This is the finished build btw!
17.12.2025 21:57
👍 1
🔁 0
💬 0
📌 0
This is the finished build btw!
And I run 250W version with i9-14900K, 7 SSD disks and Arc Pro B50! Never had a problem, but ofc i9 is power limited, though still impressive IMO.
Yep, but I a little bit regret that I didn't pull a trigger on another 96 GB kit! I thought "well, 700 is 100 euros more than I paid a year ago, will wait until prices calm down" 💀
Yep, I talked with ASRock Rack support and they basically told me that no one tests MB under those conditions so it's most likely Intel firmware or even a hardware issue which no one wants to fix.
@notthebe.ee hi! I was tinkering hella a lot with Raptor Lake W680 server board with pretty powerful BIOS, but without a lot of success unfortunately. I managed to get C10 only without GPU. Here is my full journey, feel free to ask!
But I have to add that the entire stack of 10 containers added 10W to the server's power consumption. For "doing nothing" that's quite a lot.
It turned out MAS basically replaced the legacy SSO integration, so I had to migrate. I would clarify that point in the docs better. Also, it seems that standalone Element Call doesn't support MAS for some reason. But overall, it works better than alternative servers.
I've decided to give it a second spin and picked Synapse this time; hopefully I have a more powerful machine. It worked better that time, with fewer annoying bugs. But I found the documentation unhelpful for OIDC/SSO.
Do you see the light at the end of the tunnel? Because for now I do not. Panther Lake doesn't really impress me. I don't think Windows/Linux laptops will magically last much longer after switching to a new node process. AMDs next mobile chip will probably be incremental. Qualcomm? Too many problems.
How did we end up in this situation? I don't like Apple products because of their walled-garden nature, but I have to admit that we're at the point where the MacBook is light years ahead in technological terms.
Hell, even those heavy-duty workstations are not matching Apple's single-core performance. Memory bandwidth with SODIMM DDR5 is at least two times smaller. Maybe in some cases a discrete (!) GPU can win the graphics battle, but that's it, lol.
There are no Windows or Linux laptops that will last a full day. None. There are no laptops that come even close to the M4 Pro/Max-maybe only those with a basically desktop-class chip inside a laptop-like chassis and abysmally small battery life.
I really want to buy a good Linux laptop like a ThinkPad or Framework, but I just cannot ignore how much further Apple is in terms of performance and efficiency.
The situation around Windows/Linux laptops is bizarre. While AMD clearly has a lead in the Intel vs AMD race, Apple is just smashing both of them.
And most probably this is the only thing he hosts on that Raspberry.
So we are stuck forever with Synapse which is like to use a sledgehammer to crack a nut. I just cannot host it on a small server like Raspberry Pi.
And I again faced all those bugs, slowness and everything I remembered about Matrix. But now I am pissed even more, because the protocol became SO bloated and complex that it's seemingly impossible to write a new implementation from scratch.
...and since I don't have very powerful server, I was shocked to know that Dendrite is basically abandonware now. Synapse requires a ton of resources so it's a no-go for me. So I went with Tuwunel.
Well, I've decided to give a second chance, but since I don't want to use your server (because sorry, but what is the point of your fancy protocol then if it cannot be relatively easily implemented and work properly in the federation?)
This explains a lot, I didn't know it personally. In my opinion desktop should be a priority for you, I don't use mobile app a lot, most of chats in Matrix are hobby/work related for me.
Latest Element for Desktop, use it rarely from time to time when there are no other options, my account is on envs.net instance, they use Synapse with sliding sync support.
In my case I almost never could join the big conference without pain. Missed messages, missed order of messages, missed decryption, etc. It just doesn't freaking work. And they call it a day!
I assume that "fast loading times" only applies if you are a user of matrix.org instance.
In my experience if you are trying use anything else than their flagship instance, you will instantly notice that their federation works good only on paper.
Unironically big thanks for using Discord instead of Element/Matrix.
I genuinely doesn't understand how open source community uses that mess. Literally almost anything is better.
Tried, but got an error. Probably did:web thing again.
Got it, makes sense, thank you for your response!
I understand that it will undermine your goal to force Google extend their program and allow to certify custom ROMs, but I am not sure what is more important here: potentially larger userbase or priciple prevailing.
Probably dumb question, but in case you will end up selling your own device, isn't it will be also possible to certify the device, pass all Google tests and receive their approval to pass Play Integrity checks?
This is basically what "Year N Pass" pass is.
Their Snowrunners DLCs are pretty simple. Every year they release 3 minor updates (usually it's relatively small region with 2 maps) and 1 major (with big region and 3-4 maps inside it). Each DLC (minor/major) has its transport. Any other DLC you can easily skip, they are usually junk.