Introduction

Galax GeForce RTX 5080 1-Click OC is the company's custom design graphics card priced at NVIDIA's MSRP price for the RTX 5080 of $999. It's the only custom design card we have at MSRP, and hence joins the RTX 5080 Founders Edition for our review coverage. We have a much larger number of custom design RTX 5080 graphics cards to review, but those are priced above MSRP, and we'll get to talk more about them tomorrow. Galax RTX 5080 1-Click OC runs at NVIDIA reference clock speeds, but doesn't miss out on several goodies you expect form custom design cards priced above it, such as a decent RGB LED lighting setup, a 3D metal backplate, a fairly heavy cooling solution with RGB-illuminated axial-flow fans, and a stand that counteracts PCB bending over time—stuff you expect from premium custom designs. If that's not enough, the Galax Xtreme Tuner Plus app also unlocks a software-based overclock of 2640 MHz (versus 2625 MHz reference).



The GeForce RTX 5080 is the second enthusiast-segment graphics card from the RTX 50-series Blackwell generation. It is coming in at half the price of the RTX 5090 launched last week, but addresses the same segment of the market that looks to play the latest AAA games at 4K Ultra HD resolution, with maxed-out settings and ray tracing enabled. While the previous generation RTX 4080 trailed the RTX 4090 with a price-gap of just $300, the RTX 5080 is a whole $1,000 cheaper than the RTX 5090, and $200 cheaper than the RTX 4080 at launch. Its price aligns with the launch price of the RTX 4080 SUPER. Another similarity is that the RTX 5080 maxes out the silicon it is based on, just like the RTX 4080 SUPER did.

The RTX 5080 debuts NVIDIA's second silicon implementing the GeForce Blackwell graphics architecture, the GB203, which it maxes out by enabling all 84 of its streaming multiprocessors across 7 GPCs. The Blackwell generation of GPUs are built on the exact same TSMC 4N process node as the Ada Lovelace generation. This is an NVIDIA-codeveloped variant of the TSMC N5 (5 nm EUV) node. As you'll find in the next page, the GB203 has very similar physical specs to the AD103 chip, albeit implementing a brand-new architecture with new power management technology to boot.

Since the RTX 5080 maxes out the GB203 silicon, it gets 10,752 CUDA cores across those 84 SM, besides 336 Tensor cores, 84 RT cores, 336 TMUs, and 112 ROPs. The memory bus width is 256-bit, and uses the latest GDDR7 memory standard. NVIDIA runs the memory at 30 Gbps, yielding 960 GB/s of memory bandwidth, which is a 34% increase in bandwidth over the RTX 4080 with its 22.5 Gbps GDDR6X. Significant memory bandwidth increases are a common theme across the RTX 50-series, this is needed because AI plays a bigger role in the graphics rendering stack as you'll soon find out, and AI is a memory-sensitive workload.

The GeForce Blackwell graphics architecture lays hardware-level preparation for neural rendering, a revolutionary new concept in 3D graphics, where generative AI plays a closer, more participatory role in the core rendering stack, and not just part of the DLSS super resolution feature, where it helps reconstruct details in upscaled frames. Just as NVIDIA discovered a way to combine real-time ray traced elements with classic raster 3D, it found a way to combine objects created by a generative AI with raster 3D. It worked with Microsoft to standardize this in the DirectX 12 API, letting 3D applications directly access Tensor cores. The GPU also runs generative AI and 3D graphics rendering workloads in tandem thanks to a new hardware scheduler component called the AI Management Processor (AMP).

Besides neural rendering, Blackwell also introduces DLSS 4 and Multi Frame Generation. DLSS 4 replaces the CNN (convoluted neural network) based AI models driving Super Resolution, Ray Reconstruction, and Frame Generation, with a new transformer based AI model that improves image quality at every performance tier. The updated AI models are available to even the RTX 40-series and RTX 30-series GPUs in games that implement DLSS 4, however Multi Frame Generation (MFG) is exclusive to the RTX 50-series. MFG is a technology that leverages AI to draw not just the frame succeeding a conventionally rendered frame, but up to three succeeding frames. It requires the new Flip Metering hardware capability of the Blackwell display engine.

Blackwell sees major updates to the display engine and the GPU's media acceleration engines. The display engine comes with support for DisplayPort 2.1b with UHBR20, enabling 8K high refresh-rates with a single cable with DSC, or 4K 240 Hz with DSC, both with 12 bpc HDR. It also puts out HDMI 2.1a. The hardware flip metering component of the display engine plays a role in DLSS 4 Multi Frame Generation. Meanwhile, the NVENC (encoders) and NVDEC (decoders) are updated with support for 4:2:2 video formats, and accelerate AV1 and HEVC encoding and decoding. Blackwell is the first generation of GPUs to implement PCI-Express 5.0 x16 host interface, ATX 3.1 and PCIe 5 CEM power architecture, and the 12V2x6 power connector.

The Galax GeForce RTX 5080 1-Click OC is designed to meet NVIDIA SFF-Ready logo requirements. The card measures 303 mm x 125 mm x 49.7 mm, which makes it friendly with some of the more compact mid-towers and SFF cases. It's not as compact as most of the other custom design RTX 5080 cards we are testing. The card's Xtreme Cooling System (XCS) cooling solution combines a fairly large aluminium fin-stack heatsink with a trio of 92 mm RGB-illuminated axial-flow fans. The card sticks to NVIDIA reference clock speeds of 2617 MHz boost, and 30 Gbps (GDDR7 effective) memory. It is priced at NVIDIA's baseline price of $999, which is why we get to present this review today.