25 09:37:32 <xelister> 40*5 + 350*1 + 100 = 650 W,   5 hard drives, radeon, cpu
909 2010-11-25 09:37:36 <ArtForz> stock
910 2010-11-25 09:37:47 <Diablo-D3> so a 850w psu SHOULD do what you want
911 2010-11-25 09:37:53 <ArtForz> a OCed 5970 is 350-400W
912 2010-11-25 09:38:08 <xelister> Im probably not going to OC, btw
913 2010-11-25 09:38:20 <ArtForz> then 300W per 5970
914 2010-11-25 09:38:21 <Diablo-D3> ArtForz: well like, too many PSUs dont have fucking pci-e plugs
915 2010-11-25 09:38:24 <Diablo-D3> ArtForz: or not enough of them
916 2010-11-25 09:38:25 <xelister> so what, 650 on 12 rail is too high?
917 2010-11-25 09:38:37 <xelister> count just 20 for hdds?
918 2010-11-25 09:38:49 <xelister> then 550 on 12V rail
919 2010-11-25 09:38:50 <Diablo-D3> xelister: you could do like 800 watts on a 12v rail if your PSU is beefy enough
920 2010-11-25 09:38:56 <Diablo-D3> http://www.newegg.com/Product/Product.aspx?Item=N82E16817139014
921 2010-11-25 09:38:58 <Diablo-D3> I mean hell
922 2010-11-25 09:39:01 <Diablo-D3> look at that
923 2010-11-25 09:39:13 <Diablo-D3> 1200 watts == the maximum you're going to get out of a 15 amp plug
924 2010-11-25 09:39:13 <xelister> no, Im asking,  how I should describe the PSU I need
925 2010-11-25 09:39:20 <Diablo-D3> xelister: you need a 850 watt psu.
926 2010-11-25 09:39:27 <Diablo-D3> Ive told you this three fucking times
927 2010-11-25 09:39:50 <xelister> Diablo-D3: Im not going to ask about stock W power in the name like a noob, but about the 12 V rail power
928 2010-11-25 09:39:52 <ArtForz> 850W? loaded to 600W 24/7? you like living on the edge...
929 2010-11-25 09:40:08 <Diablo-D3> ArtForz: only if he doesnt load the CPU
930 2010-11-25 09:40:19 <Diablo-D3> if he does, meh.
931 2010-11-25 09:40:26 <Diablo-D3> get that 1200 watt PSU
932 2010-11-25 09:40:28 <ArtForz> well, with cpu loaded it's closer to 700W laod
933 2010-11-25 09:40:29 <Diablo-D3> its only like $300
934 2010-11-25 09:40:47 <xelister> so should I count 40 or 20 W per sata hdd on a normal PC?
935 2010-11-25 09:40:50 <Diablo-D3> $278
936 2010-11-25 09:40:56 <Diablo-D3> xelister: 10.
937 2010-11-25 09:41:00 <xelister> will it do delayed spin up really?
938 2010-11-25 09:41:04 <Diablo-D3> remember they use 5v too
939 2010-11-25 09:41:42 <ArtForz> and btw, that corsair lowks like a CWT
940 2010-11-25 09:42:01 <xelister> 5hdd, 1 radeon, CPU, NIC:  10*5 + 350*1 + 120   = 520 W on 12 V rail.
941 2010-11-25 09:42:29 <xelister> so 520 W psu, with 6 and 8 pin PCI-e power calbes long enough to reach radeon.  Is this correct?  Anythiong else to ask for in the radeon?
942 2010-11-25 09:42:31 <xelister> in the psu
943 2010-11-25 09:42:36 <ArtForz> if it is what I think it is, it's crap
944 2010-11-25 09:46:26 <ArtForz> ahhh. flextronics OEM
945 2010-11-25 09:46:30 <xelister> anything else I need in the SPU, then 520 W in 12V rail? what else is important
946 2010-11-25 09:48:49 <ArtForz> $280 for that piece of junk is a sad joke
947 2010-11-25 10:04:58 <xelister> ArtForz: why its junk
948 2010-11-25 10:05:15 <ArtForz> look at the internal design
949 2010-11-25 10:06:10 <xelister> modecom-s are good?
950 2010-11-25 10:06:20 <xelister> modecom hidi 650 bullpower
951 2010-11-25 10:06:36 <xelister> hedy
952 2010-11-25 10:06:44 <ArtForz> modewhat?
953 2010-11-25 10:07:07 <xelister> like this http://www.ecopc.redcart.pl/p/489/66428/zasilacz-modecom-hedy-650w-bull-power-zasilacze-do-obudow-zasilanie.html
954 2010-11-25 10:08:07 <ArtForz> does that thing even have a 80+ rating?
955 2010-11-25 10:08:20 <ArtForz> and a single PCIe connector?
956 2010-11-25 10:08:31 <xelister> other PCIe woudl be from molex
957 2010-11-25 10:08:42 <xelister> what 80+ rating?
958 2010-11-25 10:08:57 <ArtForz> exactly
959 2010-11-25 10:09:01 <ArtForz> what 80+ rating
960 2010-11-25 10:09:25 <xelister> I mean, what is "80+ rating" :)
961 2010-11-25 10:09:45 <ArtForz> http://en.wikipedia.org/wiki/80_PLUS
962 2010-11-25 10:10:00 <xelister> http://en.wikipedia.org/wiki/80_PLUS
963 2010-11-25 10:10:03 <xelister> ah. yeah
964 2010-11-25 10:10:31 <ArtForz> the interesting part about 80+ certification is that they measure efficiency at 20, 50 and 100% rated load
965 2010-11-25 10:11:05 <ArtForz> = to survive testing PSU has to be able to handle 100% rated load at least for a short time
966 2010-11-25 10:11:23 <xelister> tacens radix 4, 600 W
967 2010-11-25 10:12:25 <ArtForz> that should be a HEC oem
968 2010-11-25 10:13:38 <xelister> that looks ok for 1x5970 + 6 harddries + intel quadcore ?  http://www.testseek.com/computers/powersupply/tacens_radix_4_600w-p-ce1daca4-7364-f223-bfaa-d99d0ad5f4ad.html
969 2010-11-25 10:13:41 <ArtForz> it's a oem of the cheap/crap HEXs, stay away