00:12 Sophira: So I tried 4.6.5. It didn't go well; when it initialised the modesetting driver it seemed to crash. I say "seemed to" because the screen was blank and eventually went into its power-saving mode; CTRL-ALT-DEL didn't start the process of a reboot, and there's no evidence of the boot at all in the syslogs, which suggests that it panicked before my system logger could load.
00:13 Sophira: That also means that I don't have any error messages to share; sorry. :(
00:15 Sophira: (This is before I've even tried to do anything with the nouveau git repo; this is just the in-kernel module.)
07:11 glennk: mwk, https://www.flickr.com/photos/130561288@N04/27671715841/
07:16 mwk: glisse: ah, a thing of beauty :)
07:16 mwk: 500nm, good times
09:07 karolherbst_work: Sophira: check /sys/fs/pstore
12:06 Sophira: karolherbst_work: Nothing in there, even when I boot in 4.6.5 with init=/bin/bash (since I compiled nouveau as a module) and mount /sys manually.
12:20 karolherbst_work: Sophira: ohh right, you need to configure pstore first
12:20 karolherbst_work: there are some kernel options for that
12:32 karolherbst_work: Sophira: enable those: CONFIG_PSTORE=y, EFI_VARS_PSTORE=y, EFI_VARS_PSTORE_DEFAULT_DISABLE=n, PSTORE_RAM=y
12:41 karolherbst_work: then boot with the 4.6 having that enabled and let it crash. Then under the working kernel mount it like that: mount -t pstore none /sys/fs/pstore
12:44 karolherbst_work: mhh or maybe you just have to wait, because userspace firmware is loaded
12:44 karolherbst_work: ohhhhhh right
12:45 karolherbst_work: Sophira: for maxwell2 gpus, you need the filesystem ready when nouveau gets loaded/initiliazed
12:45 karolherbst_work: so you can't really use nouveau builtin and need to load it as a module
12:46 karolherbst_work: and you need to enable ~amd64 for linux-firmware
12:46 karolherbst_work: and install it
14:08 karolherbst_work: uhhh, MS golden key for secure boot got leaked :O
14:09 karolherbst_work: are there any MS ARM thingies out there with nvidia gpus? :D
14:13 Tom^: karolherbst_work: what a surprise, so secureboot is now not so secure. xD
14:18 karolherbst_work: it is good for bringin linux to those surface tablets :O
14:18 karolherbst_work: I bet we will see some patches in the future in the kernel: surface arm support .D
14:18 karolherbst_work: :D
14:19 karolherbst_work: ohh, they usre tegras indeed
14:19 karolherbst_work: mupuf: you need a surface now :p
14:20 karolherbst_work: tegra 3 and tegra 4
14:20 karolherbst_work: would be nice to add support for those within nouveau then
14:22 Calinou: Tegra X1!
14:22 Calinou: <karolherbst_work> uhhh, MS golden key for secure boot got leaked :O
14:22 Calinou: I know the guy who did it :)
14:22 Calinou: from #fsf
14:26 karolherbst_work: :D
14:27 karolherbst_work: Calinou: why a X1?
14:28 karolherbst_work: I somehow think, that Tegra3 and Tegra4 are too different and we would need a new driver anyway
14:28 karolherbst_work: VLIW based are those
14:28 Calinou: Tegra X1 is fast, has OpenGL ES 3.x
14:29 Calinou: Vulkan
14:29 karolherbst_work: Calinou: yeah, but nouveau supports those, doesn't it?
14:29 karolherbst_work: no fun this way
19:37 NanoSector: karolherbst: linux 4.7 just landed in Arch, I'll give your branch another shot
19:39 Calinou: cool, new kernel
19:39 Calinou: will wait a bit I guess
19:39 NanoSector: ohey, you here too
19:49 NanoSector: karolherbst: your master_4.7 branch doesn't happen to include the kepler stable reclocking branch?
