I still wish I had a better GPU
My 7900xtx was a rollercoaster of emotions all the way till now, so I decided to buy nvidia even tho I don’t support them as a company in the slightest (nor do I support amd at this point, considering they dipped in same AI bubble as soon as they had the chance).
So yeah, “I will just buy next top-tier gpu as soon as it arrives” was looking like this for past couple of years:
Realistically, company shouldn’t enforce such integration in the first place, even if my distro did supporter it (I won’t install RHEL on my home desktop, dude).
They justify it by “security”, but beyond changing group policy to enforce constant password rotation and system update, haven’t noticed anything that would really justify it. I guess the only real reason of this integration existing, is that they can lock me out of my own PC if they wanted to. And considering that I’m a freelancer, that can happen at any given moment.
I’m a freelancer. Cause I had to run from my country that decided to start a war for some reason. And let me tell you - no one wants to hire a migrant from russia. So yeah, upper mentioned company won’t provide hardware for freelancers, but as alternative they do provide azure virtual desktop - which sucks and also doesn’t work on linux (due to how ms implemented auth via their token).
I can request asylum and live on government support, will probably get a rejection in the long term tho, cause according to EU and other countries - russia clearly is a safe country to live.
And suddenly me “unjustifiably” buying 128gb of DDR5 was a good investment of money.
At least in your case they didn’t give the false hope. In my case two years ago company just stated that we “enrolling in ms ecosystem”. That meant that to use your PC within company’s domain, you had to authorize it with MS account or none of sites would work. Weirdly, MS does support that not only on windows, but also on linux and MacOS. And if MacOS is simple (and supported by our tech team), linux installation is awful — it requires ubuntu lts + to install ms antivirus, that will override some system filles on top of that and will have root access afaik. Well, I don’t want company to control my personal device anyway, so I just installed win11 in qemu with tpm emulation and gpu passthrough and activated this “company account” shit there. Good luck spying on not activated bare w11 copy.
If job market wasn’t fucked, probably would just leave instead tho.
Gedit for docs (plain and simple + have basic spelling/code highlighting), kate for weird things like JSON (have actual tools to fix spacing/tabulation)
No-no, let him cook, he onto something.
Imagine how efficient it will be!
I’m not nearly educated enough to speak math, but I think acceleration/deceleration makes a big portion here. Train won’t move 300-600km/h all the time. Plus onboarding/offboarding and delays should also be considered.
After watching recent ram cartel investigation from gamernexus: fuck them. They got what they fucking deserve.
What doesn’t at this point? All repos on github contaminated at this point. In couple of years border between code created by humans and code created by llm’s will be basically non existent.
The end is near. git clone until it’s too late!
Nothing is stable in this world nowadays.
I’m doing my part!
For some reason when I use sddm shutdown sometimes just doesn’t work. Using GDM makes display doesn’t work (weird hdmi quirk with bitdepth). So instead I login via TTY and start my session manually. At least then I can shutdown my system. Usually.
Context limit is not really a problem on local models. Qwen3.6 can do up to ~256k tokens, it’s not that far from what things like cursor uses. grok in cursor have exactly 256k tokens for example. Also, you can use opencode with custom config, where you need to set trimming close to that number.
That being said, quality wise it still kinda shit and looses to paid models. I think you need something like GLM 5.2 to compete, which needs 228gb at lowest 1bit quant (not including context size, which can be up to 1m tokens, so you can round up requirements for RAM up to 256gb), so yeah, the only thing that limits you locally — the fact that “AI” companies bought all supply of focking ram and we can’t afford any.
I’ve been told that using linux similar to abusive relationships.
I wouldn’t know — I never had relationships.
Actually, I think dualshock works, it’s dualsence that have issues. Also, it works okay (I tested) in CachyOS, so whatever settings they made to udev rules or whatever — fixed it. Or they may just used old version of bluez, which is also a workaround, but you need to install insecure and 2 years old version at this point.
Also-also, it works fine via USB.
Oh, we doing “what doesn’t work for you” comment section? Count me in!
My LG TV consistently broken on multiple amd devices on linux. Something to do with how LG implemented HDMI protocol and how AMDGPU driver percepts it. I think it have something to do with color bit depth, cause all DEs allow to set 16 bit color (which is insane, not supported by any software nor my hardware and shouldn’t be a allowed as option without some special flags in my opinion) but like, there no protection and some DE/WM even defaults to max available bit depth, which in turn straight up doesn’t work and result in black screen.
Windows or nvidia — those work fine.
Another pain in the ass for me — connecting dualsense controller via bluetooth. It connects, then just disconnects. There a workaround that includes fetching it’s ID via terminal, then using commands to connect and entrust that device. After that it connects fine. Until you need to reconnect it, that is, cause it stops being recognized. Apparently that was fixed like, a year ago. Yeah, right.
Another pain for me would be steam client on wayland, but Valve finally started working in that direction so I will hold my complains for now.
Same. But more than that - I noticed that my 7700x is a huge bottleneck: desktop CPUs limited to just 2 channels, plus controller in 7700 is kinda meh, so I can’t force anything above 5600mt\s with my 4 sticks. Switching to new CPU should allow to increase speed, but will still lock me to 2 channels. So I should’ve bought cheapest threadripper instead, which allows 4 or 8 channels instead.