**** BEGIN LOGGING AT Wed Jun 29 23:59:56 2005 Jun 30 00:00:03 no...lemme know! Jun 30 00:00:11 root@maimedslug:~# cat /proc/cpuinfo Jun 30 00:00:11 Processor : XScale-IXP42x Family rev 1 (v5b) Jun 30 00:00:11 BogoMIPS : 266.24 Jun 30 00:00:24 OCSlug :) Jun 30 00:00:36 LE or BE? Jun 30 00:00:47 be Jun 30 00:00:53 Either, the core doesn't care. Jun 30 00:01:08 OK, can someone explain the technical details (not the hardware mod, the circuitry change with respect to IXP signals) ? Jun 30 00:01:10 The core has no endianness Jun 30 00:01:51 At power on and at reset the cpu reads A21:23 off the address line to determine the speed of the core clock. Jun 30 00:02:02 ka6sox: I've removed one resistor (R64). That's it. Took a whole whopping 10 seconds ;) Jun 30 00:02:12 But R64 might not be the best one to remove Jun 30 00:02:18 according to jbowler here Jun 30 00:02:31 These lines float high, but are pulled down internally (in the core) and may be pulled down externally. Jun 30 00:02:34 very cool guys! Jun 30 00:02:37 jbowler-zzz: what's the encodings on thost POR signals Jun 30 00:03:03 ka6sox: It's dyoung, bewoolie, jbowlers fault, I think. Jun 30 00:03:23 you guys are awesome! Jun 30 00:03:27 It's described in the documentation of Configuration Register 0 of the Expansion Bus Interface in the Intel Developers Manual. Jun 30 00:03:58 External to the core the encoding is accomplished by a simple 10k pull down resistor - no logic, no switching. Jun 30 00:04:06 okay so I'm worried about things getting hot. Jun 30 00:04:14 ka6sox: The core might generate a bit more heat though. I'm running it without the case, and it's somewhere around 40C. Jun 30 00:04:19 did you change the core freq... Jun 30 00:04:28 (so there is a pulldown. Jun 30 00:04:31 cool. Jun 30 00:04:36 The mapping is A23,22,21 R64,R81,R83 Jun 30 00:04:44 ah Jun 30 00:04:49 R64 and R83 are present, R81 is absent. Jun 30 00:04:52 okay got the wiki page updated yet? Jun 30 00:04:59 I've got a page ready Jun 30 00:05:02 Not saved yet. Jun 30 00:05:07 So the NSLU encodes b010 - an undefined value in the manual. Jun 30 00:05:15 you go NAiL! Jun 30 00:05:21 jbowler-zzz: url for the doco? Jun 30 00:05:25 Removing R64 changes this to b110 - also undefined Jun 30 00:05:26 ixp doco Jun 30 00:05:54 The intel part is 25248005, let me see if I can generate a URL Jun 30 00:07:08 this one: http://www.intel.com/design/network/manuals/252480.htm ? Jun 30 00:07:44 http://www.nslu2-linux.org/wiki/HowTo/OverClockTheSlug <-- feel free to correct Jun 30 00:07:50 http://download.intel.com/design/network/manuals/25248005.pdf Jun 30 00:08:35 pg 324 Jun 30 00:08:38 Looks like it, check chapter 8 Jun 30 00:09:03 Table on pg 325,326 Jun 30 00:09:53 My suggestion is to add 10k to R81 - this produces b000, which I interpret as 'clock to the max'... Jun 30 00:10:53 I'm guessing that the clock multiplier goes b010(133) b011(266) b001(400) b100(533), but I don't know how clock multipliers work... Jun 30 00:11:20 or 100 Jun 30 00:11:56 what does NAiL have? Jun 30 00:12:09 110 Jun 30 00:13:05 Of R64, R81, R83, I only have R83. Jun 30 00:13:35 so they are puuldowns? Jun 30 00:13:48 internally pulled up or down Jun 30 00:13:54 and externally pulled up or down? Jun 30 00:14:01 Yes - see 8.9.9.1 Jun 30 00:14:04 int pullup, ext pulldown IIRC Jun 30 00:15:00 removing is easier than adding Jun 30 00:15:02 Also see ftp://download.intel.com/design/network/applnots/25406701.pdf Jun 30 00:15:32 so we're changing 010 to 110, 011 or 111. Jun 30 00:15:37 Removing R81 gives b011, 266MHz - that's probably safest. Jun 30 00:16:10 the original plan (formulated in openjtag) was to move R64 to R81 Jun 30 00:16:26 but I believe beewoolie removed R64 and just wanted to see what would happen Jun 30 00:16:58 the first bit doesn't seem to matter in the table Jun 30 00:18:20 Yes, but it matters in practice - b110 != b010 Jun 30 00:18:39 (we've just proved that). Jun 30 00:18:44 indeed Jun 30 00:19:13 so someone else needs to try 011 (remove R81 only) Jun 30 00:19:36 then we're in the documented space Jun 30 00:19:57 IIRC there are higher frequency parts (650MHz?) available now (ixp625?) they may document more settings. Jun 30 00:20:25 oops: 011 is remove R83 only isn't it? Jun 30 00:20:48 only R64 present Jun 30 00:20:55 ok, my excitement has been overcome by confusion and lack of sleep Jun 30 00:21:14 * ByronT-Away puts down the soldering iron Jun 30 00:21:20 I'm heading to bed! Jun 30 00:22:05 so what's the official name? SpeedSlug, FastSlug ? Jun 30 00:22:19 SpeedySlug Jun 30 00:22:33 SlungOutOfThePark? Jun 30 00:23:04 who did the mod first? Jun 30 00:23:14 beewoolie Jun 30 00:30:43 On the IXP465 the core clock factory speed is recorded in 2 bits as 00(533), 01(400), 10(667), 11(266). Jun 30 00:35:55 which matches our bottom two? Jun 30 00:36:44 Yes, apart from the undocumented one - 010. Jun 30 06:27:15 Is it normal I still get "onotone: network exception: name resolution failure for ewi546.ewi.utwente.nl" error when doing make setup? Jun 30 06:27:42 I'm refering to the new monotone build system using the Makefile. Jun 30 08:53:22 Yo peeps Jun 30 08:54:30 beewoolie-afk, http://groups.yahoo.com/group/nslu2-linux/database?method=reportRows&tbl=11 **** ENDING LOGGING AT Thu Jun 30 23:59:57 2005