19:51 imirkin: pretty sure that one's just straight up 4.7
19:53 NanoSector: yeah thought so
20:01 NanoSector: 0a pstate is working now... it seems
20:01 NanoSector: 0f too
20:02 NanoSector: it freezes when i boost
20:06 NanoSector: so at least reclocking seems stable for me now as well... :)
20:20 NanoSector: hmm for some reason glxgears runs fine with reclocking but anything else using opengl hangs my laptop, so no improvement after all
20:22 loa: maybe applications use wrong opengl for some reason.
20:23 NanoSector: yeah i've been trying to pinpoint this issue for a while now with karolherbst, as I happen to be the only person he's come accross to have igf
20:23 NanoSector: *it
20:23 loa: in gentoo there is eselect opengl for it.
20:24 NanoSector: what's that do?
20:24 loa: nvidia proprietary drivers don't use mesa, they use their own opengl implemenation.
20:24 NanoSector: also, openGL apps run fine, but if I then try reclocking it'll run for 3 more seconds and then hang
20:24 NanoSector: oh yeah I know that, but I'm using nouveau which uses mesa libgl
20:24 NanoSector: and i have mesa libgl
20:25 RSpliet: NanoSector: that's a sign that nouveau selects the wrong voltage... karolherbst, go and fix that :-P
20:25 NanoSector: oh hey :P
20:25 NanoSector: well he thought it was voltage related
20:25 NanoSector: RSpliet: anything I can try herel ocally?
20:25 loa: kepler stable reclocking is not in master_4.7
20:25 NanoSector: I'm using his stable reclocking branch for 4.7
20:25 loa: it is in another branch.
20:25 loa: ah ok.
20:26 NanoSector: I have the source in that branch here locally, if there's something I can patch to try again that's no issue
20:26 NanoSector: or a tool i can run
20:26 NanoSector: or anything :P
20:26 RSpliet: NanoSector: just double-checking, what does git log --pretty=oneline -5 say?
20:26 NanoSector: one sec
20:27 NanoSector: oh i just downloaded the zip archive so it's not a git repo
20:27 NanoSector: downloaded it today
20:28 NanoSector: got it here: https://github.com/karolherbst/nouveau/tree/stable_reclocking_kepler_v5
20:28 RSpliet: yeah
20:30 RSpliet: sorry, I never know how to turn that into a kernel module myself... if you're sure that you are using the right module (on Fedora that includes rebuilding your initramfs using dracut and stuff like that)... I think it's supposed to make a sysfs entry that lets you select a slightly higher voltage
20:30 NanoSector: i did cd drm; make; sudo depmod -A and rebooted
20:30 NanoSector: oh and copied it into /lib/modules/...
20:31 RSpliet: what distro is that?
20:31 NanoSector: Arch linux, sorry
20:31 RSpliet: ah okay, don't know anything about that :-P
20:31 karolherbst: RSpliet: right
20:31 NanoSector: I just did what karolherbst told me to do :P so I bet it should work
20:31 karolherbst: NanoSector: can you switch to nvidia again?
20:32 NanoSector: karolherbst: no problem, hang on a minute
20:32 RSpliet: karolherbst: I'll leave this up to you ;-)
20:32 karolherbst: :D
20:32 NanoSector: with or without bumblebee, with I presume?
20:32 karolherbst: if the voltage is wrong, then there should be something really strange going on
20:32 karolherbst: or I messed up big times
20:33 NanoSector: karolherbst: switched to nvidia + bumblebee, rebooting now
20:33 NanoSector: and we're back
20:34 karolherbst: switch to the awesome_tool branch
20:34 karolherbst: and run make in the top directory
20:34 NanoSector: if qupzilla wants to launch...
