00:11 <steev> i've seen that sometimes, where i have to plug and unplug it a few times to work, not sure why

06:33 <JensGlathe[m]> are there GPU drivers to run GeekBench on sc8280xp / x1e? My (superficial) search only found announcements.

06:36 <HdkR> JensGlathe[m]: Is this the geekbench AI thing?

06:37 <JensGlathe[m]> no just GPU would be sufficient. AI would be nce, but doubt it a little

06:37 <HdkR> Geebench itself is a CPU only bench

06:37 <JensGlathe[m]> https://browser.geekbench.com/v6/cpu/8162996

06:38 <HdkR> ?

06:38 <HdkR> That's all CPU

06:39 <JensGlathe[m]> yes, but you can run compute, too afair if you have OpenCL or Vulkan

06:40 <HdkR> Oh, weird

06:41 <HdkR> Is the GPU bench only on Windows or something?

06:42 <JensGlathe[m]> no I did this on Linux IMO, but that was a few years ago

06:43 <HdkR> Huh, I guess it doesn't run the GPU bench by default? Need to pass in --compute or --gpu on latest

06:44 <HdkR> `geekbench_x86_64: src/geekbench/workload/vulkan/vulkan_kernel.h:83: void VulkanKernel::set_local_size(const std::array<uint32_t, 3> &): Assertion `cmd_buffer_.pool()->device()->properties().limits.maxComputeWorkGroupSize[i] >= local_size_[i]' failed.

06:44 <JensGlathe[m]> https://pastebin.com/pNWbdkDN

06:44 <HdkR> awww

06:49 <HdkR> JensGlathe[m]: https://browser.geekbench.com/v6/compute/2889962 Have fun with that

06:50 <JensGlathe[m]> thx, GenuineIntel Family 6 Model 166 Stepping 1 🤨

06:50 <HdkR> I have no idea what my current FEX setup is, so heft amount of salt

06:51 <HdkR> https://browser.geekbench.com/v6/compute/compare/2889962?baseline=2889976 Compared to my desktop

06:51 <JensGlathe[m]> okay, but thanks

06:55 <HdkR> `[1005/234528:ERROR:compute_workload.cpp(120)] workload 303 failed validation ` Not sure if CPU bug or GPU bug that this test failed

08:55 <SpieringsAE> JosDehaes[m]: I would say 1m42 to 1m15 is quit a bit faster lol, but yeah the Apple chips are just better. The issue for me is MacOS and the price. I dont want to pay 3k+ for a laptop with usable specs and then be stuck in the apple walled garden. Also I see now there are some other dutchies here lol

09:03 <JosDehaes[m]> Yeah apple gear is 🫰. I'd say my M1 MBA is still a great device 4 years on (and with useable specs it was 1600 euro). At that time there was nothing comparable at any price point

09:03 <JosDehaes[m]> Btw, not Dutch, Belgian 😁

09:04 <SpieringsAE> ah sorry for the assumption

11:07 <JensGlathe[m]> Some writeup re [Bluetooth on X13s and others](https://github.com/jglathe/linux_ms_dev_kit/wiki/Getting-better-range-and-quality-from-the-WCN6855-Bluetooth-chipset)

11:08 <JensGlathe[m]> From the code, looks like on could also enable it for WCN7850, but I don't have these to test

11:29 <Jasper[m]> <SpieringsAE> "ah sorry for the assumption" <- We have a lack of ae last namss

11:29 <Jasper[m]> *names

11:30 <anonymix007[m]> Jens Glathe: Looks like it's already working on WCN7850 since it asked for hmtnv20.b112 on T14s

11:37 <anonymix007[m]> And HSP looks to be an abbreviation for internal chip code names (HaStings Prime probably), i.e. WCN785x is Hamilton, WCN685x is Hastings Prime.

11:48 <JensGlathe[m]> In marketing "documentation" I've seen Headset Profile

11:49 <JensGlathe[m]> for 7850 the fallback if the specific firmware is mising isn't there I think. But thanks for testing 🙂

11:55 <anonymix007[m]> Yes, the fallback is missing. Though it's not a big deal, I'd rather have something failing loudly and not silently working bad

11:56 <JensGlathe[m]> but yes, interesting. I guess there needs to be some fine tuning. qca_get_nvm_name_generic() is a different implementation of the same thing.

11:57 <anonymix007[m]> Headset Profile doesn't need separate firmware though, and leaked qcom mobile code suggests that these abbreviations are really code names

11:59 <JensGlathe[m]> okay

12:02 <JensGlathe[m]> So for WCN6855 and WCN7850 it could be "unified" with fallback. Wit a "loud" bt_dev_info("Specific fw file %s not found, falling ack to %s - performance ma be suboptimal!"

12:19 <JensGlathe[m]> will do this as v2

13:57 <erebion[m]> By the way, the issues I have with the Pinephone dock seem to be caused by the Pinephone dock, not by the X13s or kernel. It works fine if I plug in power into the dock AFTER docking, not always if I do it before docking

13:59 <erebion[m]> Another issue I noticed with Sway:... (full message at <https://matrix.org/oftc/media/v1/media/download/AV1Hix3esfMkaFxY_GJfgoIefivG4QcR1fGMvKJ-uwZW99v0N77WVQ0Gi5TTZFUxkqetJNO2fLh3yli7MiZYL7VCeSY9gJ8AAG1hdHJpeC5vcmcvaWxQRkNHS3pNdEdJV2hCZHVEdmpLeVVz>)

13:59 <erebion[m]> I just noticed it works with external keyboards, but not the one in the laptop itself.

