**** BEGIN LOGGING AT Sat May 26 03:00:02 2018 May 26 06:47:58 [07:50:35] * #lede-dev #openwrt-devel :Forwarding to another channel May 26 06:48:08 quel surprise :--D May 26 06:49:29 Just the sort of added confusion I needed to see while I'm trying to figure out why lede-snapshot sysupgrade looks borked. May 26 07:53:35 mangix: ping May 26 08:30:17 Hi all. A quick question (i hope). I'm attempting to run openwrt on a Pi 3 B+. I'm using the snapshot for target brcm2708/bcm2710. Writing to a new, good quality SD card using dd as per the instructions. However i can't get the Pi to boot, stuck at the 'rainbow screen'. I believe the 3 B+ is now supported as per a commit in April, so i don't know if i'm being dumb? Thanks :) May 26 08:58:01 elliotp: do you have a serial console? May 26 08:58:47 morning May 26 09:30:57 russell--: i don't, to be honest i didn't know it was possible. I will look into this May 26 09:37:35 I have 8 flashes of the green led, which from what i've just read, means SDRAM not recognised (You need newer bootcode.bin/start.elf firmware.) May 26 09:46:15 Success, i've copied the bootcode.bin & start.elf from the raspberrypi master branch onto the sd card, and it's booting. May 26 10:56:08 blogic: yeah, will send a patch when I figure out the last couple of niggles. May 26 10:56:33 blogic: need to fixup/understand teh board names vs sysupgrade. but nearly there. May 26 11:03:42 coolio May 26 11:21:26 Well, *that* was 2 hours I'll never get back. May 26 11:22:23 Built a Lede snapshot, it kept complaining 'The uploaded image file does not contain a supported format. Make sure that you choose the generic image format for your platform.' May 26 11:23:03 Weird part is I scp the image to the router, and run 'sysupgrade -T' and it's happy with it, and sysupgrace -c runs fine and it reboots to the new image. May 26 11:26:21 blogic: any ruling on the lednames though? board vs vendor? May 26 11:52:39 so with the channel rename, is LEDE/Openwrt going to start actually getting maintained soon? May 26 11:55:48 ie actually fixing buggy patches that break wifi like https://bugs.lede-project.org/index.php?do=details&task_id=1180 May 26 12:08:18 Hi luke-jr You are best posting to the mail list about that bug mate May 26 12:09:34 I do agree it needs looking at. All tho I am not a dev so don't know how it could be fixed. May 26 12:10:28 Anyone on here that has a gl-inet b1300 (or any other ipq4xxx device) that has a super degraded wifi performance? May 26 12:12:57 Tapper: the most obvious fix is to just remove the broken patch :p May 26 12:13:16 there's no mailing list in the topic :/ May 26 12:13:39 https://lists.infradead.org/mailman/listinfo/openwrt-devel May 26 12:13:46 May be but I cant say anything about that mate I don't know enuff about it May 26 12:14:11 thanks Lantis that was fast May 26 12:14:14 lol May 26 12:14:47 +10 xp May 26 12:24:27 karlp: whatever upstream does May 26 12:25:12 dedomraz: bad wifi on ipq4 means bad board-2.bin May 26 12:26:08 blogic: what do you mean? My firmware is bad, or the board? sorry :-) May 26 12:26:47 dedomraz: package/firmware/ipq-wifi/ May 26 12:27:04 that file is prolly not correct or something May 26 12:27:10 Because I have tried from different repos, different versions, their official one and the only ones I am having luck are the ones from around late January from the gl-inet repo. So, in my head, it seems that its a wifi driver issue, which got updated. Just trying to confirm :-) May 26 12:27:57 ok, could be May 26 12:28:00 no idea May 26 12:28:27 you just gave me a good idea! Thanks! :-) May 26 12:28:41 erm, ok May 26 12:28:42 I will see if I can use the new openwrt repo, but use the driver from the old one May 26 12:28:49 ah ok May 26 12:29:15 just to confirm - package/firmware/ipq-wifi/ - thats where the wifi driver is, correct? May 26 12:31:33 no May 26 12:31:40 that is the board specific firmware May 26 12:31:50 wifi driver is package/kernel/mac80211 May 26 12:33:37 10-4 May 26 12:42:13 Could it be due to power manager? Does power manager at all work under openwrt? Maybe that is limiting the power of the wifi.. May 26 13:18:21 guys May 26 