20:34 karolherbst: there is also the command line :p
20:34 karolherbst: ain't that hard
20:35 NanoSector: i'm too noob for that :P
20:35 NanoSector: it's make-ing
20:36 NanoSector: fatal error in drm/nouveau/uapi/drm/nouveau_drm.h: drm/drm.h: file or folder doesn't exist
20:36 karolherbst: remove that include
20:36 NanoSector: aight
20:37 NanoSector: off we go again
20:37 NanoSector: oh, done
20:37 karolherbst: optirun sudo LD_LIBRARY_PATH=lib bin/nv_cmp_volt
20:37 karolherbst: it should print stuff
20:38 NanoSector: failed to create device, -12
20:38 karolherbst: mhhhh righto, stupid mem protection stuff going on
20:38 NanoSector: can it be disabled? :D
20:38 karolherbst: yes
20:38 karolherbst: I forgot how
20:39 karolherbst: it is in the channel somehow though
20:41 karolherbst: mupuf might now how :/
20:41 karolherbst: no luck finding it currently
20:41 NanoSector: what's it called?
20:41 urjaman: iomem=relaxed or something something?
20:41 RSpliet: karolherbst: is that the /dev/mem protection?
20:41 karolherbst: yeah
20:42 karolherbst: urjaman: you are right
20:42 NanoSector: add that to my kernel command line?
20:42 karolherbst: yes
20:42 NanoSector: will do, hang on
20:42 karolherbst: RSpliet: we need to add it to the nouveau wiki somewhere or to the envytools readme
20:42 NanoSector: booting...
20:43 urjaman: good that i wrote it right from memory :P
20:43 karolherbst: :D
20:43 NanoSector: ok different stuff now
20:43 NanoSector: nouveau 0000:01:00:0: bios: unable to locate usable image
20:43 karolherbst: :O
20:44 NanoSector: nouveau 0000:01:00:0: bios ctor failed, -22
20:44 karolherbst: what gpu do you have again?
20:44 NanoSector: GeForce GT750M
20:44 NanoSector: 4GB DDR3 variant
20:44 karolherbst: mhhh
20:44 karolherbst: how did you start the thing?
20:44 NanoSector: the command?
20:44 karolherbst: yes
20:45 NanoSector: optirun sudo LD_LIBRARY_PATH=lib bin/nv_cmp_volt
20:45 karolherbst: mhh, odd
20:45 NanoSector: it did threw errors about libdlfaker.so and libvglfaker.so not being loaded because it couldn't load them
20:45 karolherbst: could you grep a dmesg from a boot where nouveau was loaded?
20:45 karolherbst: NanoSector: that's optirun stuff, doesn't matter
20:45 NanoSector: like, remove nvidia, load nouveau and dump dmesg to a file?
20:46 karolherbst: nope
20:46 karolherbst: just check your system logs
20:46 NanoSector: might have a journalctl log
20:46 karolherbst: you can do stuff like this: journalctl --boot -2
20:47 karolherbst: mhh, tool works for me, so something is oddish on your end
20:47 NanoSector: ok boot -2 has nouveau
20:47 NanoSector: do i just > that to a file?
20:48 karolherbst: yeah
20:48 NanoSector: do i filter the nouveau messages?
20:49 NanoSector: http://hastebin.com/kuturowilu.coffee
20:49 karolherbst: mhh, let me check something
20:49 NanoSector: sure :)
20:52 NanoSector: i also keep seeing this at boot: nouveau 0000:01:00.0: bus: MMIO write of 00000002 FAULT at 4188ac [ IBUS ]
20:55 karolherbst: NanoSector: include/nvkm/core/subdev.h
20:56 NanoSector: in the awesome package?
20:56 karolherbst: there is a #define nvkm_printk(s,l,p,f,a...)
20:56 karolherbst: replace that line with this: #define nvkm_printk(s,l,p,f,a...) printf((f), ##a)
20:56 karolherbst: and run make again
20:56 NanoSector: karolherbst: is that in the awesome tool or in the module?
20:56 NanoSector: *reclocking branch
20:57 karolherbst: awesome_tool still
20:57 NanoSector: okay
20:57 karolherbst: and then run the command again, it should throw out some debugging messages
20:57 NanoSector: cool
20:58 NanoSector: repalce the entire line or just the part you posted?
