PNY GeForce RTX™ 4070 Ti SUPER 16GB VERTO™ Overclocked Triple Fan DLSS 3

The new GeForce RTX 4070 Ti SUPER is part of a mid-lifecycle refresh of the RTX 40-series Ada generation, focusing particularly on the high-end segment. The RTX 4070 Ti SUPER replaces the RTX 4070 Ti from the product stack at its $800 MSRP. You should be able to find RTX 4070 Ti cards at slightly lower prices, but the card has stopped production. The RTX 4070 Ti SUPER comes with significant improvement in specs over the original RTX 4070 Ti, with the biggest change being memory. It now gets 16 GB of memory across a 256-bit memory bus; compared to 12 GB of it across a 192-bit memory bus for the original. This change results in the engine getting fed with a 33% increase in memory bandwidth. Besides this, the card also gets a handy 10% increase in CUDA cores, and more ROPs, although the on-die cache size remains unchanged.

If you recall, NVIDIA had maxed out the AD104 silicon to create this RTX 4070 Ti, which meant that it would have to tap into the larger AD103 silicon found powering the RTX 4080 and mobile RTX 4090 series. The RTX 4070 Ti SUPER is carved out of this chip by enabling 66 out of 80 SM, worth 8,448 CUDA cores, 264 Tensor cores, 66 RT cores, and 264 TMUs. The ROP count is set at 96, out of the 112 present on the silicon. This is still 20% higher than the 80 ROPs that the RTX 4070 Ti has. The GPU runs at 2625 MHz on the PNY Verto OC (compared to 2610 MHz reference); and the memory at 21 Gbps, yielding 672 GB/s of memory bandwidth that’s 33% higher than the other RTX 4070 series cards.

The new Ada Lovelace graphics architecture driving the GeForce RTX 40-series, introduces generational uplifts in performance and energy efficiency, thanks to the new 5 nm EUV foundry process, and SIMD-level innovations. The new Ada CUDA core, in addition to IPC increases and support for higher clock speeds over Ampere; features support for shader execution reordering, a feature that improves ray tracing performance. The new 3rd generation RT core supports displaced micro-meshes, a feature that increases complexity of ray traced objects without a linear increase in performance cost; and the new optical flow accelerator, a component that lets the GPU draw entire alternate frames using AI, without involving the main graphics rendering pipeline, which is needed for DLSS 3 Frame Generation to work. The company has also redesigned the memory sub-system with larger caches on the GPU, so it could narrow the memory interfaces while increasing the memory sizes.