No? You can run any model that fits in its VRAM, and you can run larger models with layer/MoE offloading. Ask an AI what the best models you can run on that card are, then ask it for newer models than that. Ask what tuning options to pass to llama.cpp, and what the auto-tuning options are. Use ROCm builds.
It looks like your card has 16GB VRAM? Start with Qwen 3.5 9B Unsloth GGUFs (UD-Q6_K_XL) and branch out from there.
No, but yes? OmniCoder 9B at Q6 fits on my 9070 XT with 200k+ tokens of context, and it works pretty well with OpenCode. It is for sure the best local model that I've managed to squeeze onto my GPU, and it even works at 120k context at Q3 on an 8GB RX 580 GPU.
I can't imagine trying to using this model on either GPU for real work. I can use much bigger and faster models on the $3 Chutes subscription or $10 OpenCode Go subscription.
Even so, I am still excited. I don't feel like there was even a model worth using with a tool like OpenCode 6 to 9 months ago. I like the way things are heading, and I am looking forward to seeing how capable coding models of this size are in another 6 to 9 months!
You can cram absurd context into a card now, but none of that matter once you hit the VRAM wall and the whole thing slows to a crawl. Cloud is cheaper. Local still matters for privacy and weird adapter stuff, but 'usable for work' is a much higher bar than 'looks decent on benchmarks' when the task is chewing through a repo without latency going to hell.
Well, this specific solution was only set up on specific hardware, and is Nvidia dependent, as the readme stares.
That doesn’t mean the 9070XT can’t do AI stuff, quite the opposite. ROCm gets better all the time. There are many AI workloads you can do on AMD cards.
Is it a card I would choose if I was primarily working on AI? Absolutely not. But it is the card I own and it’s been a great value for gaming.
Unfortunately AMD is much worse with supporting AI features like FSR4 on older hardware generations, despite the capability and leaked INT8 models being there. Totally unlike NVIDIA.
It’s absurd I have to use open source programs to get INT8 FSR4 support.
There are two streams, video and audio are separate files. Thats the first hurdle stopping browser from presenting you with a simple "download source" button like on normal Video elements.
Seconded (& I recommended them to the author of the piece, who didn't listen. :-D )
I have an Umidigi F2, with a 38MP camera, 6GB RAM, 128GB flash, big retina screen -- I neither know nor care the rez -- and a huge battery. It lasted 3 days of intensive use when new, it has a headphone socket and both dual SIM + µSD card slots, and it was an astonishing €125 new.
And a totally vanilla Android 10 install that even says it's a Pixel in a couple of places.
I'd rather have a ~£150 phone I can replace every 2 years than a ~£300 phone that has to last ~4 years, but others' mileage varies, of course. Including David's.
This is my 3rd cheapo Chinese phone and I've been very happy with all of them.
The 2nd was a PPTV King 7, AFAIK never sold outside of China, so even when set to English the ROM was partly in Chinese. This one I reflashed with CyanogenMod and it worked well. My camera no longer focused but that was about it. For its time it was _stunning_ and it's still in use by a friend of mine about 5Y after I bought it second-hand.
It's incredible how all phones are coming along these days. It seems to be easier for other companies to get into phone manufacturing, I just hope more of them get into FOSS.
I know this! It depends on the antibody test - if you can get a test for the nucleocapsid antibodies, that will determine whether you had the disease or not. People who are vaccinated will test positive for the spike protein, but will not test positive for the nucleocapsid antibodies, at least with the vaccines currently available (I believe there are vaccines in development that target both proteins, but not yet available).
Some antibody tests target spike or nucleocapsid, some target both, so you need to check which tests they are using and verify what it tests against (or ask your doctor to order a specific test that checks for the nucleocapsid antibodies).
I learned this when I was part of a vaccine study and was curious to know whether I had gotten the vaccine or the placebo, and was able to use a spike antibody test to confirm that I was in the vaccine group.
I wonder if it’s theoretically possible to construct a historical timeline of immunity development for an individual by taking multiple tests, sounds like a proper forensics work than something to do just out of curiosity.
SSH, and also when you're on the road and want to write a simple Python script to process something, or do something with your sensor data logging. Termux has a Python API to access sensor data, it has numpy, it has requests, so you can do a lot.
For me ssh to access my main Emacs session is a big part of it but I also run some shell scripts and CLI tools and services written in Go. ssh-ing back into the phone for file transfer is another important use.
reply