intel arc b580 (i think that's the latest one) isn't obnoxiously priced but you're going to have to face the fact that your PCIE is really very slow. But it should work.
if you want to save even more money get the older Arc Battlemage GPUs. I used one it was comparable with an RTX 3060; i returned it because the machine i was running it in had a bug that was fixed 2 days before i returned it but i didn't know that.
I was seriously considering getting a b580 or waiting until the b*70 came out with more memory, although at this point i doubt it will be very affordable considering VRAM prices going up as well. A friend is supposedly going to ship me a few GTX 1080ti cards so i can delay buying newer cards for a bit.
By older Arc, I presume you're referring to Alchemist and not Battlemage in this case.
One of my brothers has a PC I built for him, specced out with an Intel Core i5 13400f CPU and an Intel Arc A770 GPU, and it still works great for his needs in 2025.
Surely, Battlemage is more efficient and more compatible in some ways over Alchemist. But if you keep your expectations in check, it will do just fine in many scenarios. Just avoid any games using Unreal Engine 5.
yeah i had an A770; it should be ~$200-$250 now on ebay, lightly used. It's, in my opinion, worth about $200, if it's relatively unused. As i mentioned, it's ~= RTX 3060 at least for compute loads, and the 16GB is nice to have for that. But for a computer from the 4th gen i'd probably only get a A380 or A580; the A380 is $60-$120 on ebay.
Note that some tinkering may be required for modern cards on old systems.
- A UEFI DXE driver to enable Resizable BAR on systems which don't support it officially. This provides performance benefits and is even required for Intel Arc GPUs to function optimally.
The reality is that advertisers will be able to inject their products into the LLMs through manufactured results, prompt engineering and possibly long term deals integrating training data for their brand and product lines.
I feel like hallucinations have changed over time from factual errors randomly shoehorned into the middle of sentences to the LLMs confidently telling you they are right and even provide their own reasoning to back up their claims, which most of the time are references that don't exist.
I recently tasked Claude with reviewing a page of documentation for a framework and writing a fairly simple method using the framework. It spit out some great-looking code but sadly it completely made up an entire stack of functionality that the framework doesn't support.
The conventions even matched the rest of the framework, so it looked kosher and I had to do some searching to see if Claude had referenced an outdated or beta version of the docs. It hadn't - it just hallucinated the funcionality completely.
When I pointed that out, Claude quickly went down a rabbit-hole of writing some very bad code and trying to do some very unconventional things (modifying configuration code in a different part of the project that was not needed for the task at hand) to accomplish the goal. It was almost as if it were embarrassed and trying to rush toward an acceptable answer.
I've seen it so this too. I had it keeping a running tally over many turns and occasionally it would say something like: "... bringing the total to 304.. 306, no 303. Haha, just kidding I know it's really 310." With the last number being the right one. I'm curious if it's an organic behavior or a taught one. It could be self learned through reinforcement learning, a way to correct itself since it doesn't have access to a backspace key.
I use incus to pass a containerized kali os the Wayland and x11 sockets, and whatever else maybe in the /run/user/1000 folder and x11 socket folder, like pipewire. It isn't perfect, but it's really nice spawning a shell/bar/etc inside the container and it goes over the current Wayland desktop. Then I am able to use it to spawn other graphical apps. It works really well. Incus is amazing, or lxc and wayland in general.
I still have the og steam controller, three of them in fact. They still work but I lost the dongle and rely on Bluetooth. It was an experience.
Definitely won't consider buying a controller from them again. The Xbox controller is perfect. Simple and good enough to use.
Similar yet opposite experience here. I had 3 Steam controllers, 2 of them started having issues after a few months. However, I really enjoyed them for certain games, and even played through most of the Dark Souls franchise with them. Definitely excited for the next generation of them and personally will be preordering one as soon as I can.
As far as I can tell, Linux will remain not targeted by attempts to sponge off all kinds of user data. Which makes me so happy that I finally made the leap.
From what I heard it's only not on Linux yet because they've had some serious crashes due to incompatibility with Wayland. Don't worry, it'll come to Linux in time.
I asked GPT 5 to spell out the individual letters of strawberry or blueberry. It did it correctly by essentially putting a space char in between the letters.
Then I simply asked it to count all unique letters in the word. GPT 5 still got it completely correct without thinking.
Lastly I asked how many r(or b) is in the word. This one for some reason switched to GPT 5 thinking with few seconds of reasoning. It out put the correct number.
I guess starting the conversation by painstakingly walking it over to the correct answer helps it out. Idk it's a silly test
I believe that is exactly the downside of using speculative decoding, which is why it is very important to have the models properly sized between each other by making sure the small use is big enough to be mostly correct while also being exceptionally faster than the larger one. However the larger one has to be fast enough that catching flaws won't introduce too manyrandom delays. Also, if the small one is incorrect then the larger one correcting the mistake is miles better than leaving in incorrect output.
It is about improving quality while allowing for faster speed most of the time. The tradeoff is that you consume more memory from having two models loaded vs one of them exclusively.
If you just focus on one then it would make sense to reduce memory usage by just running the smaller model.
Another caveat with this method is that both larger and smaller models need to behave very similar because a lot of the savings come from generating the necessary fluff around each detail such as grammar, formatting and words/letters that transition between each other.
Unsurprisingly gpt-oss has both larger and smaller models that work very similarly! Both model sizes are so similar that even if getting a few wrong would not be slowing down the performance enough to equal the speed of the larger model(which is the worst case with this setup). We want the speed of the smaller model as much as possible. That is all
Personally, I think bigger companies should be more proactive and work with some of the popular inference engine software devs with getting their special snowflake LLM to work before it gets released. I guess it is all very much experimental at the end of the day. Those devs are putting in God's work for us to use on our budget friendly hardware choices.
This is a good take, actually. GPT-OSS is not much of a snowflake (judging by the model's architecture card at least) but TRT-LLM treats every model like that - there is too much hardcode - which makes it very difficult to just use it out-of-the-box for the hottest SotA thing.
Yeah, according to the architecture it doesn't seem like a snowflake, but they also decided to invent a new prompting/conversation format (https://github.com/openai/harmony) which definitely makes it a bit of a snowflake today, can't just use what worked a couple of days ago, but everyone needs to add proper support for it.
SMEs are starting to want local LLMs and it's a nightmare to figure what hardware would work for what models. I am asking devs in my hometown to literally visit their installs to figure combos that work.
Some are asking that yeah but I haven't run an install yet, I am documenting the process. This is a last resort, hosting on European cloud is more efficient but some companies don't even want to hear about cloud hosting.
