ASUS ROG Strix SCAR 16 (2025) G635LX / ASUS ROG Zephyrus G16 (2024) GU605MI-QR194X

ASUS's US spec page shows two G635LX configurations: Windows 11 Home, 1×32 GB, 1 TB, and Windows 11 Pro, 2×16 GB, 2 TB. The G635LW with the RTX 5080 Laptop GPU is a separate model.

Intel rates the chip at 55 W base and 160 W maximum turbo power. In this laptop ASUS allocates 55 W to the CPU in Turbo mode and an extra 25 W in Manual mode; its graphic shows up to 80 W for the CPU.

The NPU runs constant AI tasks such as background blur and noise suppression at low power instead of the CPU/GPU.

TOPS is trillions of operations per second the neural accelerator can perform (Int8). 13 TOPS is below the 40 TOPS Microsoft asks for Copilot+ PCs.

175 W is the ceiling including 25 W Dynamic Boost shifted to the GPU when the CPU is lightly loaded; it is not a constant draw in every game. The RTX 5090 Laptop GPU is not the desktop RTX 5090. Graphics memory is soldered.

Dynamic Boost shifts a 20 W share of power to the GPU when CPU load is low, so 105 W is a conditional ceiling rather than a constant level. In the same series RTX 4060 SKUs are listed at 100 W and RTX 4080/4090 SKUs at 115 W. The GPU is soldered and cannot be upgraded.

Advanced Optimus can route the panel straight to the discrete GPU when a game starts without a reboot, and fall back to integrated graphics on the desktop to save battery.

ASUS quotes more than 2,000 local dimming zones and 1,200 nits peak; the peak applies to small HDR highlights, not sustained full-screen brightness. The panel is Pantone Validated.

G-Sync matches the refresh to the game’s frame rate to reduce tearing. It is not a touchscreen. Some SKUs in the same series use an IPS-level panel.

G-SYNC matches the panel refresh to the GPU frame rate to reduce tearing. The spec sheet also lists G-SYNC for external displays connected over Thunderbolt 5.

ASUS says the layer reduces reflections and improves contrast by 4.5x; this is a manufacturer claim.

SO-DIMM modules are not soldered and can be swapped under the bottom cover. The single 32 GB module configuration runs single-channel until the second slot is filled. The product page mentions DDR5-6400 elsewhere; the spec sheet value of 5600 is used here.

Because memory cannot be added later, capacity has to be chosen at purchase.

SSDs are held by tool-less Q-Latch clips. The ASUS product page calls the slots “PCIe Gen 5 ready”, but the spec sheet says PCIe 4.0 and one independent test reported no 5.0 support. Because of nearby components, only single-sided SSDs are supported.

The second M.2 slot allows an additional SSD.

120 Gbit/s is the Thunderbolt 5 one-way Bandwidth Boost rate; the normal bidirectional bandwidth is 80 Gbit/s. There is no card reader. The HDMI version is not stated on the spec sheet.

Both USB-C ports support external displays and charging; there is no Ethernet port.

ASUS notes that Bluetooth features may vary with the OS and drivers.

The third auxiliary fan moves heat from the GPU and VRAM to the heatsinks. The up-to-15 °C drop ASUS quotes for liquid metal comes from its internal test against the previous-generation compound.

These are the manufacturer ceilings under simultaneous full load; in Silent mode the GPU runs at a far lower limit.

ASUS says the mainboard frame comes off with a few screws. An independent test reported that the Wi-Fi module is soldered.

Overstroke registers a press earlier in the key travel. ASUS rates the switches for 20 million presses and lists n-key rollover.

All lighting can be switched off.

The camera has no mechanical shutter.

The Thunderbolt 5 ports accept power delivery; with a lower-power USB-C charger the battery can still drain under gaming load.

Adapter weight not included.

The MUX switch connects the display directly to the NVIDIA GPU, removing the overhead of routing through the integrated GPU in games; Advanced Optimus makes this switch automatically without a restart.

Slash Lighting is a programmable LED strip on the lid, controlled through AURA Sync.

Which adapter is in the box varies by country and model. Full GPU power should not be expected when charging over USB-C.

Accessories may vary by country.