HN Explorer

5 comments

[ 1.1 ms ] story [ 82.6 ms ] thread
karmakaze
The spec I haven't seen listed on any of these RTX Spark laptops is the memory bandwidth. Seeing as it's listed as having up to 128GB LPDDR5X, I can guess that it's about the same as an AMD Gorgon Halo's 192GB around 273 GB/s which is where I lose interest. Even the M5 Pro at 307GB/s can't compare to discrete GPUs.

Only when we get to M5 Max 614 GB/s and Ultra 1.2TB/s does it get interesting for me.

cyanydeez
I just benchmarked a halogen server running Qwen3.8-Flash-Next:

p50 decode 45.5 t/s

p50 prefill 1278 t/s

I'm able to get these engineering harnesses running for 1-2 hours at a time with the right specs.

What exactly are you looking at that needs more?

karmakaze
It's the utility value I can't see. It makes sense for as much as you can put "into a laptop" and is the same zone as AMD Strix/Gorgon Halo. I'd much rather have the high compute in my server in the basement that I have remote use of rather than burning my battery with limited thermals.

And the Apple Max/Ultra does it better. The Mac Mini/Studio form factor makes way more sense. These laptops look like conspicuous consumption brand marketing devices. CUDA is worthless--AIs can port kernels.

cyanydeez
This is a fedora server running in my basement.

I got it before the memory cartel.

Just curious what youre targeting that BW is limiting you. I churned through models ans harnesses, and settled on this setup. I also have accessto a blackwell running the samw model to 75 tk/s and 1800 prefill.

I dont really see what youre targeting besides paper specs.

karmakaze
For that kind of money I can run 2x AMD R9700 (on a server I already had) and get 150-250 tokens/sec running Qwen3.8-27B MXFP4. That speed makes all the difference as then the work is highly interactive design/development rather than semi-batch agent loops.