13:18:59 I have just bought a WRT54GL. May 26 13:19:56 now i want to install openwrt. I have downloaded this: http://downloads.openwrt.org/backfire/10.03.1/brcm-2.4/openwrt-brcm-2.4-squashfs.trx May 26 13:20:08 why on earth would you buy such obsolete hardware? May 26 13:20:24 I guess it is save to flash this firmware May 26 13:20:29 Why not May 26 13:20:38 because technology improves with time May 26 13:20:55 Now May 26 13:21:13 what about the FirmwareFlash it? May 26 13:22:16 Because in the quick start manual it says the file has to end with ...factory.bin May 26 13:24:10 openwrt 10.03 is unsupported since many years and probably has many known security problems May 26 13:24:13 mine doesnt May 26 13:24:41 … May 26 13:25:23 can i also flash a newer version? May 26 13:29:30 it seems like i cant May 26 13:30:32 what about the Firmware? May 26 13:30:38 Flash it? May 26 13:30:48 Or is there a newer version? May 26 13:34:23 anyone know if the minstrel blues patches are in the snapshots for mac80211 by default? or am I going to have to compile? May 26 13:36:11 Can i break my router by flashing something? May 26 13:37:19 Yes, not that it'd be a great loss. May 26 13:40:30 lol May 26 13:41:05 why? May 26 13:41:16 Can i flash that file? May 26 13:41:26 WRT54GL is *15 years old* now. It's beyond obsolete. It's older than my oldest child! May 26 13:41:43 your child works fine, right? May 26 13:41:47 why shouldn't a gl work fine? :) May 26 13:42:23 xD May 26 13:42:43 wigyori: even much newer wifi routers aren't compatible with recent phones May 26 13:42:52 or laptops May 26 13:42:56 hahaha May 26 13:43:07 I already have one of those May 26 13:43:28 And it worked for all laptops and phones May 26 13:43:35 wigyori: Are we proposing child labour here? :) May 26 13:44:40 Monkeh: lol May 26 13:45:32 What if i flash the firmware now? May 26 13:46:00 I just searched the router in the table of hardware May 26 13:46:21 I just searched the router in the table of hardware. May 26 13:46:47 It should be save to flash that one right? May 26 13:47:13 If it breaks you'll have an excuse to get something from this decade May 26 13:47:37 You really shouldn't be running such old hardware and firmware. May 26 13:49:28 It will not break will it? May 26 13:50:48 no it wont May 26 13:50:55 Why should it? May 26 13:51:44 lol May 26 13:51:58 It says incorrect image file! May 26 13:52:04 What now? May 26 13:52:55 what should i do now? May 26 13:55:18 just try openwrt-brcm-2.4-squashfs.bin? May 26 13:55:36 Just a different ending? May 26 13:58:38 What is a trx file? May 26 13:58:47 Why doesnt it work? May 26 14:04:09 hrm, https://www.kernel.org/doc/Documentation/devicetree/bindings/leds/common.txt actually doesn't make any guidelines on label names, other than ensuring they are unique. May 26 14:06:40 gregor2: did you tried this one http://archive.openwrt.org/backfire/10.03.1/brcm-2.4/openwrt-wrt54g-squashfs.bin ? May 26 14:08:16 is it actually on the stock firmware ? May 26 14:11:03 It is on the stock firmware May 26 14:11:20 and i did not try the one from above May 26 14:12:16 Can i break it if i flash the wrong one? May 26 14:16:03 there are some verifications before the flash but yes it's possible (but I don't think it'll happen, 'cause if it's really a wrt54gl 1.0 or 1.1, it's the good/latest one) May 26 14:18:19 but don't even try to install additionnal softwares on this thing, it only has 4MiB of flash and 95% will be used by the firmware... May 26 14:21:30 the wrt54g file can be used? May 26 14:21:38 (otherwise, the 8.09.2, which is even older, can leave some extra KiB free on the flash but it's _really_ old) May 26 14:21:59 its running now May 26 14:22:38 what is running ? owrt 10.03.1 ? May 26 14:23:05 yes May 26 14:23:22 no it just upgraded May 26 14:23:38 and it said it would be successful. May 26 14:25:01 but looks good May 26 14:25:01 gregor2: you bought OLD device, you can install OLD and non-secure firmware, or NEW and unstable (because of amount of flash/RAM memory) firmware May 26 14:25:39 gregor2: 10.03.1 and brcm-2.4 is old and hacky May 26 14:25:46 it doesn't use nice upsteam code May 26 14:25:48 etc. May 26 