Introduction

The PNY XLR8 GeForce RTX 4070 Ti OC continues the company’s hands-on foray into the custom-design graphics card market in the post-EVGA world, where the mainly creator-focused brand looks to gain market share in the gaming segment. Past generations of PNY XLR8 graphics cards had a distinct Gainward/Palit imprint, but with the GeForce RTX 40-series “Ada,” the company is taking direct charge of product design. The NVIDIA GeForce RTX 4070 Ti is the company’s latest high-performance graphics card that seeks to strike a great balance between performance and value at a starting price of just $800.

NVIDIA GeForce RTX 4070 Ti has a bit of short-term history attached to its creation. When NVIDIA debuted the GeForce RTX 40-series with the RTX 4090 flagship, at the same time they announced two RTX 4080 SKUs separated by memory size—RTX 4080 16 GB, and RTX 4080 12 GB, except that memory size is but a minor differentiator between the two SKUs. The RTX 4080 12 GB looked like a vastly different product from its 16 GB sibling, based on a physically smaller silicon, with 21% fewer shaders, RT cores, and Tensor cores, 25% lesser memory, and 25% narrower memory bus; at a starting price of $900. This caused quite some controversy on social media and the press; forcing NVIDIA to cancel the launch of the RTX 4080 12 GB, and re-brand it as the RTX 4070 Ti, which the company is finally launching today, at a slightly trimmed price. Even at its new price of $800, the GeForce RTX 4070 Ti still commands a higher MSRP than that of the original GeForce RTX 3080 of $700, which was obscured by the GPU supply chaos of 2021.

The GeForce RTX 4070 Ti debuts NVIDIA’s third-largest silicon based on its “Ada Lovelace” graphics architecture, the 4 nm “AD104.” This silicon is maxed out by the RTX 4070 Ti, enabling all 7,680 CUDA cores across 60 streaming multiprocessors (SM) physically present on it, besides 60 RT cores, 240 Tensor cores, 240 TMUs, and 80 ROPs. The chip physically features a 192-bit wide GDDR6X memory interface, which holds 12 GB of memory running at 21 Gbps, which works out to 504 GB/s of memory bandwidth, which is generationally less than not just RTX 3080 (760 GB/s), but also the RTX 3070 Ti (608 GB/s). NVIDIA introduced several architecture-level memory sub-system improvements, such as larger on-die caches, to overcome performance deficits arising from lower memory bandwidth. Even though its TGP is under 300 W, which could easily be handled by even a classic 6+8 configuration, NVIDIA is introducing the 16-pin ATX 12VHPWR connector to this segment. All custom-design graphics cards are expected to include an NVIDIA-designed adapter that converts two 8-pin PCIe power connectors (2x 150 W) to 16-pin.

The PNY XLR8 GeForce RTX 4070 Ti OC is the company’s top custom-design product based on this GPU, which it combines with a high-performance cooling solution that looks surplus to the requirements of cooling a GPU with a typical graphics power of 285 W. The company has given the RTX 4070 Ti a factory overclock of 2670 MHz GPU Boost (compared to 2610 MHz reference), while leaving the memory untouched at 21 Gbps (GDDR6X effective). PNY is pricing the card at $830, a slight premium over the $800 NVIDIA MSRP.