20:58 NanoSector: *replace
20:58 karolherbst: the entire line
20:59 NanoSector: roger, it's running
20:59 NanoSector: make , that is
21:00 NanoSector: http://hastebin.com/puwaqipuhe.scala
21:02 karolherbst: mhh right
21:03 karolherbst: I am quite sure the image is in ACPI
21:03 NanoSector: any way to find out?
21:03 karolherbst: you need to reboot with nouveau again and fetch the vbios from /sys/kernel/debug/dri/1/vbios.rom
21:03 NanoSector: okay
21:03 NanoSector: that file was there when i was using nouevau
21:03 NanoSector: hang on again
21:04 NanoSector: do i place it somewhere specific?
21:05 NanoSector: got it now
21:06 karolherbst: wait a second, I need to check how we can pass that vbios to that tool
21:06 NanoSector: aight, I'll prepare nvidia again
21:07 karolherbst: yes
21:09 NanoSector: i'm ready if you are
21:09 NanoSector: :)
21:12 karolherbst: mhh, crap, I doubt that will work out that way that easily :/ k, then we will have to mess around a little
21:12 NanoSector: haha
21:13 karolherbst: ahh wait, it will work
21:14 karolherbst: drm/nouveau/nvkm/subdev/bios/shadow.c
21:14 karolherbst: lines 183 and 184
21:14 karolherbst: optarg = "path/to/your/vbios.rom";
21:14 karolherbst: source = optarg ? kstrndup(optarg, strlen(optarg), GFP_KERNEL) : NULL;
21:16 karolherbst: recompile and try again to run it
21:16 NanoSector: done, compiling now
21:16 NanoSector: oh that was quick
21:17 NanoSector: oh
21:17 NanoSector: lots more now
21:17 NanoSector: a lot
21:17 NanoSector: think that worked
21:17 NanoSector: do i just leave it run now?
21:17 karolherbst: yeah
21:17 karolherbst: run something with optirun in addition which needs a bit more load
21:18 karolherbst: mhh and higher perf levels :p
21:19 karolherbst: I would then just need the table printed by the tool, the nouveau logging stuff I don't need
21:19 NanoSector: is primusrun okay too?
21:19 karolherbst: yeah
21:19 NanoSector: optirun crashes
21:19 karolherbst: I just used optirun for the tool, because optirun always turns the gpu on
21:19 NanoSector: i notice the values changing in the tool when using primusrun glxgears so i think that works
21:22 karolherbst: mhh
21:22 NanoSector: uploading paste now
21:22 karolherbst: is glxgear capped at 60 fps?
21:22 NanoSector: with primus??
21:22 NanoSector: *-?
21:22 karolherbst: yeah
21:22 karolherbst: usually it should vsync
21:22 NanoSector: yes, want me to run with vblank_mode=0?
21:22 karolherbst: yeah
21:23 NanoSector: will do
21:25 NanoSector: i'll push it with some ridiculous resolution and 4x MSAA in Dolphin
21:26 NanoSector: 6x native and 4x MSAA, why not
21:26 NanoSector: how about 7x native and 4x SSAA
21:26 NanoSector: dat lag
21:27 NanoSector: karolherbst: http://hastebin.com/ufukolaxit.md
21:28 NanoSector: the tab in the thing at the start is me failing to alt+tab
21:29 karolherbst: mhhh, everything looks right indeed
21:29 karolherbst: so under nouveau it only hangs after you switched the boost level?
21:29 karolherbst: mind uploading your vbios somewhere? or did you already?
21:29 NanoSector: no also when I set pstate
21:29 NanoSector: I think I did in my gdrive
21:29 karolherbst: yeah, I found yours
21:29 NanoSector: it doesn't hang with glxgears but only with opengl apps it seems
21:30 NanoSector: oh okay
21:30 karolherbst: did you try boost 1 or 2?
21:30 NanoSector: 1
21:30 karolherbst: mhh
21:31 karolherbst: and the gpu just hangs immediatly or after a while?