14:26:00 gregor2: that's why you should rather buy a more recent device May 26 14:29:47 https://www.aliexpress.com/item/802-11N-300Mbps-MT7620A-openwrt-wireless-router-300Mbps-WiFi-Router/32361575074.html May 26 14:31:48 MT7620A @580MHz with 16MiB of flash and 128MiB of ram for 25$ May 26 14:32:26 it works isnt that cool? May 26 14:32:34 and the WRT54GL == BCRM @200MHz with 4MiB of flash and 16MiB of ram... May 26 14:32:37 with a thousand security holes... May 26 14:32:44 Noooo May 26 14:32:45 congratulations! May 26 14:32:56 There are no security holes May 26 14:33:01 haha, sure :D May 26 14:33:09 ^^ May 26 14:33:19 Pull t'other one May 26 14:33:24 i'm starting to think you're just a troll... heading out, nice weather. May 26 14:33:39 Back in those times there were none to! May 26 14:34:14 Or did the hackers got better? May 26 14:34:40 How should the security holes look like? May 26 14:35:21 Well they have big neon signs over them so everyone can see them easily May 26 14:35:41 And... May 26 14:35:54 What could a hacker do here? May 26 14:36:16 Shut down the whole Wifi? May 26 14:36:31 well, he couldn't sniff your traffic as there is not enough space to even install a tcpdump May 26 14:36:31 Well he could do that anyway May 26 14:36:43 seee? May 26 14:37:28 Can he breach into the root account of the router? May 26 14:37:35 Probably. May 26 14:37:56 I could just deactivate ssh May 26 14:38:01 and httpd May 26 14:38:14 So noone can do anything! May 26 14:38:26 Just turn it off and put it in the bin, then it's safe. May 26 14:39:08 But how could that happen? May 26 14:39:28 How would anybody do that May 26 14:39:41 hacking into the root account? May 26 14:39:46 how? May 26 14:41:44 First he would have to get into wifi. May 26 14:42:02 But lets say he has accomplished that. May 26 14:42:08 What next? May 26 14:44:34 [14:36:31] well, he couldn't sniff your traffic as there is not enough space to even install a tcpdump <-- lol May 26 14:44:41 wigyori: mount a tmpfs :P May 26 15:14:05 dissent1: pong May 26 16:11:06 trolls be trollin I see. May 26 16:19:22 hello May 26 16:23:09 Hi I installed snapshot on r7800 and it bricks. Is it possible to do a binary search of snapshots to find out when it broke (and so I can pick up a most recent working snapshot?) May 26 16:23:34 ie, are snapshots kept somewhere? May 26 16:25:21 Martin-sjc: I think it would be impractical to store snapshots. What I did was git bisect with a build each time May 26 16:26:35 OK I figured :(. May 26 16:26:38 mangix: I don’t experience that pci issue on mt7621 mentioned on the forum May 26 16:27:11 mangix: and it seems that hw crypto works but it’s not better than software May 26 16:27:32 what pci issue? May 26 16:28:08 he crypto being not much faster sounds fishy May 26 16:32:10 yeah that doesn't make sense based on benchmarks i've seen May 26 16:32:11 mangix: mt76 on the forum, remember? May 26 16:32:32 the timing out issue? May 26 16:32:52 mangix: we’ve got to wait for openssl with af_alg yo see if it makes any better May 26 16:33:03 Openssl 1.1 May 26 16:33:20 there are patches floating for that yes May 26 16:34:22 mangix: to enable hw crypto in openssl now you need to enable crypto support, hw crypto and crypto digests in openssl and cryptodev in kernel modules, that’s all May 26 16:35:30 then you can test it with openssl speed -evp aes-256-cbc -elapsed May 26 16:35:58 meh. mvebu is plenty fast anyways May 26 16:36:23 i would love to see ramips support May 26 16:47:07 i took a brief look at the pci changes (nit on a computer atm). i don't see any problems. dmesg output changed as a bunch of printf's got removed. May 26 16:47:14 *printk May 26 16:48:15 on the bright side, the kernel's staging driver is now 1:1 with OpenWrt's May 26 16:53:25 what happens if you accidentally misconfigure openwrt? May 26 16:53:31 Are you doomed? May 26 16:54:08 Can you never flash the firmware again because you have no access to the router anymore? May 26 16:54:35 can you lock yourself out of the router forever? May 26 16:57:06 luke-jr: tmpfs with 16mb ram? :D May 26 16:58:26 I'm pretty sure I once upon a time successfully tcpdump'd on a WRT54GL, somehow May 26 16:59:35 what about the reset button? May 26 16:59:42 What does it? May 26 17:00:03 nothing? May 26 17:00:41 gregor2: depend on the platform, some of them have serial port where you can get into a boot loader and download a new image to it. May 26 17:01:25 and reset does nothing? May 26 17:01:43 gregor2: again, depends on platform. Some support factory defaulting via reset button, some don't. May 26 17:02:25 SwedeMike: he's the guy trying to setup on a WRT54GL <.