Introduction

Palit GeForce RTX 4070 SUPER JetStream OC is a swanky new overclocked custom design graphics card and draws from a long line of premium boards by the company. This is the company’s top RTX 4070 SUPER product. In the company’s lineup, the JetStream tier is located above the GamingPro OC, but just a notch below GameRock, which the company reserves for enthusiast-class GPUs. The new GeForce RTX 4070 SUPER is part of a three-GPU refresh by NVIDIA for the upper end of its RTX 40-series. The mid lifecycle refresh sees more performance on offer at give price points, there are no new features to be had, since these GPUs are still based on the existing Ada Lovelace graphics architecture. The GeForce RTX 4070 SUPER is recommended by NVIDIA for bleeding edge gaming at 1440p with ray tracing; we’ve seen over the past year’s additions of new games, that even the regular RTX 4070 is capable of 4K Ultra HD, if you’re a little clever with your game settings, or use DLSS, or better yet DLSS 3 Frame Generation, a feature NVIDIA introduced with the RTX 40-series.

The GeForce RTX 4070 SUPER comes in at the same starting price that the original RTX 4070, which it displaces to $550, in line with the idea behind the SUPER series. NVIDIA had left a rather large performance gap between the RTX 4070 and the RTX 4070 Ti, which needed to be filled in the face of competition from the AMD Radeon RX 7800 XT. We hence have the RTX 4070 SUPER, which on paper leans very close to the specs of the RTX 4070 Ti. The RTX 4070, the new RTX 4070 SUPER, and RTX 4070 Ti, are all based on the 5 nm AD104 silicon. The RTX 4070 Ti maxes it out, while the RTX 4070 features just 46 out of 60 streaming multiprocessors or just 76% of the shaders. The RTX 4070 SUPER gets close, feature 56 out of 60 SM, or 93% of the available shaders. In addition, NVIDIA enabled all 80 ROPs, and the entire 48 MB of L2 cache on the silicon; the RTX 4070 had reduced amounts of these.

The underlying technology is still the Ada Lovelace architecture, which introduces a new generation CUDA core that supports shader execution reordering that improves ray tracing performance; the 3rd generation RT core with its increased ray intersection performance and support for displaced micro-meshes that allow game developers to increase geometric complexity of ray traced objects; and the optical flow accelerator, a component needed by DLSS 3 Frame Generation, to draw alternate frames entirely using AI, and without involving the graphics pipeline. Lastly, Ada re-balances the memory sub-system, thanks to large on-die caches, which allow NVIDIA to generationally increase memory sizes without resorting to wider memory interfaces—they’re in fact narrower, but more on that later.

The GeForce RTX 4070 SUPER is carved out of the AD104 silicon by enabling 56 out of 60 SM, which works out to 7,168 CUDA cores, 224 Tensor cores, 56 RT cores, 224 TMUs, and 80 ROPs. The chip’s 192-bit memory interface drives 12 GB of GDDR6X memory running at 21 Gbps, which yields 504 GB/s of memory bandwidth that’s accelerated by the large 48 MB on-die cache.

Palit GeForce RTX 4070 SUPER JetStream OC runs the GPU at a boost frequency of 2640 MHz, a respectable factory overclock, compared to 2475 MHz reference. Palit has also increased the power limit of the card to 245 W out of the box, compared to 220 W reference. The card draws power from a 12VHPWR connector, and a cable that converts two 8-pin PCIe power to one of these, is included. Palit hasn’t shared the pricing of the RTX 4070 SUPER JetStream OC with us, so we are assuming a market price of $650 considering the product on offer. This price will be used in our performance per Dollar calculations, and conclusion.