NVIDIA GeForce RTX 3000 SUPER, the alleged technical specifications appear

It has been rumored for some time that NVIDIA intends to introduce, before the new GeForce RTX 4000 range, a refresh of Ampere solutions under the brand SUPER, introduced for the first time with the GeForce RTX 2000 in the summer of 2019. The rumors are different and in the past few hours HP has indicated, we do not know whether or not correctly, configurations of the new Envy 34 desktop up to the new GeForce RTX 3080 SUPER.

To give further life to the rumors a leaker who has repeatedly revealed the specifications of the new cards in advance, Kopite7kimi. According to what was widely known, the SUPER range would include the following models: RTX 3090 SUPER, RTX 3080 SUPER, RTX 3070 SUPER e RTX 3060 SUPER. If so, NVIDIA would add a model over the previous generation, the RTX 3090 SUPER.

Regarding the specifications always good to remember that since these are “rumors” they may not be definitive and we have also recently had proof of it with the appearance of an RTX 3080 Ti on the Russian market: for a long time that seemed to be the correct memory configuration of the new card unveiled this summer, but then NVIDIA changes its mind by falling back on different specs and 12 GB of memory. Furthermore, the same leaker says he is uncertain about some data and the final name of the RTX 3090 SUPER.

Having said that, let’s go. According to Kopite7kimi the “alleged” GeForce RTX 3090 SUPER should be based on a GPU GA102 fully active with 10752 cores side by side with 24 GB GDDR6X. NVIDIA would unlock the two disabled Streaming Multiprocessors (SM) on the current card, offering the GA102 in all its glory. Accordingly, the card should also offer 336 Tensor cores and 84 RT cores. The core data matches that indicated by another leaker in recent weeks. With such specs, the TDP could aim for 400W, unless NVIDIA does a painstaking job of selecting the best dies.

Regarding the GeForce RTX 3080 SUPER, the card could boast a GPU (GA102 or GA103?) with 8960 CUDA core and 12 GB of GDDR6X memory. Remember that the RTX 3080 has a GPU with 8704 CUDA cores and 10 GB of GDDR6X memory, while the more recent RTX 3080 Ti uses 10240 CUDA cores and 12 GB of GDDR6X. The RTX 3080 SUPER it would then position itself in the middle, leaving the suffix Ti a position of superiority in the hierarchical scale.

The GeForce RTX 3070 SUPER should see instead only an improvement on the memory front over the RTX 3070, in fact we are talking about 5888 CUDA core (GA104?) and the transition to GDDR6X. The recently introduced RTX 3070 Ti figures 8GB of GDDR6X memory, but a fully active GA104 GPU with 6144 CUDA cores.

Finally, here is the GeForce RTX 3060 SUPER, a card that it should show up with 5632 CUDA core and 12 GB of GDDR6 memory, far more than both the 4864 CUDA cores of the RTX 3060 Ti (equipped with 8 GB GDDR6) and the 3584 cores of the RTX 3060 (equipped with 12 GB GDDR6). The latter based on a partially activated GA106 GPU, with 2 SMs less than the maximum 30 SM for a total of 3840. Consequently the RTX 3060 SUPER could be based on a GA104 chip.

The latter model, on paper, could present the highest performance boost compared to the other SUPERs compared to the cards already on the market, but at the same time it would position the SUPER brand above the Ti one, while for the other models it would be the opposite. This may imply the departure of some of today’s models as soon as NVIDIA introduces the SUPER brand, perhaps at the beginning of 2022. We just have to wait for more concrete information.