Introduction

NVIDIA today debuted the GeForce RTX 4080 Super, the third and final part of the Super mid-lifecycle refresh of its RTX 40-series Ada generation. NVIDIA focused on housekeeping the upper end of its GeForce RTX product stack with this refresh. For the RTX 4070 Super and RTX 4070 Ti Super, this meant more performance at their existing $600 and $800 price points, respectively. The company’s approach to the RTX 4080 Super is slightly different. There is some increased performance to be had over the original RTX 4080; but at a 20% lower price point of $1,000, compared to the RTX 4080, which launched at $1,200. Much like the RTX 4080 and the flagship RTX 4090, the new RTX 4080 Super is recommended by NVIDIA for maxed out AAA gaming at 4K Ultra HD, including with ray tracing.

Back when rumors of the RTX 40-series Super refresh started first pouring in, it was widely expected that the RTX 4080 Super would be carved out of the larger AD102 silicon that the RTX 4090 is based on; by giving it a higher number of shaders than available on the AD103 silicon, and a wider memory interface, such as 320-bit, driving a 20 GB video memory. This could have well been something NVIDIA continued to sell at $1,200, but the RTX 4080 Super turned out a whole different proposition. The RTX 4080 enabled 76 out of 80 streaming multiprocessors available on the AD103 silicon (95% of the SM enabled). To create the RTX 4080 Super, NVIDIA simply enabled all 80 SM, slightly increased the memory speed from 22.4 Gbps to 23 Gbps, and lowered the price by 20%, down to $1,000.

The RTX 4080 Super hence has the AMD Radeon RX 7900 XTX in its crosshairs, which had originally launched at the same $1,000 price, but is now available at a street price of around $900. Since the RX 7900 XTX is AMD’s fastest graphics card, which maxes out its biggest GPU, there is no threat to the RTX 4090 from Team Red, and there was no need for NVIDIA to make the RTX 4080 Super a significantly bigger SKU based on the AD102, while holding onto the $1,200 price-point. Instead, the RTX 4080 Super is designed to be the RTX 4080 with a few extra shaders, and a touch faster memory, at a lower price.

With the AD103 silicon maxed out, the RTX 4080 Super gets 80 streaming multiprocessors, which work out to 10,240 CUDA cores, 320 Tensor cores, 80 RT cores, 320 TMUs, and 112 ROPs. The chip’s memory interface is 256-bit wide, driving 16 GB of memory, but powered by the fastest graphics memory chips in the industry, 23 Gbps GDDR6X. This yields a bandwidth of 737 GB/s, which is cushioned by 64 MB of on-die L2 cache. The GPU frequency is slightly increased, from 2505 MHz boost, to 2550 MHz. Interestingly, NVIDIA didn’t tinker with the total graphics power (TGP) of 320 W, which is unchanged from the RTX 4080.

As with the other GeForce RTX Super products, the RTX 4080 Super doesn’t see the introduction of any new feature, it is an exercise and bringing more performance at existing or lower price points; the underlying technology is still the GeForce RTX Ada Lovelace architecture. Ada leverages the new 5 nm EUV foundry node to significantly dial up transistor counts, and improve energy efficiency. The new generation CUDA core, in addition to supporting higher clock speeds and providing higher IPC, supports shader execution reordering, something that speeds up ray tracing performance. The new 3rd generation RT core supports displaced micro-meshes, a technology that allows game production designers to increase complexity of ray traced objects. The new optical flow processor lets the GPU draw entire alternate frames using AI, and without involving the graphics pipeline, which is how the new DLSS 3 Frame Generation technology works.

Unlike the GeForce RTX 4070 Ti Super from last week, the new RTX 4080 Super gets the Founders Edition treatment from NVIDIA. The GeForce RTX 4080 Super Founders Edition isn’t intended to be a reference design, but a first-party custom design that sells at the $1,000 MSRP, and sets design and quality standards for the other board partners to live up to, especially with their products priced at MSRP. The new RTX 4080 Founders Edition card features a physically identical product design to the RTX 4080 Founders Edition, except for a few visual changes—the gunmetal finish makes way for metallic matte black; with a distinctive RTX 4080 Super logo on the backplate, with the “RTX 4080” portion in recessed lettering, and the Super brand extension raised. The RTX 4080 Super Founders Edition, as we mentioned, comes in at the company’s baseline price for this GPU, at $1,000.