NVIDIA RTX 4080/4070 SUPER update might involve a mix of AD103/102/104 GPUs

NVIDIA RTX 40 SUPER with three different GPUs

Another day, another RTX 40 SUPER rumor. 

Kopite7kimi joined the ongoing discussion regarding the expected NVIDIA RTX 40 SUPER refresh in the upcoming months. In contrast to some other leakers, Kopite was quick to point out that the rumor of a 20GB memory capacity for the RTX 4080S may not hold true.

Most of the leakers that are discussing the RTX 40S update have a good track record, which means that RTX 40 update is no doubt discussed behind the closed doors. However, it should also be noted that NVIDIA very often shares varying specifications with multiple sources, including their board partners. This practice frequently results in conflicting information and some details never materializing in the final product SKU. Nevertheless, in certain cases, rumors do eventually receive confirmation, even if it takes up to a year, as we saw with the RTX 3080(Ti) 20GB models. That said, something is definitely happening, but it may take weeks before we get the full information.

Now, according to Kopite7kimi, the RTX 40 SUPER refresh might involve GPU mixing. This concept is not widely understood because NVIDIA typically releases such SKUs primarily in the Chinese market. The rumor suggests that the NVIDIA RTX 4080 SUPER could potentially utilize either the AD103 or AD102 GPU. If this turns out to be true, it would clash with the previous rumor regarding a potential 20GB memory upgrade, albeit not entirely. This is because, to accommodate 20GB of memory, NVIDIA would need to employ either a 320-bit or 160-bit memory bus. Only the latter option is supported by the AD103 GPU, which is limited to a 256-bit memory bus, but it may look strange next to RTX 4080 with 256-bit bus.

As for the RTX 4070 SUPER, there have been reports solely about a potential upgrade to the AD103 processor, which would imply an increase in core count. However, Kopite claims that the specifications are actually slightly lower than those of the RTX 4070 Ti, meaning that the Ti variant would remain the faster option.

By using two different GPUs, NVIDIA has to make sure that both processors perform identically, not causing any frequency or power differences; otherwise that would call for a different SKU variant. NVIDIA is not using such chips for the US or EU market because striker consumer laws prevent such disproportions. As we learned from Chinese models using different GPUs, there are indeed sometimes some visible differences, like higher power consumption or access to different video coding capabilities, but this problem was rather seen on entry-level series.

We are constantly discussing the RTX 40 SUPER update with our sources, which confirm that these new SKUs should not be expected any time soon. Once NVIDIA submits final specs to them, we finally start talking launch dates and pricing.

Source: kopite7kimi