Powerful creator tablet with Snapdragon X2 Elite - Asus ProArt PZ14 Review

[Redaktion]

The Asus ProArt PZ14 is a sleek 14-inch tablet with a Snapdragon X2 Elite SoC, a gorgeous 144 Hz OLED touchscreen, stylus support and a great keyboard. It is powerful and extremely versatile, and it might be perfect for creators on the road.

https://www.notebookcheck.net/Powerful-creator-tablet-with-Snapdragon-X2-Elite-Asus-ProArt-PZ14-Review.1359622.0.html

[davidm]

I don't think it's reasonable for "soldered" RAM on a tablet to be the first item on the cons. Is the lack of a replaceable CPU, math co-processor and cache memory a problem? Why not ask for a couple of PCIe 5.0 slots? x16 naturally.
People buying this want compact as possible performance using leading techniques. Something Apple has achieved to great success but x86 heritage is still not that great. It's the entire point of the design (efficiency through integration). It's very unlikely RAM will fail. Upgrading RAM is a separate issue, but going aftermarket is not really going to save anything these days.
Honestly NBC should really try to enter the mid 2020s. Repair-ability is an important modern consideration particularly for batteries, keyboards and ports (things that will definitely wear) and screens, but it's not so much "soldered" RAM, which implies the vendor arbitrarily soldered DIMMs to a board, but in reality it is fully integrated at a low level that does away with layers of connections and transmission distances.

[dry->cookie,wet->cake]

I don't think it's reasonable for "soldered" RAM on a tablet to be the first item on the cons.

Agreed. Someone didn't pay attention.

Since the 32 GB RAM in this tablet are not upgradable: If you care about using this for AI extensively, know that with 32 GB of memory (RAM+VRAM) and current SOTA AI LLM Qwen3.6-35B-A3B model[1], for its class and size (this is like the smallest size where a model become usable at all kind of tasks), and its best bang for the buck, 4-bit quant, you will be restricted to about this amount of context:

.. I've come to the conclusion that (1) 32768 is the biggest context I can get away with in an adequately smart model, and (2) it just ain't enough.

More about running AI locally: notebookchat.com/index.php?topic=315954.0 ("Your own ChatGPT, offline: AI without the cloud on your laptop") and comments.

[1] artificialanalysis.ai/models/open-source?models=qwen3-6-35b-a3b%2Cqwen3-6-27b#artificial-analysis-intelligence-index (as of this writing, there's no other model that outperforms it at same speed and tasks, the dense 27B model is better, but it's much slower) (you can search for other models in the same table)

[x864ever]

please asus release a version next year with a full fat nova lake processor with Intel's highest grade igpu and up the PPI to 250+ and keep everything else the same and this will be a creator's dream device.