< May 26 17:02:52 ok, I didn't have that context. I don't have any specific knowledge about that platform and what it does or doesn't do. May 26 17:04:08 obviosly the reset button does not work May 26 17:21:14 * ldir laughs - a wrt54gl in 2018?!!!! May 26 17:26:24 i am back in! May 26 17:26:28 bam! May 26 17:26:42 what do you say now?! May 26 17:26:50 Not so tough now huh? May 26 17:40:34 it worked May 26 17:40:55 All i had to do was pressing a Butten during boot May 26 17:41:06 then connect via telnet and May 26 17:41:14 firstboot May 26 17:41:19 finished May 26 17:43:19 ldir: They're not even good paperweights May 26 18:00:42 hi guys, just a short question, I already made my make menuconfig and now I just want to build an Image for TP-Link Archer C5 so I dont have to build all images and might save time, is it possible and how could this be achieved? Thanks in advance. May 26 18:41:24 privatereese: pretty sure there's something in menuconfig to choose that May 26 19:16:05 heh, just bought this zbt we1326, and finally xiaomi comes out with the rev 4 May 26 19:16:52 bit low on ports though, and they dropped usb so probably for the best May 26 19:35:40 ldir: wrt'ing with style May 26 19:37:03 i might order one of ebay just for the giggles May 26 19:38:43 can anyone here recommend a cheap and decent ata adapter? May 26 19:47:13 [how] can i patch and build odhcpd with the imagebuilder and/or sdk? May 26 19:57:43 mangix: May be same shit as with r7800 that is initially misconfigured unlike other boards on same chipset May 26 20:12:17 karlp: the xiaomi mi 4 'looks' better, but in terms of functionality the zbt we1326 wins easily (sad that xiaomi has removed the usb port - and it would be a lot more useful with 4+1 ethernet ports instead of just 3) May 26 20:26:26 am i supposed to be able to use "make package/$package/compile" to build packages that are in the sdk's base feed? May 26 21:45:51 howdy guys i have a wuestion about firstboot, how long does in normally take to complete the script? May 26 21:47:06 few seconds usually, is yours broken? May 26 21:48:12 yea May 26 21:48:22 i can't remember what i did to it May 26 21:48:29 so im just resetting to default May 26 21:48:35 that depends a lot, first if you are in failsafe mode == jffs2 is not mounted and gets actually formatted or with the overlay mounted, so you can just delete the files (faster) - and it also depends a lot on the flash chip, its size and its sector size. e.g. a 16 MB spi-nor flash with 4 KB sector size may take over five minutes, a NAND chip a couple of seconds May 26 21:49:02 trusay about nor May 26 21:49:03 should i remove overlay or overlay/upper? May 26 21:49:10 nand ftw May 26 21:52:15 like how if overlay/upper exists is that the one I should delete? or is it all routers contain overlay/upper and it depends on the router when you should just delete overlay or overlay/upper/*? May 26 21:52:19 MrFantastik: you can look in /etc/rc.button/reset May 26 21:52:41 jffs2reset -y && reboot & May 26 21:53:29 what device are you working with? May 26 21:53:43 a netgear router May 26 21:53:53 i can find the exact one in a sec May 26 21:53:54 there are about 500 of those May 26 21:54:13 netgear r6100 May 26 21:55:41 the "factory" reset process is in that /etc/rc.button/reset script May 26 21:56:06 at least on the devices i have (not that one) May 26 21:56:24 so jffs2reset -y && reboot & May 26 21:56:42 that works on my devices May 26 21:56:53 okay ill give it a try May 26 22:06:39 i think it worked May 26 22:06:45 but i can't access the router now May 26 22:06:59 should be 192.168.1.1 right? May 26 22:10:30 how often does the builtin ntp client sync time? May 26 