**** BEGIN LOGGING AT Fri Jan 10 02:59:59 2014
Jan 10 03:13:11 <SAiF> :)
Jan 10 12:36:49 * sixwheeledbeast checks scrollback regarding smartreflex.
Jan 10 12:50:29 <sixwheeledbeast> SR is considered unstable by TI and Nokia for stock kernel. Whereas SR is considered stable by KP team on kp50> up to ~900Mhz I believe. I would like to hear KP teams opinion on SR, myself.
Jan 10 12:58:14 <ArkanoiD_> I had one n900 that ran fine with SR and one that failed miserably when SR and OC were both enabled
Jan 10 12:58:30 <ArkanoiD_> so it is unpredictable
Jan 10 13:09:13 <hxka> ArkanoiD_: When that was?
Jan 10 13:16:28 <hxka> ArkanoiD_: IIRC you swithed to N950 in 2011, that's before kp50 was released
Jan 10 13:20:25 <ArkanoiD_> it think it was 2012. at least i remember i spent some time playing with OC and power saving options
Jan 10 13:26:32 <TheCakeIsAlive> Even in 2012 it could be kp49
Jan 10 13:50:24 <DocScrutinizer05> sixwheeledbeast: KP team is rejecting/ignoring the Nokia/TI finding, based on own assumptions about why SR got disabled as well as a general WFM rationale
Jan 10 13:50:41 <DocScrutinizer05> so: based on essentially nothing
Jan 10 13:50:46 <DocScrutinizer05> ;-D
Jan 10 13:51:41 <ArkanoiD_> we were told OC is dangerous all the time, but for all years passed not a single toasted n900 yet to be seen
Jan 10 13:51:53 <DocScrutinizer05> haha
Jan 10 13:52:03 <jaska> 55.5% oc seems a lot.
Jan 10 13:52:40 <jaska> the most ive done on a pc was 59%, and that was a l2-cacheless celeron266 to 450
Jan 10 13:52:46 <DocScrutinizer05> sure, the gallery labeled "look this device fails on video playback all the time, i probably cooked it by OC" is empty for sure
Jan 10 13:52:50 <ArkanoiD_> and Nokia's default OS tuning (swappiness, scheduler etc) proved to be all wrong
Jan 10 13:53:18 <jaska> erm hmm
Jan 10 13:53:27 <jaska> wrong %s
Jan 10 13:53:29 <jaska> stupidme
Jan 10 13:53:49 <jaska> 80% oc that is
Jan 10 13:54:04 <jaska> and the celeron is 69%
Jan 10 13:55:02 <DocScrutinizer05> fact is: TI checked life expectation of their chips and declared it to be 100k-hours on 500MHz and 15kh on 600MHz
Jan 10 13:55:09 <DocScrutinizer05> no extrapolate that
Jan 10 13:55:33 <DocScrutinizer05> now*
Jan 10 13:55:45 <jaska> could use a couple more data points there :)
Jan 10 13:55:53 <ArkanoiD_> but they did not take lower voltage and more sleep state into account
Jan 10 13:56:12 <DocScrutinizer05> what will it be at 800MHz? 1500h? 500h?
Jan 10 13:57:02 <DocScrutinizer05> lower voltage is another gambling
Jan 10 13:57:21 <DocScrutinizer05> and sleep states are not related to the above figure
Jan 10 13:57:35 <DocScrutinizer05> except for inplicitly not applying
Jan 10 13:57:51 <DocScrutinizer05> ~omap-oc
Jan 10 13:57:52 <infobot> omap-oc is, like, http://mg.pov.lt/maemo-irclog/%23maemo.2010-08-01.log.html#t2010-08-01T22:16:05  read that!, or and this http://mg.pov.lt/maemo-irclog/%23maemo.2011-03-11.log.html#t2011-03-11T03:04:11
Jan 10 13:59:15 <DocScrutinizer05> I simply say that TI's and nokia's statements on OC make sane and are yet to get proven wrong. While all the "we haven't seen yet" is an age old moot statement
Jan 10 13:59:50 <DocScrutinizer05> s/sane/sense/
Jan 10 13:59:51 <infobot> DocScrutinizer05 meant: I simply say that TI's and nokia's statements on OC make sense and are yet to get proven wrong. While all the "we haven't seen yet" is an age old moot statement
Jan 10 14:01:04 <DocScrutinizer05> sur you can OC omap, when you are willing to take compromises on stability and endurance of the product. Nokia opted for the former rather than the OC
Jan 10 14:03:28 <ArkanoiD_> I am yet to see a single n900 that is unstable at 900MHz
Jan 10 14:10:47 <freemangordon> ArkanoiD_: the question is - at what voltage?
