2ghz does not measure it’s computing power though, only the cycle speed. Two very different things.

An objective measure is a simple benchmark:

Here’s a quad core 1.5ghz RISC-V SoC (noted as VisionFive 2) vs a quad core 1.8ghz ARM chip (noted as Raspberry Pi 400).

It’s not even remotely close to usable for all but the most basic of tasks https://www.phoronix.com/review/visionfive2-riscv-benchmarks/6

Well that’s true if you have a live animal producing your meat. Not sure that applies if the meat is lab grown though?

100% they absolutely were.

Give geneticists 20 years, we’ll have lab grown T-Rex in the grocery store

Depends on your goals. For raw tokens per second, yeah you want an Nvidia card with enough^™ memory for your target model(s).

But if you don’t care so much for speed beyond a certain amount, or you’re okay sacrificing some speed for economy, AMD RX7900 XT/XTX or 9070 both work pretty well for small to mid sized local models.

Otherwise you can look at the SOC type solutions like AMD Strix Halo or Nvidia DGX for more model size at the cost of speed, but always look for reputable benchmarks showing ‘enough’ speed for your use case.

Yeah… I mean, I did hedge by saying “depends on your CPU and your risk profile”, but I understand your point and will edit my comment to caution readers before playing with foot finding firearms.

From my understanding it’s a mixed bag. Some of those vulnerabilities were little more than theoretical exploits from within high levels of trust, like this one. Important if you’re doing a PaaS/IaaS workload like AWS, GCP etc and you need to keep unknown workloads safe, and your hypervisor safe from unknown workloads.

Others were super scary direct access to in-memory processes type vulnerabilities. On Linux you can disable certain mitigations while not disabling others, so in theory you could find your way to better performance at a near zero threat increase, but yes, better safe than sorry.

I apologize for being glib.

Agreed, shouldn’t affect performance. But also depends on how they see best to patch the vulnerability. The microcode patch mechanism is the currently understood vector, but might not be the only way to exploit the actual underlying vulnerability.

I remember the early days of Spectre when the mitigation was “disable branch prediction”, then later they patched a more targeted, performant solution in.

no performance change

You must be new here.

Joking. In reality it depends.

The first iteration of this comment had a cheeky observation about the performance impact of these CPU mitigations on Linux, some of which have nearly no real world threat to people not running cloud providers.

And while that’s true to a degree, tests disabling some or all of the most modern set of mitigations show that most have become highly optimized and the CPUs themselves have iterated over time to increase the performance of the mitigations as well.

And many of these CPU vulnerabilities actually had in the wild use and can still do horrible things with very little surface exposure from your system. Apologies to the people who read the first version of this comment and took the time to rightly push back.

Yup. When you “both sides” insanity, you normalize it.

Doubt it was on purpose. Bastard is a MAGA sycophant.

But hey, we can hope he’s having a stroke.

Can you provide your docker-compose entry or your docker run command?

Okay, thanks for giving me that, I’ll investigate further tonight

So I just spot checked. Both shows work, you just have to not click an episode anymore.

E.g, https://pbskids.org/videos/design-squad -> design-squad

Thank you for telling me, I’ll update the readme

Hmmm. I just double checked and my episodes are still downloading. But maybe newer shows have a different format… What’s the exact error? I’ll try to reproduce and fix.

https://gitlab.com/mwirth001/elinorr

https://gitlab.com/mwirth001/elinorr/container_registry

https://gitlab.com/mwirth001/elinorr

https://gitlab.com/mwirth001/elinorr/container_registry

The streaming was easy, just declared I wasn’t paying for it anymore lol. We still have a crappy version of Spotify for free because of another service (ISP or phone plan something like that), but it’s purely used as a backup.

Jellyfin’s interface is a bit clunky as a music client in my experience. FinAmp looks cool but it’s still early on.

Navidrome does smart playlist, crossfading, gapless, flac streaming, and flac to opus transcoding. Those are sorta my core requirements, and Navidrome + the clients we use handles them all with aplomb.

As for the user playlist thing… I haven’t seen anything like that but maybe I’m misunderstanding.

Yeah Music Assistant uses Snapcast, which has been fun. I did try squeeze, but haven’t had a reason to switch so far

It’s a crappy python script I packaged in a docker container lol. Turns out PBS kids uses an open unauthenticated CDN for serving videos to the website and apps.

I can share if you want, but it’ll take me until tomorrow to make it public

Certainly!

Jellyfin I use for video content. I find its music functions lackluster.

Navidrome I use (and my family uses) for personal listening.

Music around the house, like on one or more of my casting capable speakers / tvs I use Music Assistant. Also let’s me do automations easily, and doesn’t tie up an android phones media’s output. Struggled with earbuds while casting taking over audio for too long before deploying Music Assistant!

I will add, what helped me the most with Plex/Jellyfin load was using Tdarr to normalize my library’s formats into something easy to direct stream to any device without transcoding.