Laptop review · Snapdragon X2 Elite · Windows on ARM

ASUS A16 X2 Elite: 48GB and OLED worth another look.

Strong CPU results, a large OLED and software caveats. My two-week A16 review, with game excerpts and a dated Best Buy offer for the 48GB model.

Alan Nafarrete · Blendlogic TechVideo: 2026-04-22Article updated September 27, 2026

Offer spotted September 27, 2026

A16 price drop: $1,369.99

Best Buy showed $530 savings against its $1,899.99 “Comp. Value.”

48GB RAM · 512GB SSD · 16-inch 3K OLED touchscreen
Snapdragon X2 Elite Extreme · SKU 6674414

If you have been waiting for an A16 discount, this is worth a look. Best Buy displayed roughly eight hours remaining when checked on September 27. Check the retailer’s current price, stock and countdown before ordering.

The lower price does not erase my Windows-on-ARM caveats. I returned my review unit for my own software and gaming needs; read the testing below and check your must-have apps. This offer is specifically the 48GB/512GB configuration.

Check Best Buy price ↗

Affiliate link: I may earn a commission if you buy through it, at no extra cost to you.

Price snapshot: September 27, 2026. The sale can change or end; this article does not reserve the offer.

ASUS Zenbook A16 with Kingdom Come: Deliverance II on screen — thumbnail from my April 22 two-week review.

ASUS Zenbook A16 with Kingdom Come: Deliverance II on screen — thumbnail from my April 22 two-week review.

2 modesCPU chart keeps power modes separate

2 weeksThe published daily-use follow-up

20.3 tok/sNPU table value · description disagrees

01

The question was whether it could replace my daily machine.

The ASUS Zenbook A16 gave me the large, light Windows laptop I wanted to try. My April 9 first look and April 22 follow-up reached a less tidy conclusion: the CPU scores were promising, individual games could run well, and the experience still did not replace my Mac for the work and games I used.

The machine in these videos

ASUS Zenbook A16
Snapdragon X2 Elite family
16-inch Windows-on-ARM laptop

The evidence window

First look: April 9, 2026
Two-week follow-up: April 22, 2026
Game behavior and software from those dates

These are historical tests of this unit. Windows game compatibility, drivers and translation software can change. The footage does not establish a current compatibility list or a matched cross-platform FPS ranking.

02

A CPU result needs its power-mode label.

The first video’s CPU chart includes two X2 power modes. The single-core result changes much more than the multi-core result. That is useful context before quoting one number as “the X2 score.”

X2 Elite · Geekbench single-core

Geekbench 6 score · Higher is better

Explore a result

Best Performance · published chart3,692

Balanced · published chart1,983

Bars start at zero. Scale resets for each selected metric.

The selected power mode belongs beside the score.

April 9 video, on-screen comparison at 0:58. These are the two X2 rows on that card. Run count, temperature and power connection are not specified there. A later individual result screen has different numbers; it is not substituted into these rows.

All measurements and test conditions

Published results and their limits
Test	Results	Unit and conditions
X2 Elite · Geekbench single-core	Best Performance · published chart: 3,692 Balanced · published chart: 1,983	Geekbench 6 score April 9 video, on-screen comparison at 0:58. These are the two X2 rows on that card. Run count, temperature and power connection are not specified there. A later individual result screen has different numbers; it is not substituted into these rows.
X2 Elite · Geekbench multi-core	Best Performance · published chart: 22,635 Balanced · published chart: 18,722	Geekbench 6 score Same published comparison card at 0:58: 22,635 and 18,722. This is a pair of recorded results, not proof that every machine will change by the same amount. A CPU score does not measure game compatibility or frame pacing.

Inspect the original CPU comparison card

The original Geekbench chart, including X2 Best Performance and Balanced rows.

This article plots only the two X2 rows. Other machines on the original card were not tested under a documented common condition set. Original card at 0:58, with audio →

03

Inspect the games that drove the decision.

The Dota sequence is why I did not let the CPU chart settle the gaming verdict. I tested an effects-heavy scene and compared how the machines behaved. The clip lets you see that session; it is not a substitute for a synchronized frame-time capture.

Dota 2 in the first-look testing sequenceThe filmed display shows an effects-heavy in-game scene. The original comparison moves between machines; this excerpt is observational footage, not a settings-matched performance measurement. Silent passage from my published video, at its original speed. Press play to watch.

The later video also shows Kingdom Come: Deliverance 2 running on the A16. The live overlay changes throughout the sequence. I have not converted those changing counter readings into an average or a 1% low.

Kingdom Come in the two-week follow-upThe A16 displays a moving game scene with a live overlay. The chapter calls this “Native ARM”; the executable architecture was not independently established, so the article treats it as demonstrated gameplay on this Windows ARM machine. Silent passage from my published video, at its original speed. Press play to watch.

War Thunder appears later in the same follow-up. Working examples matter, but they do not establish that another game, anti-cheat system or translation path will work equally well.

04

The NPU result needs the model, not just a speed.

The follow-up includes an NPU language-model result. The published table shows 20.3 tokens per second for the X2 Elite NPU running Llama 3.1 8B, with a 4.7GB model and 8K context noted on the card.

20.3 tok/sPublished X2 NPU row · Llama 3.1 8B · 4.7GB · 8K context

The actual local-AI table from the follow-upThe table also includes different Gemma models on other machines, with different context lengths. Those rows are not a matched hardware ranking. Silent passage from my published video, at its original speed. Press play to watch.

The source disagrees with itself here. The video description’s chapter label says 15.16 tokens/s. The visible table says 20.3. This article identifies the on-screen value and its model, rather than averaging the two or treating them as the same run.

05

Why I returned it after two weeks.

After two weeks, I chose to return the A16 and keep using a Mac for my needs. That is a fit decision based on the games, software and ecosystem in these videos. It does not mean the X2 lacks CPU performance, or that the demonstrated working games do not count.

The original prices and price gap belong to April. I would reassess a new purchase against current software behavior and current listings, rather than promote that old price comparison as a live deal.

What would change this verdict?

A game update, a driver or translation improvement, and a different set of required applications could change the fit. A matched replay and raw frame-time capture would strengthen the gaming comparison. A repeat NPU run with the exact model and context would resolve the conflicting AI numbers.

Watch the complete review

The short clips above support individual findings. The full video includes narration and the rest of the testing.

Open the full review on YouTube ↗ · Subscribe for the next review ↗

Watch the earlier A16 first look →

Published-source snapshots: x2-first · x2-followup. Chart values and conditions.

The rest of this laptop cycle

The Acer versus MSI memory comparison →