Introduction

The EVGA GeForce RTX 3090 Ti FTW3 Ultra is the company's premium air-cooled rendition of the newest big kid on the block. The RTX 3090 Ti is the crescendo of the "Ampere" graphics architecture, a have-it-all SKU. It is designed to be a double-digit percentage faster than the RTX 3090, the de facto flagship for the RTX 30-series until today. The RTX 3090 Ti is designed by enabling all 84 streaming multiprocessors (SM) on the GA102 silicon, dialing up clock speeds, and using even faster memory. The design goal is maxed-out gaming at native-4K resolution with ray tracing, as well as several creative use cases that could benefit from the large 24 GB of fast memory.



The GeForce RTX 3090 Ti maxes out the GA102 silicon, enabling all 10,752 CUDA cores, 336 Tensor cores, and 84 RT cores. That's not all, as NVIDIA is introducing new 21 Gbps-rated GDDR6X memory with the RTX 3090 Ti, as opposed to the 19.5 Gbps memory on the RTX 3090, or the 19 Gbps on the RTX 3080. This pushes memory bandwidth past the 1 TB/s mark, compared to the 940 GB/s on the RTX 3090. Memory size on both cards is 24 GB.

NVIDIA is first-to-market with graphics cards that feature the new ATX 16-pin power connector designed for the next-generation of graphics cards. The 16-pin connector is capable of delivering up to 600 W of power, and it was needed as the typical board power for the RTX 3090 Ti is rated at 450 W, which is 100 W higher than the RTX 3090. Something like this would have required three 8-pin PCIe power connectors on cards with even close-to-reference specs, hence the switch to the new connector. NVIDIA understands that the new connector is very new to the market, so all board partners are made to include an NVIDIA-designed adapter that converts three 8-pin PCIe connectors to one 16-pin.

The GeForce Ampere graphics architecture introduces the second generation of NVIDIA's path-breaking RTX technology, which fuses real-time ray traced elements with traditional raster graphics. These include almost physically accurate lighting, reflections, lensing, illumination, and motion-blur. The Ampere architecture introduces new 2nd-generation RT cores which provides hardware acceleration to the most computationally demanding parts of ray tracing, including ray-traced motion-blur. The 3rd-generation Tensor core leverages the sparsity phenomenon to increase AI-DNN inference performance by an order of magnitude.

The EVGA RTX 3090 Ti FTW3 Ultra features the company's latest iCX 3 smart cooling solution that combines a powerful heatsink with a trio of individually controlled fans and a network of temperature sensors across the PCB that let the card optimally manage cooling on its own. The PCB is intelligently designed with cutouts that have some of the airflow from the fans pass through, and out vents on the backplate. The card is smartly illuminated with a large ARGB diffuser along the top. Compared to the 1860 MHz reference, EVGA's generous factory-overclock for the card has a 1920 MHz maximum boost frequency.