00:13 <craftyguy> calebccff: ya I feel your pain wrt missing kvm, but on my x13s (ya you're probably tired of hearing me say that D:)

01:35 <anthony25> robclark: thanks again for the debugging for the glitches on Hyprland, I sent a PR that uses the caps checks before enabling explicit (aquamarine was checking it already, just the Hyprland side never used it…)

01:35 <robclark> thx

18:09 <pengyu[m]> they said fedora aarch64 iso is signed with Microsoft key, we can install it with secure boot, but I cannot boot into the grub provided by iso on Qualcomm laptop

18:37 <robclark> konradybcio, jhovold: with the devkit no longer being a thing, is there any other NuC form-factor (ie. mainly no-battery) that someone has started looking at? (I'm looking for something that we could use for mesa+drm CI .. we need to be able to remotely cycle power..)

18:38 <SpieringsAE> robclark: the only thing I know that is coming is this one: https://liliputing.com/geekom-qs1-is-the-first-mini-pc-with-a-snapdragon-x-elite-processor/

18:39 <Jasper[m]> Wanted to mention that

18:39 <konradybcio> robclark: x1e?

18:39 <Jasper[m]> As far as lower end sku's go, there were some announcements at CES iirc

18:39 <robclark> purwa or hamoa

18:39 <Jasper[m]> I think Asus and Lenovo are making their models

18:39 <SpieringsAE> this is hamoa

18:39 <SpieringsAE> I believe its the 80 or 84

18:39 <Jasper[m]> Jasper[m]: These are purwa or lower

18:39 <robclark> deqp is mostly cpu limited so maybe hamoa would have some benefits

18:40 <Jasper[m]> (if that even exists)

18:40 <robclark> but I'll take what I can get

18:40 <SpieringsAE> yeah that geekom one has the X1E-80-100

18:40 <SpieringsAE> so little bit of extra boost on some cores as well

18:41 <SpieringsAE> does the 80 also have the higher gpu clock? Or is that 84/DE

18:41 <robclark> I think maybe 80 does.. although tbh that probably won't matter, for the most part the deqp tests don't submit a whole lot of rendering

18:43 <konradybcio> robclark i dont think theres anything on the market

18:43 <konradybcio> only laptops and these mini pcs

18:43 <konradybcio> but doesnt seem like there's a HDK like board

18:44 <robclark> ahh, lenovo has one too (mini pc)... but are any of those shipping yet?

18:44 <maz> ah?

18:45 <robclark> https://www.qualcomm.com/news/onq/2025/01/the-desktop-pc-reborn-snapdragon-x-series-powers-lenovo-mini-desktop-pcs

18:45 <robclark> hmm, ok, one of the lenovo ones is expected in Feb

18:45 <SpieringsAE> Only see X plus and just X mentioned there

18:45 <SpieringsAE> no X elite :(

19:02 <robclark> for what I need, with purwa I'd just buy 50% more devices to offset the core count difference ;-)

19:09 <JensGlathe[m]> purwa seems pretty efficient, though

19:09 <JensGlathe[m]> roughly scales linearly for CPU

19:12 <robclark> for mesa CI we run as many threads as we can on a runner (lightly oversubscribing, so like 20% more threads than cpu cores, seems to work well), and then shard across as many runners as we need to keep runtime down to ~10min for pre-merge CI..

19:22 <HdkR> Sharding is definitely interesting. So far I've been able to get away with throwing faster hardware at the problem and lowering the number of tests on slower hardware :D

19:31 <SpieringsAE> would love it if arch linux arm got some better build servers lol

19:31 <SpieringsAE> almost on week 2 of sleepy time

19:34 <robclark> HdkR: runtime of deqp-vk, in particular, seems to keep growing ;-)

19:35 <SpieringsAE> oof

19:36 <HdkR> eef

21:22 <\[m]> https://marcan.st/2025/02/resigning-as-asahi-linux-project-lead/ that sucks

21:22 <abby> yeah :/

21:28 <HdkR> Looks like Asahi has a good path forward planned with a good number of contributers though

21:28 <HdkR> So keep marching forward as usual

23:16 <icecream95> Huh, so Windows (both native and WSL) is about twice as fast as Linux for llama.cpp generation on the CPU. So that means that there definitely is some "secret" switch to double the read memory bandwidth?

23:26 <bamse> icecream95: x1e?

23:29 <icecream95> bamse: yes