22:10:49 MrFantastik: gibe it a few minutes May 26 22:10:57 give May 26 22:11:37 okay May 26 22:39:38 hmm May 26 22:39:44 so May 26 22:41:01 about the mvebu did someone measure the drift with ntpclient to get new adjtimex frequency value? May 26 22:41:13 *rtc May 26 22:44:16 http://doolittle.icarus.com/ntpclient/HOWTO May 26 23:19:14 hmm May 26 23:21:26 i didnt follow the entire last conversation about the mvebu rtc issue. but one thing i noticed. with disabled rtc, adjtimex gives a frequency of 32768000 (or something like that). With enabled rtc adjtimex gives a frequency of 1570702 May 26 23:21:34 which seems quite wrong May 26 23:49:54 also return value: 5 (clock not synchronized) May 27 00:36:40 Morning. Is anyone able to quickly point me towards code that gets run by setting CONFIG_ARM_ATAG_DTB_COMPAT_CMDLINE_FROM_BOOTLOADER=y please?? May 27 00:38:40 I'm afraid all I can do is run screaming from such a terrifyingly long option. May 27 00:40:59 lol, that is kinda ridiculous May 27 00:41:19 also ive come to feel sorry for any country that the british visit :D May 27 00:42:19 Do you have a flaaaaag? May 27 00:43:48 sat in edmonton May 27 00:43:56 surrounded by british May 27 00:44:02 coz the football is on May 27 00:44:25 4200 milee and i still cant escape my fellow countrymen May 27 00:44:39 miles May 27 00:45:34 idiots wherever we go apparently, no wonder they dont want us :D May 27 00:45:37 Lantis: Brief grep, check arch/arm/boot/compressed/atags_to_fdt.c May 27 00:45:51 jwh: I resemble that remark May 27 00:47:23 you resemble a typical brit? :D May 27 00:47:50 No, but I'm pretty much the same idiot wherever I go May 27 00:47:58 see what i did there May 27 00:48:02 heh May 27 00:49:39 bunch of essex wideboyshere being noisy just like home May 27 00:50:28 All over a ball May 27 00:52:34 jwh: http://www.bbc.co.uk/news/uk-england-suffolk-44223220 May 27 00:55:33 um, what May 27 00:55:51 also screw the bbc, directimg me to bbc.com May 27 00:55:56 directing* May 27 00:56:11 more adverts than the daily mail May 27 00:56:21 heh May 27 00:56:33 I just figured you'd appreciate a typical brit showing his mental age.. :) May 27 00:56:36 lol May 27 00:57:00 well May 27 00:57:02 My cat isn't that dumb May 27 00:57:16 these guys live here so a permit must be easy May 27 00:57:29 they dont appear very clever May 27 00:58:18 ive been up in the country with the natives all week, no british there its been lovely May 27 00:58:32 i can now curse people in Cree May 27 01:03:44 Monkeh: what did you grep? It does not appear to be in that file (but i agree it is the right place) May 27 01:07:19 Lantis: the linux kernel commit 88bc4a36a985ac14bd60fdc4fee00c68874c52f1 and d0f34a11ddab9b456e4caf9fc48d8d7e832e0e50 has some relevant log messages May 27 01:07:55 Lantis: See arch/arm/Kconfig May 27 01:09:39 just got asked for id too for smokes, pretty rude May 27 01:12:41 Monkeh: it is a bit obscure what depends on that setting; pretty clear how to set it, but not what happens if it is/isn't set. May 27 01:14:31 russell--: I didn't say I'd read the code May 27 01:14:31 :) May 27 01:15:58 it looks like ATAGs are a way of passing parameters to the kernel instead of via device trees or something May 27 01:16:39 I'm more interested in the interaction between CONFIG_ARM_ATAG_DTB_COMPAT_CMDLINE_FROM_BOOTLOADER and CONFIG_ARM_ATAG_DTB_COMPAT_CMDLINE_MANGLE May 27 01:16:56 Which works fine on WRT32X, but breaks all other mvebu devices i have tested May 27 01:18:01 did you check what happens if you enable append-rootblock for the broken devices as well? (not really a solution, but interesting to check) May 27 01:18:21 https://stackoverflow.com/questions/21014920/arm-linux-atags-vs-device-tree May 27 01:18:31 I have not, but i don't see that as a viable solution anyway. May 27 01:22:42 i didn't really want to pull the case apart on another linksys device but here we go again May 27 01:23:10 it's a bit sad that there doesn't really appear to be a way to override CONFIG_ARM_ATAG_DTB_COMPAT_CMDLINE_EXTEND from dts May 27 01:24:23 hey folks! May 27 01:27:22 ugh. I really need to rig up something so I can connect network devices