Introduction

ASUS ROG STRIX GeForce RTX 3090 OC is the flagship implementation of the RTX 3090 Ampere by ASUS. The GeForce RTX 3090 by NVIDIA represents the pinnacle of the GeForce “Ampere” architecture for the client segment. ASUS builds on this technology with a product targeted at premium gaming PC builds and overclockers, as the card offers many conveniences for both kinds of users. The RTX 3090 transcends the barrier between the gaming and professional visualization markets, much like the TITAN brand of graphics cards, especially when combined with NVIDIA’s highly capable Studio drivers. This begins to explain why NVIDIA referred to the RTX 3080 from last week as its “flagship” gaming product.

NVIDIA reshaped the top-end of its GeForce product stack with the RTX 30-series. The RTX 3080 is designed to offer premium 4K UHD gaming with raytracing on, which should cover almost every gaming use-cases, hence the “flagship” badge. It’s been extensively compared to the RTX 2080 Ti. The RTX 3090, on the other hand, is being compared by NVIDIA to the $2,500 TITAN RTX, a “halo” product based on Turing. What’s also new is that both the RTX 3080 and RTX 3090 share a common silicon under the hood, which is the largest based on the GeForce “Ampere” architecture.

The GeForce Ampere architecture uses the 2nd generation of NVIDIA’s path-breaking RTX technology, which introduced real-time raytracing to the gaming segment. NVIDIA perfected a way to combine conventional raster 3D rendering with certain real-time ray-traced elements, such as lighting, shadows, reflections, ambient-occlusion, and global-illumination. Processing even these few elements required NVIDIA develop fixed-function hardware. Ampere combines a new generation CUDA core that doubles FP32 throughput over Turing and performs concurrent RP32+INT32 operations, the company’s 2nd generation RT core, which has double the performance over the previous generation, and hardware to pull off raytraced motion-blur. The 3rd generation tensor core accelerates AI and leverages the sparsity phenomenon in deep-learning neural nets to increase AI inference performance by an order of magnitude. NVIDIA leverages AI for de-noising its raytraced elements and features such as DLSS. Find more details about the architecture in our NVIDIA Ampere Architecture article.

The RTX 3090 isn’t strictly a creator’s card, NVIDIA is also making a bold stab at the 8K frontier. 8K is four times the pixels of 4K and sixteen times that of 1080p, which is an enormous performance and memory demand on any GPU. NVIDIA believes that its DLSS 8K feature can unlock this resolution for gamers by leveraging AI deep-learning to render the game at 1440p and upscaling it to 8K with an AI-based 9X super-sampling algorithm that leverages ground-truth data from NVIDIA that’s been rendered at 16K on its render farms. 8K gaming monitors are still a fair bit away, but TVs with this resolution are hitting shelves. Helping things is NVIDIA’s implementation of HDMI 2.1, which enables 8K HDR 60 Hz using a single cable. Besides 8K gaming, the RTX 3090 also has the muscle to drive e-Sports games at insane refresh rates of 360 Hz at lower resolution, with the NVIDIA Reflex technology working to reduce whole-system latency to give professional e-Sports gamers the edge they need.

As we mentioned earlier, the RTX 3090 is based on the same silicon as the RTX 3080—the 8 nm “GA102.” The RTX 3090 nearly maxes this chip out by enabling all but one TPC (two SMs) on the chip, resulting in a CUDA core count of 10,496, along with 328 Tensor cores, 82 RT cores, 328 TMUs, and 112 ROPs. The RTX 3090 maxes out the 384-bit wide GDDR6X memory interface of the “GA102” and comes with 24 GB of memory. This memory is ticking at the highest memory frequency in NVIDIA’s product stack, 19.5 Gbps, which works out to an astounding 940 GB/s of memory bandwidth,and took expensive HBM2E solutions in the past.

The GeForce RTX 3090 starts at a price of $1,500, which is exactly 50% more than the RTX 2080 Ti, but $1,000 less than the launch price of the TITAN RTX and probably explains the hyphenation. Unlike the TITAN series, the RTX 3090 can be made into custom-design cards by NVIDIA’s board partners. ASUS flexed its vast engineering muscle to come up with its latest-generation ROG STRIX design scheme. Matte black surfaces make way for brushed aluminium and a vast fin-stack cooler, and a swanky new VRM solution keeps the feisty 350-watt GPU fed through three 8-pin power inputs. But wait there’s more—ASUS is including a manual power limit setting of up to 480 W! This is much higher than any other RTX 3090 card out there. We did a special round of testing at the maximum power limit in our RTX 3090 STRIX OC review to see what additional performance can be gained. ASUS has also given the RTX 3090 a factory overclock of 1860 MHz, while the memory is left untouched at 19.5 Gbps. The ROG Strix RTX 3090 OC is priced at $1800. In this review, we take a look at the gaming and overclocking credentials of the ASUS flagship, with impressive results.