Jan 10 14:11:29 <freemangordon> most oh them need voltage that is above that for 600MHz without SR
Jan 10 14:11:45 <freemangordon> like 1375mV and above
Jan 10 14:14:26 <freemangordon> DocScrutinizer05: I am not speculating on the reason why SR is not enabled, it is a fact of life that OMAP in n900 has calibration screwed (by TI I guess) for OPP 3 4 and 5
Jan 10 14:15:06 <DocScrutinizer05> which is the whole point of the core volatge reglation, it's not meant for user to pick what looks most pretty to him7her, but for adjusting CPU voltage to the needs of that clock frequency. Higher clock needs _higher_ core voltage, not lower
Jan 10 14:15:41 <freemangordon> sure
Jan 10 14:16:09 <freemangordon> but without SR CPU runs on overvoltage most of the time
Jan 10 14:16:19 <DocScrutinizer05> yes
Jan 10 14:16:35 <DocScrutinizer05> particularly on the seldomly used 250MHz
Jan 10 14:17:16 <DocScrutinizer05> unless some smartass implemented "manual" core voltage adjustment into freq-governor
Jan 10 14:17:24 <freemangordon> hmm, which reminds me, I should send a patch to Pali to increase a slope of SR calibration recalculation function in KP
Jan 10 14:18:07 <DocScrutinizer05> SR is just about *automatic* adjustment. "manual" adjustment can be done any time
Jan 10 14:18:27 <freemangordon> I'd trabe 1-2% of battery life to be on the safe side
Jan 10 14:18:37 <freemangordon> DocScrutinizer05: sure
Jan 10 14:19:09 <DocScrutinizer05> tel4030 using a dedicated I2C link to CPU/SoC for SR exclusively
Jan 10 14:19:14 <freemangordon> but what SR driver in KP does is not a blind "manual adjustment"
Jan 10 14:19:22 <freemangordon> :nod: there is
Jan 10 14:19:27 <DocScrutinizer05> maybe that link has data integrity issues
Jan 10 14:19:40 <freemangordon> why do you think that?
Jan 10 14:20:03 <DocScrutinizer05> I don't suggest it has, I'm just showing possibilities
Jan 10 14:20:30 <DocScrutinizer05> there's quite a bunch of possible issues that made SR instable
Jan 10 14:21:17 <freemangordon> DocScrutinizer05: there is exactly 1 reason why SR is unstable on n900 - the calibration values of OPP 3 4 and 5 are bogus. period
Jan 10 14:21:35 <DocScrutinizer05> and when Nokia and TI(!) detected a problem they didn't know how to fix nor where's the root cause, I'd consider that a lost case
Jan 10 14:22:17 <DocScrutinizer05> you realy say TI is not expereinced enough to detect and fix that obvious issue?
Jan 10 14:22:36 <freemangordon> DocScrutinizer05: I would really love some of the ex-Nokians hanging on the channel to prove me wrong re screwed calibration
Jan 10 14:22:46 <freemangordon> DocScrutinizer05: it is in efuse ;)
Jan 10 14:22:56 <DocScrutinizer05> nobody doubts the calibration is wrong
Jan 10 14:23:00 <freemangordon> when it is detected, I'd say it is too late
Jan 10 14:23:31 <freemangordon> DocScrutinizer05: so why you search for some other hipothetical "integrity issues" etc when there is an obvious reason
Jan 10 14:23:32 <DocScrutinizer05> when you can enable and fix it, you think Nokia and even TI couldn't?
Jan 10 14:23:47 <freemangordon> DocScrutinizer05: did they care?