and hax without fscking up my actual network connection lol May 27 01:33:24 classic... May 27 01:33:26 Kernel command line: May 27 01:33:26 May 27 01:33:42 when you use both kconfigs together, you get a blank cmdline May 27 01:33:50 heh. May 27 01:41:24 yeah there's some sort of wierd interaction with base config files and subtarget config files May 27 01:42:20 any clever ideas? I assume it isn't possible to set a KConfig for a single device? May 27 01:43:15 Lantis: offhand no, but in ar71xx targets there's a thing to patch the command line; its specified in the makefile definition for the devices there. don't know what you're working on, however. May 27 01:43:31 a dualboot mvebu device, so hardcoding the cmdline is out May 27 01:43:42 that would require a dedicated subtarget, which is adds considerable load to the buildds May 27 01:55:03 man, ryzen 1700, really speeds up compiles :) May 27 02:01:50 if it doesnt crash :D May 27 02:02:22 jwh: honestly after disabling c6 I've had 16 days uptime, prior to a manual reboot for a kernel and bios update :) May 27 02:02:46 pretty good going May 27 02:03:15 ryzen is a nightmare atm May 27 02:03:52 sticking with intel for now heh May 27 02:04:09 have power and cooling limits to consider also May 27 02:04:26 I'm seriously considering setting aside whole paychecks to eventually get a talos 2 workstation :P May 27 02:10:16 heh May 27 02:11:27 neber trusted amd, they did good woth barton May 27 02:11:30 with May 27 02:11:40 but thdy have been garbage since May 27 02:11:43 they May 27 02:12:21 those new ryzen+ chips look nice. I maybe should have held out for one of those, but the fx6300 build I had wasn't cutting it anymore :) May 27 02:12:50 mayne for a desktop i would May 27 02:13:06 but they are pretty terrible for server duties May 27 02:13:15 mayne May 27 02:13:27 evaluated it, dont like it May 27 02:13:32 eh... I wouldn't say that, epyc and threadripper look pretty awesome tho. May 27 02:13:40 but hey, opinions :) May 27 02:13:42 high power, high heat, not that fast May 27 02:14:00 not stable under linux yet either May 27 02:14:14 still mamy problems with the chipsey May 27 02:14:15 jwh: at least AMD solders their chips May 27 02:14:17 chipset May 27 02:14:26 sorry, IHS May 27 02:14:37 how do you mean? May 27 02:14:49 google delid intel May 27 02:14:51 stable for me :). granted, I like to do kernel stuff a lot so reboots are fairly frequent for me, so I probably won't get a very accurate uptime reading for a while May 27 02:15:15 yeah this is production, cant risk it May 27 02:15:19 gets you a 14 degree improvement in the typical case May 27 02:15:26 because intel decides to be cheap May 27 02:15:36 jwh: fair. May 27 02:15:51 mangix: oh, that May 27 02:16:21 honestly May 27 02:16:22 AMD chips also perform better under LN2 May 27 02:16:31 if the per core was better on amd May 27 02:16:32 no stupid DRAM issues May 27 02:16:51 i would jyst engineer around heat and power May 27 02:16:54 just May 27 02:17:21 but many applications need high perf and may not be multithreaded May 27 02:17:53 i get 1kw per 42u rack May 27 02:18:01 so do the mathsheh May 27 02:18:35 and cooling ability is based on that May 27 02:19:03 its pretty tighy May 27 02:19:10 ugh, tight May 27 02:19:46 few cisco 7200s, some servers, x86 lede routers May 27 02:19:54 already at 900w May 27 02:21:07 iys an old building too, wasnt designed to cool so much equipment May 27 02:21:10 its May 27 02:21:41 wben it was designed p1s were a thing, no heat and little power really May 27 02:25:08 mangix: Who cares how they perform wildly outside their specified operating conditions? May 27 02:25:33 Monkeh: crazy people, that's who :P May 27 02:32:04 lol May 27 02:32:33 tbf i used to be an amd man May 27 02:32:56 i think i've spotted the issue, if someone could sanity check me please May 27 02:32:57 coz they eere cheap and intel sucked May 27 02:33:15 but now, intel every time May 27 02:33:23 Monkeh: delidding allows lower fan speed in addition to longer boosting May 27 02:33:24 Monkeh: I mean, I get it, its cool (heh, bad pun), but I'd personally never do that. hell I struggle to convince myself