14:00 <erebion[m]> Is there any reason Super+Shift+Right would not work on that keyboard?

14:05 <macc24> <erebion[m]> "By the way, the issues I have..." <- huh, i never had issues with it

14:15 <erebion[m]> Maybe my dock has issues or my power supply then?

14:24 <robclark> HdkR: re: compute benchmarks, there is a branch with rusticl support.. idk if clvk+tu would be better, I suspect that cl kernels (since you don't know the local wg size at compile time) will have a lot more register spilling than needed, so probably a bit suboptimal. (Not sure how/if clvk handles that.)

17:03 <macc24> o/ alpernebbi any plans to have efivars supported in libreboot on kevin?

17:03 <macc24> and other rk3399 machines

17:14 <alpernebbi> not soon I guess, but I want to, everything else keeps getting in the way

17:16 <macc24> okay, i'll just work around this

17:16 <alpernebbi> any specific problem?

17:17 <macc24> by default in cadmium i'm putting systemd-boot in esp/efi/systemd/ and that would require adding boot variable to point efi at it, for now i've added a workaround of putting it into esp/efi/boot/bootaa64.efi

17:17 <alpernebbi> (fwiw I just put rEFInd in /boot/EFI/efi/bootaa64.efi and it kinda works)

17:17 <alpernebbi> (or whatever path it was)

17:20 <alpernebbi> yeah you'd need efivars support for that, sorry about it

17:21 <macc24> it's fine, don't worry about it

18:28 <SpieringsAE> wait, there are open uefi firmwares for arm chromebooks?

18:29 <SpieringsAE> I knew it for x86 ones, I have one myself with mrchromebox firmware, but I thought arm was stuck to funky boot partitions, at least thats how I got linux booting on my girlfriends chromebook with a mediatek chip

18:29 <macc24> SpieringsAE: u-boot can run efi binaries

18:30 <macc24> i don't recall anyone having u-boot running on an mtk chromebook tho

18:30 <SpieringsAE> Thats true, never done that though, does that actually work properly though with stuff like efibootmgr?

18:32 <macc24> in libreboot efibootmgr doesn't really work as efivars don't work - but aside that it's usable enough for linux

19:55 <erebion[m]> Wait, the X13s actually does not suspend. It really just turns off the screen. :D

19:55 <abby> it's more than the screen

19:55 <abby> it's s2idle

19:55 <erebion[m]> What would be a USB-C dock that works well with the X13s? I need USB-C to plug in stuff and Diplay port and charging. Nothing fancy. :D

19:56 <erebion[m]> Any Thinkpad docks that are known to work well without pain?

19:59 <JensGlathe[m]> The without pain part is... ephemeral. I have a display that connects with type-c, with integrated dock, and that is quite useful. But all dock functions are limited to USB-2 as of now, its getting only one lane for data.

20:04 <erebion[m]> Well, that would at least allow me to connect keyboard, mouse and my Debian phone, which only has a 2.0 port anyway.

20:05 <JensGlathe[m]> iiyama ProLite XUB2792QSN-B5 27" Monitor 2560x1440 USB-C LAN HDMI DP Dock USB3.0

20:05 <JensGlathe[m]> https://www.ebay.de/itm/195990504695

20:07 <erebion[m]> Suspend produces:

20:07 <erebion[m]> Lockdown: systemd-logind: hibernation is restricted; see man kernel_lockdown.7

20:07 <erebion[m]> So I don't think it suspends, considering that the screen stays on and it is still reachable via SSH

20:08 <JensGlathe[m]> I get mine to suspend, Ubuntu 24.04 (no dock)

20:08 <erebion[m]> Though I already have a screen, which I actually bought just a couple weeks ago, so a dock would be a better option

20:09 <erebion[m]> Probably without Secure Boot, then?

20:09 <JensGlathe[m]> yes, without secure boot

20:10 <erebion[m]> Although, suspend works fine on a different laptop with Secure Boot enabled.

20:11 <erebion[m]> Also, why " Lockdown: systemd-logind: hibernation is restricted; see man kernel_lockdown.7"? I'm trying to suspend, not hibernate.

20:16 <erebion[m]> I just researched s2idle and realised that it does what it is supposed to do and I just did except a different thing to happen.

20:30 <JensGlathe[m]> when the ath11k firmware acts up with wrong ex buffer lengths, it also seems to tie up the complete bandwith of the router (or something to that effect)

20:31 <JensGlathe[m]> Wife was complaining that her iPad movie has stopped

20:35 <JensGlathe[m]> probably a good idea to just reboot the firmware when that happens

22:09 <erebion[m]> Wifi is a bit slow for me, right now i3status tells me 180 Mbit/s. I'd love a bit more as I'm also shoveling a lot of data around my network.

22:25 <erebion[m]> Also, should I change any configs to get more than 7 hours on full charge? That's what it says when unplugging the charger without it doing more than running Sway.

22:47 <zdykstra> What's your power draw at idle?