Introduction

NVIDIA today launches the GeForce RTX 4070 Ti SUPER, the second in its three part series of new high-end graphics cards bearing the SUPER brand extension. The three form a mid lifecycle refresh for the GeForce RTX 40-series Ada generation, and are developed to offer more performance at existing price points. The new RTX 4070 Ti SUPER comes in at a starting price of $800, which is exactly the same as the RTX 4070 Ti, which the company will retire from its lineup. At its price, the new RTX 4070 Ti SUPER offers more shaders, and other on-die components; but more importantly, comes with a major upgrade to its memory sub-system. Unlike the GeForce RTX 4070 SUPER from last week, there is no NVIDIA Founders Edition graphics card for the RTX 4070 SUPER in existence, and so we have with us the MSI GeForce RTX 4070 Ti SUPER Ventus 3X, a custom design card from MSI that you will find at the NVIDIA MSRP, which covers all the basics, and has everything you need if you’re in the market for an RTX 4070 Ti SUPER. This card was sampled to us by NVIDIA, and comes both at baseline pricing, as well as reference clock speeds.

NVIDIA had originally launched the GeForce RTX 4070 Ti as the “RTX 4080 12 GB,” which it had to rename as the naming caused some controversy owing to its significantly different specs from the then “RTX 4080 16 GB,” which went onto be known simply as the RTX 4080. The company had maxed out all available shaders on the 5 nm AD104 GPU to create the RTX 4070 Ti, and so to create the RTX 4070 Ti SUPER, it’s tapping into the larger AD103 silicon powering the RTX 4080 and the upcoming RTX 4080 SUPER that you’ll hear more about next week. The AD103 gives NVIDIA as many as 33% more shaders than the AD104, spread across 80 streaming multiprocessors; but more importantly, features a wider 256-bit GDDR6X memory bus, compared to the 192-bit bus of the AD104 that powers the original RTX 4070 Ti, RTX 4070 SUPER, and the original RTX 4070. This increase in memory bus width, along with a few more shaders, sits at the core of the new GeForce RTX 4070 Ti SUPER.

The new RTX 4070 Ti SUPER comes with 16 GB of GDDR6X memory across the full 256-bit bus width of the AD103 silicon. The memory ticks at 21 Gbps, yielding 672 GB/s of memory bandwidth, which is a significant 33% increase compared to the 504 GB/s of the original RTX 4070 Ti. NVIDIA carved the RTX 4070 Ti SUPER out of the AD103 by enabling 66 out of 80 streaming multiprocessors, which is a 10% increase over the original RTX 4070 Ti. This results in 8,448 CUDA cores, 268 Tensor cores, 66 RT cores, and 262 TMUs. There are, however, two major on-die changes compared to the RTX 4080. NVIDIA enabled 96 out of the 112 ROPs physically present on the AD103; which is still a 20% increase over the 80 ROPs of the RTX 4070 Ti. The other change is the on-die L2 cache. From the 64 MB available on the AD103, NVIDIA enabled just 48 MB of it. This is the same amount of cache as the RTX 4070 Ti and RTX 4070 SUPER. NVIDIA has given the RTX 4070 Ti a total graphics power (TGP) power limit of 285 W, which is the same as the RTX 4070 Ti. This power limit isn’t all that far from the 320 W of the RTX 4080.

The SUPER moniker denotes more performance at given price points, it doesn’t indicate the introduction of any new architecture level features. The GeForce Ada Lovelace architecture introduces the 3rd generation of RTX real time ray tracing technology. Its new generation CUDA core, in addition to support for higher frequencies and increased IPC, introduces shader execution reordering, a tech that should speed up ray tracing workloads. The new 3rd generation RT core, in addition to increased ray intersection performance, supports displaced micro-meshes, which allows greater complexity for ray traced objects. Lastly, the new optical flow processor is needed for DLSS 3 Frame Generation to work.

NVIDIA is positioning the GeForce RTX 4070 Ti SUPER as a 1440p graphics card. These kind of specs are frankly overkill for 1440p, given that even the regular RTX 4070 can provide around 60 FPS. What the RTX 4070 Ti SUPER does, however, is enable high refresh-rate 1440p gameplay. There are plenty of well priced 1440p displays with 144 Hz or 165 Hz refresh rates these days, and if the RTX 4070 Ti SUPER averages close to 3-digit FPS, then these displays have enough frame-rate to go on for a superior gaming experience compared to 60 Hz monitors. As with even the regular All the three RTX 4070-series cards we’ve tested before this one are capable of 4K gaming even though NVIDIA doesn’t recommend it; and so the RTX 4070 Ti SUPER should only be more capable of 4K. You have GeForce Experience, which can pick the right game settings for you, and then there are always DLSS and DLSS 3 Frame Generation.

The MSI RTX 4070 Ti SUPER Ventus 3X features a pretty spartan design, with a 2-tone black+silver cooler shroud that’s a callback to some of the older NVIDIA Founders Edition cards. The card is three slots thick, and looks fairly premium when installed in your machine. It sticks to NVIDIA reference clock speeds of 2610 MHz boost, and 21 Gbps memory. The power limit is set at the NVIDIA reference 285 W. The card draws power from a 12VHPWR connector, an adapter is included that converts two 8-pin PCIe power connectors to one of these. MSI is pricing this card at the NVIDIA MSRP of $800.