If recent leaks are accurate, Nvidia will add 'Super' GPUs to its RTX lineup in July

Attachment 26985

It has been three weeks to the day since Nvidia posted a cryptic tweet from its GeForce account saying, "Something super is coming..." Since then, AMD has come out and announced its first run of Navi graphics cards—Radeon RX 5700 XT and 5700—but whatever "super" surprise lies in wait from Nvidia has yet to reveal itself. Well, officially anyway.

There have been a few leaks pointing to several upgraded Super variants of Nvidia's existing RTX cards. Namely, it's rumored Nvidia is getting ready to announce three new RTX cards, including Super versions of the RTX 2080, 2070, and 2060.

Nothing has been confirmed, so a dose of skepticism is in order, or a grain of salt if you prefer. However, both Videocardz and Igor Wallossek from our friends at Tom's Hardware Germany are reporting the same SKUs and specs, with the former saying the latter "was the first to report the correct variants." In other words, Videocardz is claiming to have independently verified Wallossek's information.

Let's break this down into two parts. First, here is a look at the GPUs that the new cards will be using, along with the GPUs employed in currently shipping RTX cards, and the accompanying number of CUDA cores:

• GeForce RTX 2080 Super—TU104-450 (3,072 CUDA cores)
• GeForce RTX 2080—TU104-400 (2,944 CUDA cores)
• GeForce RTX 2070 Super—TU104-410 (2,560 CUDA cores)
• GeForce RTX 2070—TU106-400 (2,304 CUDA cores)
• GeForce RTX 2060 Super—TU106-410 (2,176 CUDA cores)
• GeForce RTX 2060—TU106-200 (1,920 CUDA cores)

According to the leaked specs, the RTX 2070 and 2060 Super models will have 256 more CUDA cores than the regular variants, while the RTX 2080 Super gains 128 CUDA cores over the standard version. There is no mention of clockspeeds, but the additional CUDA cores would give the Super cards more raw muscle to throw at games and graphics.

Now let's have a look at the memory configuration of each card, including the amount, speed, and bandwidth, alongside their non-Super original versions:

• GeForce RTX 2080 Super—8GB GDDR6 @ 16Gbps, 512GB/s memory bandwidth
• GeForce RTX 2080—8GB GDDR6 @ 14Gbps, 448GB/s memory bandwidth
• GeForce RTX 2070 Super—8GB GDDR6 @ 14Gbps, 448GB/s memory bandwidth
• GeForce RTX 2070—8GB GDDR6 @ 14Gbps, 448GB/s memory bandwidth
• GeForce RTX 2060 Super—8GB GDDR6 @ 14Gbps, 448GB/s memory bandwidth
• GeForce RTX 2060—6GB GDDR6 @ 14Gbps, 336GB/s memory bandwidth

The regular RTX 2060 is the only card of the bunch with 6GB of GDDR6 memory mated to a 192-bit bus. Every other SKU has 8GB of GDDR6 memory and a 256-bit bus, including the RTX 2060 Super, hence the higher memory bandwidth.

If the leaked specs are accurate, the RTX 2070 Super will not see a change in the memory configuration, while the RTX 2080 Super will use faster clocked memory chips, resulting in a bit more bandwidth.

Both Videocardz and WCCFTech have pegged next week as a possible time frame for an announcement from Nvidia, while noting that an actual launch might be sometime in July. We'll have to wait and see if that's actually the case.

The same goes for pricing—nothing substantial has been leaked, though Videocardz seems to think that the Super variants will slip into the non-Super MSRPs, which in turn would push the pricing down on those existing models.

If all this comes to fruition, things will get interesting in a hurry. During its unveil, AMD showed slides claiming that its Radeon RX 5700 XT is generally faster than the RTX 2070 when playing games at 1440p. The same goes for the non-XT model, compared against the RTX 2060.

We obviously won't draw any conclusions until we've fully tested AMD's upcoming hardware, along with whatever Nvidia might have in store.