21:31 NanoSector: after like 3 seconds, sometimes immediately
21:32 NanoSector: no, often immediately i should say
21:32 karolherbst: odd
21:33 NanoSector: so it's not a voltage issue?
21:33 karolherbst: don't think so
21:33 karolherbst: mind switching to nouveau again with nouveau.runpm=0 ?
21:33 NanoSector: yeah sure
21:34 NanoSector: mehh stupid bootloader not recognising me pressing its show key...
21:35 NanoSector: i'll just add it permanently, hang on lol
21:36 NanoSector: or set a timeout
21:36 NanoSector: ok booted
21:36 NanoSector: karolherbst: and now just test reclocking again?
21:38 karolherbst: yeah, but without running stuff
21:38 karolherbst: just go all the way up
21:38 karolherbst: and then execute sensors
21:38 karolherbst: it should print the currently set votlage
21:39 NanoSector: for state 07: 0.82V
21:39 NanoSector: for state 0a: 1,06V
21:39 karolherbst: I only care about 0f and boost 2 :p
21:40 NanoSector: oh
21:40 NanoSector: lol
21:40 karolherbst: and maybe boost 1 too
21:40 NanoSector: 1.18V for boost 2
21:40 NanoSector: and 0f
21:40 NanoSector: 1.14V for boost 1 and 0f
21:40 karolherbst: mhhh
21:40 karolherbst: let me check
21:40 NanoSector: max is 1.20 volts and min is 0.60
21:41 karolherbst: those 1.14V sound not enough
21:41 RSpliet: karolherbst: didn't you have that sysfs file that lets nouveau select a voltage slightly higher?
21:41 RSpliet: that'd confirm whether your "it's not voltage" theory is true or false ;-)
21:41 karolherbst: I removed it, because it didn't help at all
21:41 RSpliet: oh
21:41 karolherbst: it would be more usefull for undervolting actually
21:42 karolherbst: but I think we do something wrong regarding maxwell gpus on the clock domains
21:42 karolherbst: they changed stuff quite much there
21:43 karolherbst: mhh nope, 1.14 V sounds right
21:44 karolherbst: RSpliet: on nvf1 cards we can't read out the clocks right on high cstates. at one point the read out clock starts falling allthough nvidia dispalys higher clocks in nvidia-settings
21:45 RSpliet: mmm okay
21:45 karolherbst: the step sizes also changes at one point
21:45 RSpliet: btw, the output of your voltage measurement for the blob goes up to 1,18V, doesn't it?
21:45 NanoSector: I'm just hoping I'm helping you figure out something that's related to general reclocking, or at least more people than just me... :)
21:47 karolherbst: RSpliet: yes
21:47 karolherbst: NanoSector: well, I am sure it only affects maxwell
21:47 NanoSector: isn't the 750M still kepler?
21:47 karolherbst: nope
21:47 karolherbst: orrr
21:47 NanoSector: oh?
21:47 karolherbst: ohhh
21:47 karolherbst: you are right :D
21:47 karolherbst: it is a kepler
21:47 RSpliet: it's a DDR3 Kepler
21:47 NanoSector: :D
21:47 karolherbst: my mistake
21:47 karolherbst: forget what I said then
21:47 NanoSector: hehe np
21:48 NanoSector: RSpliet: there's also GDDR5 versions iirc, those perform close to a 760M
21:48 karolherbst: NanoSector: mind trying out boost 2?
21:48 NanoSector: with 0f?
21:48 karolherbst: yeah
21:48 NanoSector: or you mean running anything?
21:48 karolherbst: I mean running stuff
21:49 NanoSector: yeah sure
21:49 NanoSector: hmm
21:49 NanoSector: it fails to use nouveau
21:50 NanoSector: nvc0_screen_create:803 - Error allocating PGRAPH context for M2MF: -16
21:50 NanoSector: that's because I'm using that nouveau.run_pm=0 thing?