Jan 10 14:23:53 <DocScrutinizer05> yes, they did
Jan 10 14:24:00 <freemangordon> is is the same as with thumb IMO
Jan 10 14:24:04 <DocScrutinizer05> ~sr
Jan 10 14:24:05 <infobot> [SmartReflex] >>Again, TI and we [Nokia] couldn't fix SmartReflex - we say memory corruption in front of our own eyes with that enabled, so we had to ship that disabled.<<
Jan 10 14:24:15 <freemangordon> noone took the risk
Jan 10 14:25:09 <freemangordon> DocScrutinizer05: sure you'll have memory corruption if CPU is undervolted
Jan 10 14:25:34 <DocScrutinizer05> when wrong SR calibration would have been the clearly spotted culprit, then there wouldn't be *any* risk in fixing it the same way you did
Jan 10 14:26:19 <freemangordon> again - why do you think anyone did care enough to fix it?
Jan 10 14:26:58 <DocScrutinizer05> which is a proof that your fix is NO valid fix for the original issue, just catching up on some fallout from not caring about SR after it got aboandoned and deprecated due to another problem TI and NHokia spotted. At least in my book
Jan 10 14:27:28 <freemangordon> DocScrutinizer05: "my" fix is not different from the "USB host mode" IMO
Jan 10 14:27:29 <DocScrutinizer05> again, because I talked to the guy who stated the above quote
Jan 10 14:27:41 <freemangordon> they said the same for the host mode, ain't?
Jan 10 14:27:47 <DocScrutinizer05> no
Jan 10 14:27:52 <freemangordon> ok
Jan 10 14:28:02 <DocScrutinizer05> and they been right about hostmode
Jan 10 14:28:11 <freemangordon> I think we already had that conversation :)
Jan 10 14:28:17 <freemangordon> several times :D
Jan 10 14:28:17 <DocScrutinizer05> it doesn't work the way it's supposed, in N900
Jan 10 14:28:27 <freemangordon> the point is - it works
Jan 10 14:28:42 <DocScrutinizer05> the point is, it doesn't, really
Jan 10 14:28:57 <DocScrutinizer05> what you see *looks* like hostmode, but isn't
Jan 10 14:28:57 * ShadowJK looks at output of first line of cpustat.sh..
Jan 10 14:29:06 <ShadowJK> 92% cpu off :)
Jan 10 14:29:17 <ShadowJK> (over 4 days uptime)
Jan 10 14:29:27 <TheCakeIsAlive> ArkanoiD_: My N900 is unstable at anything above 600 MHz
Jan 10 14:30:03 <DocScrutinizer05> so much for "I've yet to see any device that..."
Jan 10 14:30:26 <freemangordon> and I am sure my SR recalibration outputs a slighly higher voltages than those if TI didn't screw it
Jan 10 14:30:41 <DocScrutinizer05> it's easy to not see any poor man when you have a habit to look the other direction whenever you pass by at one
Jan 10 14:31:04 <freemangordon> ShadowJK: with SIM card in it?
Jan 10 14:31:30 <freemangordon> TheCakeIsAlive: what voltages?
Jan 10 14:31:53 <ShadowJK> freemangordon; with sim
Jan 10 14:31:54 <TheCakeIsAlive> freemangordon: default profile of k-p-s
Jan 10 14:32:09 <freemangordon> afk
Jan 10 14:33:45 * DocScrutinizer05 thinks that all statements like "I am yet to see a single n900 that is..." are generally just rhetorical
Jan 10 14:34:28 <DocScrutinizer05> I'm yet to see any N900 besides my 8 and a maybe 5 from freinds
Jan 10 14:34:36 <ShadowJK> hm.. so if CPU had lifetime of 1000h, and 92% was true for longer periods than 4 days, the cpu would still last longer than the modem has in my 4 previous N900
Jan 10 14:34:57 <DocScrutinizer05> sure
Jan 10 14:35:57 <DocScrutinizer05> and that's what I think is pretty much exactly what we see in RL regarding all the OC stuff
Jan 10 14:38:00 <DocScrutinizer05> prolly CPU life expectation is actually reduced to 500...2000h, and just the idle/OPP800 ratio spreads that time over a reasonable lifespan of device usage - **under standard usecase conditions**
