I think Gufo is optimized only for the Strix Halo but llama.cpp should get you running with expert offloading no problem.

Is this the future version of holding two phones together to transmit files via infrared?

I’m running Gufo on a Strix Halo PC with 128GB shared memory.

The quant is unsloth’s UD-Q4_K_XL. Speed is really good with this configuration (1600 t/s prefill and 100 t/s decode). I wanted to try some higher quants as well but I’m waiting for Qwen 4 for now to do more testing.

You should be able to get QFN running quite well on your system as well when offloading unused experts to RAM.

I don’t know what they are testing but I use QFN for everything.

Assuming it doesn’t have to rely on internal knowledge too much by giving it tools to retrieve information, it does everything I ask it to.

I challenged two of your claims.

I have not made any statements on whether LLM generated decompilations are good or bad.

This is a shit load of code, serious devs capable of doing this would be doing something that paid them instead.

That’s just not true. There’s plenty of passion projects out there of people hand decompiling code for massive games.

Decompilations and emulators have been a thing long before LLMs even existed.

Which is why we have playable Ocarina of Time remake emulators and decomps within 2 or 3 days all of a sudden.

No, we have Ocarina of Time emulators and decomps within 2 or 3 days because the yuzu folks spent years reverse engineering and documenting every detail about the Switch 1.

The Switch 2 is so similar to the Switch 1, the first proof of concept Switch 2 emulator is 95% just Eden/yuzu. Things would have taken much longer if yuzu didn’t exist, even with LLMs.

Please tell me this didn’t break SteamVR again…

I would have been fine with them releasing the same game, but they released an inferior game for the third time.

I don’t really want to use Steam as a launcher. It’s proprietary and not the most intuitive when it comes to navigation outside of gaming, and most of the usage will be non-Steam stuff.

I also don’t want SteamOS on my HTPC, I want something more open that does not send requests to Valve on every boot. That’s fine on my Steam Deck but not on my HTPC.

Will definitely test the power consumption and wake up on Steam Controller power on in third party distros.

I assume the low power carries over since they did a lot of work in the kernel.

I’m planning to eventually do a fresh Linux install on my Machine which launches straight into Plasma Bigscreen.

Already got the multimedia situation sorted out with HDR, 4k and audio passthrough via Jellyfin MPV Shim. Now I just need to check if the Machine still gets firmware updates when launching Steam even when it is not running SteamOS.

Didn’t have time to do the setup yet but I’m planning to replace my Odroid N2+ running CoreELEC with this.

For controlling it I wanted to use my Remote 3 and/or the Steam controller.

Bitwarden keeps a copy of your vault on clients.

So even if you lose your Vaultwarden/Bitwarden server files, you will have a local copy on every device you use.

It comes with defaults that are very close to what I’m looking for.

Fedora lost me to the mesa-freeworld dance I had to do every major update.

Fantastic news. My only gripe with 3.8 is that 262k is very limiting with all the reasoning the model does.

Pi ends up compacting history 3-4 times for larger tasks before it finishes.

I use the spacer as well to fit my neanderthal eyebrows in the headset.

I played around with the left/right only setting a little bit and it appears the right eye causes the jump to the bottom right when blinking.

The left eye appears to mostly work but it has trouble reaching the bottom, I need to look a little bit into the dark area before the eye tracking goes further down. Which would explain why I sometimes can see the foveated streaming.

Not sure what causes these issues with my lenses but I wrote them a mail about it and maybe they can swap me to the eye-tracking variant since they promised working eye tracking on the Frame.

Now that seems to be a useful option to look into.

From the first quick test, the debug region does follow my vision pretty well, but it’s not what I would call “instantaneously”. For small eye movements it works but for flicking left/right it takes a tiny moment to get there.

Is there a way to get this overlay without streaming? Would be useful to test without any network delay, just to rule one component out. I might have to pick up some standalone games with foveated rendering to do some more testing.

For the blinking issue, it might be the VROptician comfort lenses. With the lenses installed, the debug region flicks to the bottom right for a moment after opening my eyes again. I sent them a mail asking about it.

So that’s where it was! The menus are really a mess on the Frame.

The system profile is already set to performance (I think that’s the default?). Still the same issues.

Do you happen to know which settings?

I know that there was a performance and quality setting when I got it but I can’t seem to find it in the current beta version. No other power saving settings I could find.

Streaming gets a little bit better when using 144 Hz instead of 90 Hz but it’s still not what I would call “unnoticeable”.

Use Proton-GE! It has fixes to make the overlay work under Wayland.

I have been playing in HDR with a Steam Controller (which needs the overlay) for a while and it works perfectly.

I really hope that all the work that has been going into the process freeze/unfreeze functionality on Linux at some point leads to the ability to have universal save states for every game.

Would be so cool to just restore to one of the last savestates in any game, no matter how crappy the save system in the given game might be.