21:50 karolherbst: shouldn't matter#
21:51 NanoSector: also when the GPU hangs normally when reclocking, the values after AC in pstate go to 0
21:51 NanoSector: so core 0MHz and memory 0MHz
21:51 NanoSector: it's still showing valuesn ow
21:52 imirkin: NanoSector: runpm=0
21:52 imirkin: not run_pm
21:52 NanoSector: yeah sorry, i have it without underscore
21:52 NanoSector: no it's not running anything with PRIME now
21:52 NanoSector: it defaults back to the intel graphics and throws that error
21:53 imirkin: generally that happens when pgraph fails to come up, you should see stuff in dmesg yelling at you
21:53 NanoSector: heh, yeah
21:53 NanoSector: nouveau 0000:01:00:0: fifo: SCHED_ERROR 06 []
21:53 NanoSector: and that times a million
21:55 karolherbst: mhh
21:55 karolherbst: slowly I get the feeling, that we still mess something up with ddr3 on kepler
21:55 karolherbst: NanoSector: mind trying out 0a with boost 2?
21:56 NanoSector: it went flat
21:56 NanoSector: no pulse :P
21:56 NanoSector: oh, not true
21:56 NanoSector: wat
21:57 NanoSector: um okay
21:57 karolherbst: so?
21:57 NanoSector: PRIME doesn't work
21:57 NanoSector: the echo 0a > pstate command is stuck
21:57 karolherbst: well
21:57 NanoSector: and it went dead for a few seconds before coming back
21:57 karolherbst: you should reboot when the gpu got stuck
21:58 NanoSector: rebooting
21:58 NanoSector: does the SCHED_ERROR mean that the gpu is stuck?
22:00 karolherbst: yeah
22:00 karolherbst: or something like that
22:00 NanoSector: fifo: FB_FLUSH_TIMEOUT before it starts spamming SCHED_ERROR 0c's
22:01 NanoSector: [ 143.236003] nouveau 0000:01:00.0: fifo: write fault at 0000001000 engine 05 [BAR2] client 08 [HOST_CPU_NB] reason 02 [PTE] on channel -1 [00ffbf6000 unknown]
22:01 NanoSector: [ 143.237045] nouveau 0000:01:00.0: fifo: FB_FLUSH_TIMEOUT
22:01 NanoSector: [ 145.237168] nouveau 0000:01:00.0: timeout at /home/rkerkhof/Downloads/nouveau-stable_reclocking_kepler_v5/drm/nouveau/nvkm/subdev/mmu/gf100.c:190/gf100_vm_flush()!
22:02 karolherbst: does that happen after reclocking?
22:02 NanoSector: oh now it's actually throwing [ 242.913577] nouveau 0000:01:00.0: fifo: CHSW_ERROR 00000001
22:02 NanoSector: karolherbst: I reclocked before running DRI_PRIME=1 glxgears
22:03 karolherbst: boot with nouveau.config=NvMemExec=0
22:03 NanoSector: also the runpm=0?
22:03 karolherbst: you can remove that
22:03 NanoSector: ok will do
22:04 NanoSector: and off we go again
22:05 NanoSector: core and memory are 0MHz now after a fresh boot, I assume it's just powered off?
22:05 karolherbst: yes
22:06 NanoSector: do i reclock first or run an app first/
22:06 NanoSector: ?
22:06 karolherbst: well, while the gpu is suspended you can't reclock
22:06 NanoSector: oh right
22:06 NanoSector: so boost 2 and pstate 0a again?
22:08 NanoSector: immediate hang
22:08 karolherbst: yes
22:08 NanoSector: http://hastebin.com/unagacedoh.avrasm
22:09 NanoSector: then nouveau: waiting for kernel channels to go idle...
22:09 NanoSector: nouveau: failed to idle channel 0 [DRM]
22:10 karolherbst: mhh
22:12 NanoSector: so no assumptions yet?
22:12 NanoSector: :P
22:15 karolherbst: not yet
22:15 karolherbst: does it run stable with boost 0?
22:15 NanoSector: and 0a?