Jan 10 14:38:27 <DocScrutinizer05> doing a lot of VoIP calls or whatever can significantly change that picture
Jan 10 14:40:22 <DocScrutinizer05> also the same low active/idle ratio makes for a significantly lower chip temperature, which again mitigates the electromigration
Jan 10 14:42:11 <ShadowJK> I don't remember seeing the cpu temp meter react at all unless I make extra efforts
Jan 10 14:43:11 <DocScrutinizer05> otoh a 800MHz OC on a device that's running on 100% CPU busy will probably not only burn out the CPU from overtemperature after 30min, but as well cause the chip temeprature exceed the 105°C specified for the OPP numbers TI stated, and thus you start with a life expectation that's way lower than those 10kh resp 50kh TI claimed for normal operation
Jan 10 14:43:34 <ShadowJK> hm
Jan 10 14:44:45 <ShadowJK> where is that spreadsheet... I did work/watt figures well into OC range to find the "sweet spot" of the cpu
Jan 10 14:45:08 <DocScrutinizer05> maybe that been a bit fuzzy. lemme retry: TI statements as of http://mg.pov.lt/maemo-irclog/%23maemo.2011-03-11.log.html#t2011-03-11T03:04:11 are for a max die temp of 105°C (A chip)
Jan 10 14:46:12 <DocScrutinizer05> when you run OC and the CPU is 100% busy and not idling a lot, then you for sure exceed that temperature, making the lifetime of CPU even shorter than it would be for mere OC alone
Jan 10 14:46:51 * ShadowJK sacrifices device
Jan 10 14:47:09 <ShadowJK> ok, 900MHz, 100% busy, how long to hit 50C do you think?
Jan 10 14:47:12 <DocScrutinizer05> thus the summed up CPU actibe time @800MHz will be considerable shorter when CPU idles less in between
Jan 10 14:47:28 <DocScrutinizer05> depends on what else the device does
Jan 10 14:47:36 <DocScrutinizer05> thermal design is pretty complex
Jan 10 14:48:01 <DocScrutinizer05> last 2 @ ShadowJK
Jan 10 14:48:17 <ShadowJK> well, xchat on wifi, 3g standby, screen on
Jan 10 14:48:35 <DocScrutinizer05> I think device heats up most from WWAN traffic
Jan 10 14:49:01 <DocScrutinizer05> and CPU/SoC has a TDP of maybe <1W
Jan 10 14:49:18 <ShadowJK> yep
Jan 10 14:49:33 <DocScrutinizer05> this will get easily dissipated by PCB groundplane, unless it gets heat from other components as well
Jan 10 14:49:42 <ShadowJK> My first got to 65C without OC or even 100% cpu
Jan 10 14:50:34 <DocScrutinizer05> it's all a very complex equation to evaluate
Jan 10 14:52:05 <DocScrutinizer05> will you get away with 800MHz OC for 3 years? yes, probably. Will this guarantee that the device doesn't burn out in 4 weeks under certian usage conditions? definitely not
Jan 10 14:53:00 <DocScrutinizer05> watching video for 8h a day during your holiday could already suffice to kill the device for good
Jan 10 14:53:46 <ShadowJK> video for me seems to be mostly 500MHz, with occasional 550 and 250
Jan 10 14:53:50 <DocScrutinizer05> or whatever else usecase, the ebove I made up without good backup from numbers
Jan 10 14:54:48 <ShadowJK> 5min, bq27200 gone from 33C to 37C, SoC still reporting 27
Jan 10 14:54:49 <DocScrutinizer05> so make that 3 months of manager hassle with a 6h of phonecalls per day, or whatever
Jan 10 14:55:08 <DocScrutinizer05> SoC temp sensor freezes eventually
Jan 10 14:55:23 <DocScrutinizer05> it's fubar
Jan 10 14:55:47 <DocScrutinizer05> SiErr, known
Jan 10 14:58:14 <DocScrutinizer05> OC is not about "it's safe / not safe", it's all about "what amount of reliability and lifetime reduction I'm willing to accept. And particularly whicj ammount of uncertainty regarding the aforementioned parameters?"
