CES: NVIDIA GeForce RTX 4070 Super, RTX 4070 Ti Super and RTX 4080 are now official – data and information

Now it’s official: NVIDIA today announced the GeForce RTX 40 SUPER series at CES 2024 – including the GeForce RTX 4080 SUPER, GeForce RTX 4070 Ti SUPER and GeForce RTX 4070 SUPER. This latest iteration of GPUs based on NVIDIA Ada Lovelace architecture delivers up to 52 shader TFLOPS, 121 RT-TFLOPS and 836 AI-TOPS to accelerate gaming and creativity. The GeForce RTX 4070 SUPER starts at 599 US dollars.

PC gamers naturally want the best in visual quality for their hard-earned cash, and things like AI-based NVIDIA Deep Learning Super Sampling (DLSS) Super Resolution, Frame Generation and Ray Reconstruction combine with ray tracing to deliver breathtaking worlds. With DLSS, as many as seven out of eight pixels can end up being AI-generated, speeding up full ray tracing by up to 4x with better image quality. So much for the theory, but practice has already proven that there seems to be no way around such technologies.

I have already compiled the most important official slides and benchmarks, although these are of course manufacturer benchmarks and unfortunately they are only compared with their own cards. But even that can be quite informative in the end, until the first real benchmarks are published:

Cards

 

NVIDIA GeForce RTX 4080 SUPER specifications and performance

The NVIDIA GeForce RTX 4080 SUPER will use the AD103-400 GPU (PG139 SKU 355) with a full 10,240 cores, 320 TMUs, 112 ROPs and 64 MB L2 cache. The GPU is also operated at a slightly higher clock rate of 2295 MHz base and 2550 MHz boost compared to the RTX 4080 non-SUPER variant with 2205 MHz base and 2505 MHz boost. The chip will use a 256-bit bus interface, confirming the 16 GB GDDR6X memory, but utilizing faster 23 Gbps modules, allowing for an effective bandwidth of up to 736 GB/s, an approximately 3% increase over the non-SUPER variant. The NVIDIA GeForce RTX 4080 SUPER will have the same TGP as the RTX 4080 Non-SUPER at 320W.

The graphics card is expected to be 3 to 5% faster than the previous RTX 4080, which is not a major improvement, but factory overclocked models could offer a performance increase of almost 10%. This (slight) performance boost could help the card to be a tick faster than the Radeon RX 7900 XTX (which should be the intention behind it). If it were to be available at a price of just over 1000 EURO, this would of course be competitive and also a clear challenge.

 

NVIDIA GeForce RTX 4070 Ti SUPER specifications

The NVIDIA GeForce RTX 4070 Ti SUPER will be equipped with both the AD103-275 and the AD102-175 GPU SKUs (other board, PG141 SKU 323) with a total of 8,448 cores and 48 MB L2 cache. The graphics card will also have 16 GB of GDDR6 memory, which corresponds to a VRAM upgrade of 4 GB compared to the RTX 4070 Ti Non-SUPER and will therefore support a 256-bit bus interface. The graphics card will have a TGP of 285 watts, which in turn corresponds to the value of the non-SUPER variant.

On paper, the NVIDIA GeForce RTX 4070 Ti SUPER 16 GB GPU offers an average performance increase of 15% compared to the non-SUPER variant and is expected to compete well with the Radeon RX 7900 XT, perhaps the street price will even be lower or AMD will pick up the gauntlet. However, as with the non-SUPER cards, there will be no NVIDIA’s own Founders Edition (FE), only AIC cards.

NVIDIA GeForce RTX 4070 SUPER specifications

The NVIDIA GeForce RTX 4070 SUPER will use theAD104-350 in full configuration (or the AD103-175 SKU in the leftover version), which offers a total of 7,168 cores. The GPU has 36 MB of L2 cache and will be equipped with 12 GB of GDDR6X memory via a 192-bit bus interface. Compared to the existing RTX 4070, the TGP of the RTX 4070 SUPER will be 20 W higher and will then be 220 W.

The performance of the NVIDIA GeForce RTX 4070 SUPER is expected to be around 14% faster than the RTX 4070 Non-SUPER. The graphics card will also receive the new CEM 5.1 Gen5 power connector. The existing RTX 4070 graphics cards come with either an 8-pin connector or a 12-pin power connector (supplied with adapter).

 

The information was provided by NVIDIA in advance under NDA with the condition that the embargo period be strictly adhered to.