22:16 karolherbst: for example
22:16 NanoSector: hang on booting freshly
22:18 NanoSector: it's rock solid on boost 0 and pstate 07
22:18 NanoSector: pstate 0a seems to work too
22:18 NanoSector: and even 0f works
22:18 NanoSector: i booted with the nouveau.config=NvMemExec=0 btw
22:19 karolherbst: k
22:19 NanoSector: memory clocks don't move above 810 MHz now though
22:19 karolherbst: then something goes wrong on higher clocks
22:19 karolherbst: yeah
22:20 karolherbst: that's because nvmemexec disables memory reclocking
22:20 NanoSector: ahh okay
22:20 NanoSector: echoing 1 or 2 to boost causes an insta freeze
22:20 karolherbst: k
22:21 karolherbst: mupuf: we should check all the kepler/maxwell1 gpus you have and find any with reclocking issues
22:22 NanoSector: i'm getting gr: TRAP ch 2 (00f882000 glxgears[1127]) now
22:22 karolherbst: NanoSector: on boost 0?
22:22 NanoSector: no after issuing boost 1
22:22 karolherbst: ahh k
22:22 karolherbst: the thing is, your gpu has brutally high clocks
22:23 NanoSector: not normal for a 750M?
22:23 karolherbst: let me rephrase: the 750m has brutally high clocks :p
22:23 NanoSector: ah okay :P
22:23 karolherbst: only the 755m has higher ones
22:23 NanoSector: so in theory anyone with a 750M should have these issues right?
22:24 karolherbst: might be
22:24 karolherbst: I guess we do something wrong with high clocks
22:24 karolherbst: my gpu just goes up to 862MHz
22:24 karolherbst: and it is a 770m
22:25 NanoSector: but how can it be that much faster than a 750M then, though?
22:25 karolherbst: I could try to increase the voltage and clocks on mine and see if I can trigger some instabilities on higher clocks
22:25 karolherbst: more cores
22:25 NanoSector: just more cores?
22:25 karolherbst: yours have 384
22:25 karolherbst: mine has 960
22:25 karolherbst: :D
22:25 NanoSector: oh, haha
22:25 NanoSector: that's almost triple the amount :P
22:28 karolherbst: with boost 0 mine is actually at 705MHz
22:28 karolherbst: I will try to test stuff on the weekend
22:29 karolherbst: and you could do a mmiotrace of nvidia while letting nvidia reclock around a little
22:29 NanoSector: didn't i send you one a while back as well?
22:30 karolherbst: mhhh right
22:30 karolherbst: I didn't download it
22:30 karolherbst: do you still have the link?
22:30 NanoSector: ah i see
22:30 NanoSector: do you remember if i uploaded it to my google drive or not?
22:30 karolherbst: I think you did
22:30 NanoSector: i'll login, a sec
22:31 karolherbst: NanoSector: just to compare: a 780 ti also hardly reaches 1GHz
22:31 NanoSector: oh wow
22:31 NanoSector: i have a trace.log.xz
22:31 NanoSector: think that'si t
22:31 NanoSector: https://drive.google.com/open?id=0B2F3P3d66S-1aWt1S0hDc0xkbFU
22:31 karolherbst: and boost 1 already wants like 1.085 GHz or so for you
22:32 karolherbst: yep, looks good
22:34 NanoSector: think that was on linux 4.6 though, do you want me to do another on 4.7?
22:35 karolherbst: RSpliet: any idea about stuff we might miss out for high engine clocks?
22:35 karolherbst: NanoSector: doesn't matter
22:35 NanoSector: ah ok
22:46 NanoSector: karolherbst: well, poke me if you find anything new or want to test anything, I'll be right here :) and i hope i was able to help you out
22:53 RSpliet: karolherbst: no idea, never looked into Kepler PLLs
22:59 karolherbst: ever encountered something like that on a fermi?
23:26 RSpliet: as if clock changing works on those :-P
23:26 RSpliet: but not for core
23:39 Tom^: hardly reaches 1ghz ?
23:39 Tom^: im on 1100 all the time
23:40 Tom^: or more specificly 1097 :D