Jan 10 14:59:09 <ShadowJK> 10min 39C, 413-420 mA
Jan 10 14:59:30 <DocScrutinizer05> without OC, the device is probably guaranteed to stay alive for 3 years continuous usage, no matter what you do to it
Jan 10 15:00:16 <DocScrutinizer05> with OC... up to everybody's guess, but defintely magnitudes less
Jan 10 15:01:01 <DocScrutinizer05> magnitudes literaly
Jan 10 15:01:15 <DocScrutinizer05> like /10 or /100
Jan 10 15:01:27 <DocScrutinizer05> even /1000
Jan 10 15:01:56 <DocScrutinizer05> depending on your usecase
Jan 10 15:02:09 <ShadowJK> 300mA @ 600MHz, 39C
Jan 10 15:02:42 <DocScrutinizer05> hmm yeah, I've however seen bq27k logs where device ate >1A
Jan 10 15:03:00 <ShadowJK> I've managed 1.3A :)
Jan 10 15:03:20 <DocScrutinizer05> and most of that in the end converted to heat to dissipate
Jan 10 15:03:45 <ShadowJK> can't do that anymore, 3g signal too good everywhere now
Jan 10 15:03:53 <DocScrutinizer05> hehe
Jan 10 15:03:55 <DocScrutinizer05> yeah
Jan 10 15:04:15 <DocScrutinizer05> it's a complex equation with a lot of variables you can't control
Jan 10 15:04:36 <DocScrutinizer05> especially WWAN
Jan 10 15:04:59 <ShadowJK> xchat on gsm out at sea is a pretty good power hog too
Jan 10 15:05:06 <DocScrutinizer05> we had a good share of lessons regarding that, when trying to reproduce buzz issue at openmoko
Jan 10 15:06:11 <ShadowJK> Did I tell you I overclocked an AC induction motor at $work? :)
Jan 10 15:06:30 <DocScrutinizer05> "when I hold the device out the window, buzz is gone. When I get it back into room, touching the window, buzz returns after 5 seconds like somebody switched it on"
Jan 10 15:07:05 <ShadowJK> The plastic fan blade disintegrated at 115 Hz :o(
Jan 10 15:07:19 <DocScrutinizer05> the joy of you not controlling the complete testbed - here: BTS
Jan 10 15:07:38 <DocScrutinizer05> LOL at motor-OC
Jan 10 15:08:54 <DocScrutinizer05> finally I found that placing the Freerunner on a steel table was a good method to make it ramp up TX power and thus start buzz
Jan 10 15:08:58 <ShadowJK> It's running happily at 50Hz now though, with a tribladed fan that used to be four-bladed :D
Jan 10 15:09:24 <DocScrutinizer05> :-o
Jan 10 15:09:32 <jaska> wobble wobble
Jan 10 15:10:27 <ShadowJK> the desk I used to place my n900 on for service as modem, had bunch of papers, and underneath a shielded usb cable that acted as antenna booster for 3g
Jan 10 15:10:55 <DocScrutinizer05> :nod:
Jan 10 15:11:03 <ShadowJK> only found out when speed collapsed from 5Mbit/s to .5 when I needed the cable
Jan 10 15:11:09 <kerio> lmao
Jan 10 15:11:43 <DocScrutinizer05> good USB cable
Jan 10 18:22:32 * ShadowJK enjoys a beer from Nokia Brewery
Jan 10 18:58:22 <DocScrutinizer05> Nokia brewery ?? o.O
Jan 10 18:58:45 <DocScrutinizer05> no more rubberboots?
Jan 10 19:03:03 <jaska> rubberboot malts
Jan 10 20:01:42 <ShadowJK> (Brewery in the city Nokia)
Jan 10 21:54:00 <DocScrutinizer05> city Nokia? OMG
Jan 10 21:54:48 <DocScrutinizer05> the heck! http://en.wikipedia.org/wiki/Nokia,_Finland
Jan 10 23:16:46 <sixwheeledbeast> DocScrutinizer05: I must thank you for your wisdom :)
Jan 10 23:17:00 <DocScrutinizer05> hmm?
Jan 10 23:18:14 <DocScrutinizer05> do I sense sarcasm?
Jan 10 23:18:39 <sixwheeledbeast> I have added initial first startup information to FlopSwap
Jan 10 23:18:47 <DocScrutinizer05> ooh
Jan 10 23:19:02 <sixwheeledbeast> and in the process found a bug and fixed that too ;)
Jan 10 23:19:10 <DocScrutinizer05> hah!
