ASUS ProArt P16 RTX 5090 review: big projects, small footprint

Which ProArt P16 this is

This review covers the H7606WX-XH99T, the factory build listed in the ASUS US configuration table with a Ryzen AI 9 HX 370, an RTX 5090 Laptop GPU, 64GB of memory, two 2TB SSDs, a 3840x2400 120Hz tandem OLED touchscreen and Windows 11 Pro. It is a late-2025 machine being bought in late 2026. Older P16 models used a 60Hz panel. If a listing shows a different model code, assume different specs until proven otherwise. Newer does not automatically mean better for your workflow.

The H7607 RTX Spark models share the ProArt name but are separate machines built on an Arm platform with shared memory, rather than the dedicated graphics memory in this laptop.

NVIDIA's launch materials describe native Photoshop and Premiere Pro as betas, with native Substance support to follow. If Spark tempts you, verify every plugin you depend on first. The rest of this review covers the x86 RTX 5090 model.

The 24GB question: capacity, compute and your real bottleneck

The RTX 5090 Laptop GPU has 24GB of GDDR7 against 16GB on the RTX 5080 Laptop, according to NVIDIA's spec listing. Think of memory as the size of a workbench and compute as how fast the person at it works. The two chips differ in both, so the 5090 is not simply a 5080 with a bigger bench. The memory gap, though, is the clearest and most predictable difference, and it is the strongest reason to pay for this configuration.

When a scene, texture set or local model does not fit in GPU memory, what happens depends on the software. Blender's manual, for example, says its CUDA, OptiX, HIP and Metal backends can try to use system memory when the GPU fills, with a performance cost. Other tools may slow down, throw an error or refuse to load a model. More memory can turn those cases into clean runs. Two cautions: a project that overflows 16GB will not automatically fit in 24GB, and a meter showing high GPU memory use is not proof on its own, because many apps reserve or cache memory they do not strictly need. Look for reproducible slowdowns or errors on the same project instead.

Below 16GB, the case rests on compute, and that is less predictable. We have no matched creator benchmarks against an RTX 5080 laptop, so we will not promise a percentage.

4K video is the common overreach. Editing has separate bottlenecks: decoding camera files leans on media engines and the CPU, real-time effects lean on the GPU, and exports depend on codec and settings. A plain 4K cut can run well on far less graphics memory, while heavy noise reduction, stacked GPU effects and AI tools are where capacity starts to bite. Bottlenecks also coexist: a project can be short on GPU memory and limited by the CPU or storage at the same time.

Judge by your worst project as well as your average one. A regular job hitting a GPU-memory wall is a practical reason to compare extra capacity, then verify that the job actually fits and benefits. It does not make 24GB automatically sufficient or the P16 the winner. The Blade 16 also offers 24GB, so compare power, screen, service and price. If memory is not the bottleneck, compare compute performance and the screen instead.

Memory and storage: what is fixed and what is not

The 64GB of LPDDR5X is mounted on the board, according to the ASUS technical specifications, so it cannot be upgraded. That is comfortable for most editing, photo and 3D work, but think about the next few years of projects. If you already exceed 64GB on a desktop, a bigger GPU will not fix it. The 24GB of graphics memory is separate. It does not combine with system RAM into an 88GB pool, and running short of either can slow or fail a job.

Storage is two 2TB SSDs filling both slots. The primary link is PCIe 4.0 x4 and the secondary is x2, a narrower path that can cap peak throughput but does not mean every application runs at half speed on that drive. Cache traffic can be part of a bottleneck, so let actual drive performance and your project's file activity decide where active footage and caches live, rather than assuming either slot is the slow one. Archives are the easy low-demand fit.

Do not assume a particular SSD brand or identical drives. Fast external storage is an option for big libraries if it suits how you work. And two drives are not a backup: keep an off-device copy of client work.

The 4K touch OLED: when it earns its place

The 16-inch 3840x2400 tandem OLED is the other half of what you are paying for. The current ASUS specification lists an anti-reflection treatment and stylus support. Anti-reflection reduces glare without eliminating it, so check the screen in your usual workspace, especially near windows. ASUS quotes a 1600-nit HDR peak. Read that as a manufacturer figure for HDR highlights, not a promise of full-screen or everyday desktop brightness.

Resolution helps less than it sounds at this size: with Windows scaling, the extra pixels mostly mean sharper text and finer detail, not more workspace. The 120Hz refresh smooths scrolling and scrubbing but does nothing for export times. Stylus support is listed, but this is a clamshell, not a drawing tablet, so check pen comfort for your own tasks before relying on it, and do not assume a pen is in the box.

A bright, wide-gamut panel is not automatically a reference monitor. If you deliver HDR or color-critical work, color management, the target gamut and the delivery context matter more than how vivid the panel looks. Check exports on the kind of screen your audience actually uses.

The value turns on where you work. Docked to a large external monitor most of the day, you rarely see the premium panel. On location, in suites without spare displays, or anywhere the laptop is the screen, it earns its share of the price.

Power, noise, battery and buying: the ownership costs

Notebookcheck tested an ASUS-supplied HX 370/RTX 5090 P16 in October 2025, not the modified Amazon offer. GPU limits were 120W (Performance), 105W (Standard) and 55W (Whisper); fan noise ranged from about 42 to 53 dBA. Browsing lasted 8h23 at 150 nits and 5h12 at full brightness. Its 1.95kg body came with a roughly 730g adapter. These are browsing results, not unplugged rendering estimates; try Standard first and compare your actual project.

Against the same-chip 2025 Razer Blade 16, PugetBench Photoshop Overall 1.01 scores were 8,893 (P16) and 8,879 (Blade); Premiere Pro Overall 1.1 scored 11,107 and 12,662. That is a near-tie and a Blade lead in those aggregate tests, not export timings or a universal application ranking. Neither predicts the 2026 Blade, the modified P16, or every project. Choose by your software and test your own demanding files.

Ultrabookreview reached a more cautious picture from a pre-production sample tested in October 2025, with the page later updated. The reviewer flagged a possible CPU cooling issue, and an early battery reading of around three hours was revised to roughly five to nine hours of browsing after a Windows reset, at 120 nits and 60Hz. We cannot say what caused the change. The useful lesson is about method: sample status, software state and settings can move results a lot. It does not prove retail units run hot, and it is not a reason to ignore the concern. Watch temperatures and fan behavior on your own unit.

Ports include full-size SD Express, HDMI 2.1, two USB-A, one 10Gbps USB-C and one 40Gbps USB4. Confirm your dock drives your displays and charges as expected. A USB-C dock may top up the battery under light load, but do not assume it replaces the 240W adapter for full performance.

CDW listed this exact factory configuration at $5,718.99 when we checked on October 21, 2026. That is a dated snapshot, not a stock or checkout guarantee. When comparing listings, match 64GB, two 2TB drives and Windows 11 Pro, then compare warranty, service and return terms.

Start with untouched factory stock from sellers on ASUS's shopping list. Some listings are retailer-modified units, and a modification is not evidence of better performance, so pay extra only for support or return terms you actually value.