yxc999

Yeah I'm gonna use that from now on, that's hilarious.

Honestly, in that case I would buy one. But otherwise this much processing power is just absolutely unwarranted.

Gen 1 X13 currently go for 100-150 bucks, they are the best value but compiling time is gonna be pretty trash

Well I do want GrapheneOS on it so they aren't an option. I just wonder why noone is capitalizing on a ~200 bucks GrapheneOS phone for people like me who don't need the processing power or even lower-spec models for criminals who literally only have Signal on it. I mean which criminal organization or activist group or whatever wouldn't buy a box full of sub 200 bucks GrapheneOS phones?

Yes, but like half the SoC power, half the RAM wouldn't change anything about how I use the phone. It's just not needed at fucking all, waste of money. I'd rather have even more battery life lmao

Damn I guess most of us didn't know that, thank you for sharing.

Honestly, I'd replace this fuckass Google device in a heartbeat, but I will never in my life understand why people want flagship phones. I got a b-ware 9a for 300 bucks and even this one is way overkill for what I want in a phone. Why can't there just be a low-spec one instead of ones that cost more than an entry class laptop?

cry

I feel like you definitely can get away with as much as even 5, but if you actually look at the CLI, it will tell you how many draft tokens actually got accepted. In some situations like writing simple text, over 90% of my two draft tokens get accepted so setting it to five would have made it faster. However if you run a task which requires heavy reasoning, even two draft tokens start approaching close to only 60%. Considering Qwen3.8-27B overthinks more than an anxious teenage girl, you should generally with two. If you take my example with the quant dropped to UD-Q5_K_XL with 262K tokens and reasoning low for a task that isn't actually complex, just long, then go right ahead and let it run with 5.

Edit: excuse the horrendous grammar, I was distracted

I'm using this on Linux, so the first two lines won't apply to you. Besides that, you are completely right, llama.cpp with the SYCL backend and MTP is the fastest way to run LLMs on Intel Arc GPUs, it's just braindead people and bots on Reddit repeating the same line while they run a NVIDIA GPU and Ollama.

source /opt/intel/oneapi/setvars.sh
/opt/llama.cpp-sycl/bin/llama-server \
  --model /path/to/your/model/Qwen3.8-27B-UD-Q6_K_XL.gguf \
  --device SYCL0 \
  --n-gpu-layers 999 \
  --load-mode none \
  --flash-attn on \
  --jinja \
  --reasoning-preserve \
  --ctx-size 100000 \
  --cache-type-k q8_0 \
  --cache-type-v q8_0 \
  --temp 1.0 \
  --top-p 0.95 \
  --top-k 20 \
  --min-p 0.00 \
  --presence-penalty 0.0 \
  --repeat-penalty 1.0 \
  --spec-type draft-mtp \
  --spec-draft-n-max 2 \
  --port 9931

Huh, my B70 gets peaks of 37 tok/s TG and 800 tok/s PP, what are your settings?

I think there are two sides to this.

1. You are right, I'd rather just get them on Windows 11 Enterprise IoT LTSC than goddamn ChromeOS.

2. ChromeOS will likely lower the bar for switching to a Linux distro so you might consider it a stepping stone. A stepping stone with nails and spiders coming out of it, but still a stepping stone nonetheless.