to try watercooling. May 27 02:33:37 https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/ipq806x/patches-4.14/0067-generic-Mangle-bootloader-s-kernel-arguments.patch;hb=5857088c5eb3a5a2409e3c57dbfa2487e08bbf4a#l60 May 27 02:33:41 Also everyone bragging about the solder hasn't bothered addressing the fact that the Ryzen die is 50% larger. The supposed improvement in performance seems rather less significant when you factor that in.. May 27 02:34:10 am i crazy, or if you have no "append-rootblock" section in the dts, does dest = "" ?? May 27 02:34:12 i absolutely would not delid a cpu May 27 02:34:28 they are designed to specs May 27 02:34:36 i know what they are May 27 02:34:39 jwh: Those specs can be faulty May 27 02:34:41 ;) May 27 02:34:49 But the whole thermal paste issue is.. hugely overblown May 27 02:34:58 in which case the vendor replaces them May 27 02:35:14 Have you had a replacement CPU for spectre and meltdown? May 27 02:35:16 but ive neber had an intel cpu failure May 27 02:35:19 nope May 27 02:35:23 never* May 27 02:35:33 Classic case of faulty specs! May 27 02:35:47 closed systems, deemed the effort to replace not worth it May 27 02:36:19 more chance of an sql injection than spectte May 27 02:36:26 spectre May 27 02:36:33 jwh: and i would, given the crap intel calls thermal paste May 27 02:36:43 well May 27 02:36:44 mangix: Please address the area issue. May 27 02:36:49 retaik is a disgrace May 27 02:36:57 i will give you that May 27 02:37:08 but oems test them May 27 02:37:24 Lantis: only two ipq806x use append-rootblock, the other have no negative impact May 27 02:37:34 they do the work and insure their designs May 27 02:37:45 Lots and lots of harping on about how the paste isn't good enough, nobody addressing how lacklustre the soldered performance is considering the 50% greater area involved.. May 27 02:37:47 so i know they will work May 27 02:38:11 ig May 27 02:38:12 Lantis: the patch/ config symbol are obviously active for the whole target May 27 02:38:26 if a cpu faikls, i get a new one May 27 02:38:32 fails May 27 02:38:32 Monkeh: to be fair, a delidded ryzen CPU (1700x i think) was tested to have a 4 degree reduction in temperature May 27 02:38:37 pkgadd: do the ones that don't use it provide "bootargs"? May 27 02:38:44 ok May 27 02:38:48 probably mostly due to removal of silicone adhesive May 27 02:38:52 im going to counter this May 27 02:39:03 .. the adhesive isn't going to have any impact May 27 02:39:12 modern cpus are designed for 90 degrees May 27 02:39:18 And frankly, 4C is margin of error with most of these test setups. May 27 02:39:19 Monkeh: yes it is May 27 02:39:20 Lantis: some do May 27 02:39:20 sustained May 27 02:39:39 the adhesive is an insulator May 27 02:39:48 Lantis: e.g. all the netgear ones May 27 02:39:49 ... but it's not between the heatspreader and the die May 27 02:40:15 we worked out how to make cpus cope with high temps May 27 02:40:28 coz we couldnt make them more efficient May 27 02:40:43 jwh: Actually we made them scarily more efficient. May 27 02:40:47 But they got smaller. May 27 02:40:49 or rather all netgear ones, except for the r7800 May 27 02:40:50 we dont May 27 02:40:52 it contacts the IHS. It's been tested. If you reapply silicone adhesive after wiping it off on intel CPUs, delid performance can be worse May 27 02:41:03 they are worse now than 20 years ago May 27 02:41:11 This means higher thermal resistance which means higher core temperatures. May 27 02:41:16 more hear, higher power May 27 02:41:21 heat May 27 02:41:31 because of new fab tech May 27 02:41:39 Yes, but the work done per watt is insanely higher. May 27 02:41:46 yesx it is May 27 02:41:53 ie. they use power more efficiently. May 27 02:42:05 but temps and waste arent decreasing May 27 02:42:12 They did May 27 02:42:15 they are increasing May 27 02:42:26 Hell, for the average computer they're down quite a lot over the P4 era May 27 02:42:29 pkgadd: i'm just trying to follow the logic here. If you have no "bootargs", cmdline is "". If you then don't have a valid "append-rootblock" clause, at no point does anything get memcpy'd to dest May 27 02:42:29 hence, they