AMD Strix Halo with 32 GB RAM for $1899 - Asus ProArt PX13 2026 Convertible Review

[Redaktion]

The base model of the compact Asus ProArt PX13 convertible is equipped with the AMD Ryzen AI MAX+ 388 and the powerful Radeon 8060S graphics. You also get the familiar 2.8K OLED screen, 32 GB RAM and a 512 GB SSD for $1899.

https://www.notebookcheck.net/AMD-Strix-Halo-with-32-GB-RAM-for-1899-Asus-ProArt-PX13-2026-Convertible-Review.1352377.0.html

[Limited AI use]

Almost no matter how I look at it: Attaching only 32 GB RAM to the rather big and expensive 256-bit Strix Halo APU, that is supporting up to 128 GB RAM, is waste of silicon.

demanding games that can have issues with comparable dedicated GPUs with just 8 GB VRAM.

Fair argument, but still such a waste of silicon.

Why is AIDA64 memory read/write only measuring 61859 MB/s? This is the speed of a 128-bit APU, running 5600 MT/s RAM speed, not a 256-bit APU, running 8000 MT/s RAM speed. It should at least be roughly double. (Actually more like 200,000 MB/s (=256-bit * 8000 MT/s / 1000 / 8 * 0.65-0.85 [theoretical ->practical]), but AIDA64 shows a different value.)
The low memory bandwidth score does not affect the 3DMark scores, so not sure what's going on.

[Limited AI use]

If you care about using this laptop for AI extensively, know that with current SOTA AI LLM Qwen3.6-35B-A3B model and its lowest usable 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 here: notebookchat.com/index.php?topic=315954.0 ("Your own ChatGPT, offline: AI without the cloud on your laptop") and comments.
(PS: On Strix Halo on Windows, only 75% of the RAM can be freely allocated to the iGPU for fitting an AI LLM model + its context. In this case, it's fine in the sense that the OS needs 8 GB for itself anyway. But on a 128 GB RAM config, that's only 96 GB for fitting AI model(s) and a whole 32 GB for the OS itself, which is a bit too much.)

[Nobody cares about AI]

Silent profile - 50w/40w - TimeSpy Gfx 6279 - CP2077 48.2 fps

Should be more clear about wattage. This isn't SoC power which was more likely operating closer to 30w-35w(?) but total device consumption here.

In any case, it just shows how far intel arc gfx has progressed when they're doing 40 FPS @ 25w and just 1-2 FPS slower @ 35w.

As it was not mentioned in this review, people interested in this laptop should be aware of this potential issue:
youtu.be/AzBjtmJCbhE?si=bZfxWb8gp-oCdo3o

Will you guys be reviewing 15ASH11? It's been out for awhile now. When you do, please test at lower wattages too (20w, 25w, 30w, 35w) for gaming - not everyone is interested in hearing how a blow-dryer sounds at 80+ w. :')

[Andreas Osthoff]

The listed TDP values are not total system power.