The latest leaks about the GeForce RTX 40 SUPER left us with very detailed information about its possible specifications, as well as consumption levels, prices and even a release date. In order to put some order among all the rumors, I wanted to make this article as a summary with information more reliable and interesting.

Before we get down to business, it’s important to be clear about what the GeForce RTX 40 SUPER is. it’s about intergenerational renewal graphics cards that will use the Ada Lovelace architecture and that will be made up of more powerful models than the originals. This means that they do not bring changes at the architectural level or support new technologies, and therefore are just a refresh of the current generation.

Possible specifications of the first three models

GeForce RTX 4070 Super

• Graphics core AD104 in 5nm.
• 7,168 shaders at an unspecified frequency.
• 224 texturing units.
• 96 raster units.
• 224 tensor kernels of the fourth generation.
• 56 cores of the third generation RT.
• 192-bit memory bus.
• 12GB GDDR6X.
• TGP 225W.

The performance difference between this model and the GeForce RTX 4070 would be quite large since it has 5,888 shaders. In general, if these specifications are met, the SUPER model will be only slightly slower than the Ti version.

GeForce RTX 4070 Ti Super

• Graphics core AD103 in 5nm.
• 8,192 shaders at an unspecified frequency.
• 256 texturing units.
• 96 raster units.
• 256 tensor cores of the fourth generation.
• 64 cores of the third generation RT.
• 256-bit memory bus.
• 16GB GDDR6X.
• 285W TGP.

There is also talk of a configuration with 8,448 shaders, but all information agrees that it will come with a 256-bit bus and 16 GB of graphics memory. That memory system configuration should significantly boost 4K performancethe resolution in which it will show the biggest differences compared to the current Ti model.

GeForce RTX 4080 Super

• Graphics core AD102 in 5nm.
• 10,240 shaders at an unspecified frequency.
• 320 texturing units.
• 112 raster units.
• 320 tensor cores of the fourth generation.
• 80 cores of the third generation RT.
• 320-bit memory bus.
• 20 GB of GDDR6X memory.
• TGP 320W.

The increase in shaders over the original model would be small, but the new 320-bit memory bus should improve the performance data more noticeably, especially in 4K, which This is where the difference that more bandwidth can make is most noticeable. in graphics memory.

Well, if everything you said is fine, let’s see:
RTX 4080 Super, AD103-400, 10240FP32;
RTX 4070 Ti Super, AD103-275 or AD102-175, 8448FP32, 48M L2;
RTX 4070 Super, AD104-350 or AD103-175, 7168FP32, 48M L2.
I still have my doubts with them, especially Ti Super. I can’t fully agree.

— kopite7kimi (@kopite7kimi) October 28, 2023

I don’t want to close this section without talking about consumption, and the first information assures that they will be almost identical to the current models. As we can see in this summary of specifications, only the GeForce RTX 4070 SUPER would have a slightly higher consumption than the standard model (20 watts more), while the other two they would have the same TGP as the original models.

Possible prices and release date

NVIDIA is expected to introduce the GeForce RTX 40 SUPER sometime in the future. first quarter of 2024, and CES in January sounds strong as a possible gala venue for said presentation, though nothing is confirmed. If this happens, it is likely that the launch of these new graphics cards will end between the first and second quarter of 2024.

As for the price, there is nothing definitive, but it is rumored that the GeForce RTX 40 SUPER They will cost the same or slightly more than the original version. If NVIDIA goes this route, it’s clear that it would have to make a significant price adjustment to better organize its catalog and make room for these new additions.