Jan 10 23:19:13 <DocScrutinizer05> :-)
Jan 11 00:20:56 <ChiaSmurf> did microsoft buy that too?
Jan 11 00:26:45 <DocScrutinizer05> ChiaSmurf: eh?
Jan 11 00:27:03 <ChiaSmurf> the town
Jan 11 00:27:05 <DocScrutinizer05> ooh, Nokia city?
Jan 11 00:27:20 <DocScrutinizer05> possibly ;-P
Jan 11 00:27:24 <ChiaSmurf> rename it billburg
Jan 11 00:27:31 <DocScrutinizer05> rotfl
Jan 11 00:31:13 <DocScrutinizer05> copying a private query since it should belong to #maemo to start with. sorry for the flooding
Jan 11 00:31:24 <DocScrutinizer05> [2014-01-11 01:13:54] <vakkov_> Hi! Do you know any way to see the dmesg of n900 other than serial console? I am debugging a kernel here but the panic about missing symbols disappears too fast
Jan 11 00:31:25 <DocScrutinizer05> [2014-01-11 01:14:27] <DocScrutinizer05> lemme think a while
Jan 11 00:31:27 <DocScrutinizer05> [2014-01-11 01:14:56] <DocScrutinizer05> I gather you're not interested in dmesg but in early boot console output
Jan 11 00:31:28 <DocScrutinizer05> [2014-01-11 01:16:03] <DocScrutinizer05> basically kernel should dump any oops and panic to log partition in NAND
Jan 11 00:31:30 <DocScrutinizer05> [2014-01-11 01:16:09] <DocScrutinizer05> cat /proc/mtd
Jan 11 00:31:31 <DocScrutinizer05> [2014-01-11 01:17:06] <DocScrutinizer05> IroN900:~# sp-oops-extract --help
Jan 11 00:31:33 <DocScrutinizer05> [2014-01-11 01:17:06] <DocScrutinizer05> Oops Log extractor (build Feb  4 2009 16:33:27).
Jan 11 00:31:34 <DocScrutinizer05> [2014-01-11 01:17:06] <DocScrutinizer05> sp-oops-extract: Unable to stat file '--help': No such file or directory
Jan 11 00:31:36 <DocScrutinizer05> [2014-01-11 01:17:36] <DocScrutinizer05> IroN900:~# cat /proc/mtd
Jan 11 00:31:37 <DocScrutinizer05> [2014-01-11 01:17:36] <DocScrutinizer05> dev:    size   erasesize  name
Jan 11 00:31:39 <DocScrutinizer05> [2014-01-11 01:17:36] <DocScrutinizer05> mtd0: 00020000 00020000 "bootloader"
Jan 11 00:31:40 <DocScrutinizer05> [2014-01-11 01:17:36] <DocScrutinizer05> mtd1: 00060000 00020000 "config"
Jan 11 00:31:42 <DocScrutinizer05> [2014-01-11 01:17:36] <DocScrutinizer05> mtd2: 00040000 00020000 "log"
Jan 11 00:31:43 <DocScrutinizer05> [2014-01-11 01:18:13] <DocScrutinizer05> IroN900:~# cat /proc/cmdline 
Jan 11 00:31:45 <DocScrutinizer05> [2014-01-11 01:18:13] <DocScrutinizer05> init=/sbin/preinit ubi.mtd=rootfs root=ubi0:rootfs rootfstype=ubifs rootflags=bulk_read,no_chk_data_crc rw console=ttyMTD,log console=tty0 snd-soc-rx51.hp_lim=42 snd-soc-tlv320aic3x.hp_dac_lim=6
Jan 11 00:31:46 <DocScrutinizer05> [2014-01-11 01:18:56] <DocScrutinizer05> >>>> console=ttyMTD,log
Jan 11 00:31:48 <DocScrutinizer05> [2014-01-11 01:19:21] <DocScrutinizer05> dunno about your particular kernel
Jan 11 00:31:49 <DocScrutinizer05> [2014-01-11 01:22:42] <DocScrutinizer05> you still listening?
Jan 11 00:31:51 <DocScrutinizer05> [2014-01-11 01:25:49] <DocScrutinizer05> :-S
**** ENDING LOGGING AT Sat Jan 11 02:59:58 2014