make them withstand higher temps May 27 02:42:40 peer watt, yes May 27 02:42:48 and per hz May 27 02:42:53 mangix: A little adhesive holding the edges of the IHS down to the carrier is not going to make any notable difference. May 27 02:43:00 but we havebt really made them better May 27 02:43:16 lol May 27 02:43:18 current physics doesnt allow us too May 27 02:43:19 it's not a little May 27 02:43:51 And seeing as we were talking about Ryzen, where the dies are soldered to the IHS.. May 27 02:44:29 we need more work pdr cyvle May 27 02:44:32 per May 27 02:44:39 not more cycles May 27 02:44:48 I mean sure, it's going to seal a little air under the IHS, but it's not nearly enough to care about May 27 02:45:17 cavium hot the work per cycle down btw May 27 02:45:20 got May 27 02:45:32 but general cpis csnnot do that May 27 02:45:36 cpus May 27 02:47:19 Lantis: I can only speak for the nbg6817 (and didn't actually introduce the ATAG patch to the target, that was already introduced for the ea8500 before), but as far as I understand it, you should get the bootloader cmdline if you don't override it via bootargs or append-rootblock May 27 02:48:02 mangix: https://img.purch.com/o/aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9NL0MvNjYxNzY0L29yaWdpbmFsL0ltYWdlMTAucG5n as an example - 1800X vs 7700K, similar power consumption, 6.6C difference - and the Ryzen has 50% (at least) greater contact area between die and IHS - where's this insane solder advantage again? May 27 02:48:33 8.6C. Pff, 2C, tired eyes. May 27 02:51:59 if yiu care about 6c youve fucked up yoir cooling design :D May 27 02:52:03 you May 27 02:52:08 also your May 27 02:52:44 jwh: It's like the people who argue about which brand of paste to use May 27 02:53:00 shluld account for 30 either way May 27 02:53:13 esoecially in a shared facility May 27 02:53:19 especially May 27 02:53:21 Oh look, this one is 0.3C cooler in this single sample test! Clearly it's superior! May 27 02:53:24 heh May 27 02:53:50 Seriously, the range between 'toothpaste' 'cheap and nasty paste' and '$40 tube' is about 8C May 27 02:53:52 over 7 days my kit has reported +-20 deg c May 27 02:54:07 And the range between '$4' and '$40' is about 2C May 27 02:54:19 I've been using a cheap noname thermal paste bought long before CPUs needed them so far - am I a bad person? ;) May 27 02:54:21 obviously ill be raising this with the facility, but it isnt much May 27 02:54:34 And aaaaaalllllllll these tests are a single run on a single sample May 27 02:54:39 pkgadd: artic silver or death May 27 02:54:51 jwh: Overpriced underperformer. ;) May 27 02:54:59 haha May 27 02:55:10 AS5 was awesome when it came out May 27 02:55:14 i havent built a maxhine since i was a kid May 27 02:55:20 18 years ago May 27 02:55:21 .. it hasn't got any cheaper May 27 02:55:31 Monkeh: alternate suggestion? its what I use but I'm willing to be convinced to use another brand :) May 27 02:55:33 jwh: you you mistake mine with white toothpaste... I wouldn't recommend mixing them up though ;) May 27 02:55:39 Pk2 May 27 02:55:42 ugh May 27 02:55:45 pkgadd: ha May 27 02:55:53 hopefully toothpaste May 27 02:56:31 hanetzer: Currently I'm using a giant tube of Arctic MX-2. It's fine, it's really cheap, it was to hand.. May 27 02:58:00 hanetzer: https://img.purch.com/image006-png/w/711/aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9YL0YvNjk4NDUxL29yaWdpbmFsL2ltYWdlMDA2LnBuZw== for example May 27 02:58:14 You can see how massively important your choice of normal price pastes is. May 27 02:58:43 .. the liquid metal stuff is pretty impressive but not impressive enough to care. May 27 02:58:55 Please hold the gallium. May 27 02:58:58 tbh May 27 02:59:06 if you buy oem stuff May 27 02:59:15 its already been tested May 27 02:59:26 dont need to csre May 27 02:59:33 care May 27 02:59:38 Prebuilt stuff will use one of the big names I can't honestly remember right now May 27 02:59:44 dell et sl are so cheap now May 27 02:59:48 why bother May 27 02:59:50 It pretty much sits smack in the middle of that May 27 03:00:00 And they pay about a hundredth as much for it. **** ENDING LOGGING AT Sun May 27 03:00:02 2018