The latest Nvidia RTX specs rumour makes RTX 4090 seem like a goddamn clown card

If the specs of the Nvidia RTX RTX5090 and RTX5080 graphics cards are true, we're looking at a GPU with a staggering amount of processing power. And a second-tier processor that's already pissing a lot off gamers.

The latest spec leak' comes from kopite7kimi - a more or less reliable X-leaker. It tells a tale of an almighty RTX-5090 with a specs boost that gives it a mouthwatering amount TSMC-built technology in a single graphic card. It's still a reference card, and reportedly a two-slot one at that.

The RTX 4090 will look like a clown in more than one way. This Founders Edition graphics card is so large and oversized that it still makes me laugh when I put it in the office test rig. This will make the card look even bigger, as it's going up against what is reportedly a dual slot card. Before you even get to the spec of this new GPU, it's a mammoth beast. It has 21,760 CUDA Cores (effective FP32 Units)--that is still not the full GB202 complement--which is a 33% increase over the RTX 4090. The RTX 3090 had 50% more cores.

The new Nvidia Blackwell flagship GeForce cards will reportedly be flexing, with a 512 bit aggregated memory bus as well as a full 32 GB GDDR7 RAM. Guess what? That's a 33% boost in both memory and the memory bus. The increased power numbers (presumably, the TGP of card) suggest a 600 W maximum, which, if math is your thing, also hits the seemingly magical 33% mark.

What kind of performance improvements will Nvidia be touting for its new card? It's got to be better than 33%, right? In our testing, we found that the RTX 4090 performed 72% faster on average than the RTX 3090 when it came to our 4K gaming tests. I'm hoping that it's actually more in line with an increase in memory bandwidth.

Memory bandwidth is one of the few numbers that doesn't follow this trend. The RTX-5090 will supposedly have a memory bandwidth of 1,792 GB/s. We're looking at 1,792 GB/s of raw memory bandwidth with the 512-bit bus, and 28 Gbps GDDR7. That's a 78% increase compared to the RTX4090.

I don't understand how Nvidia will squeeze all this extra power into a dual slot GPU design. Rumours suggest that the only change will be a small one in the process node (N4P instead of N4 for Ada generation), so you won't see a significant difference in cooling. We could be looking at a Founders Edition that is very special indeed, possibly some sort of liquid-cooled beastie.

It's doubtful that third-party GPU vendors will still add monstrous cooling to their own versions. I also doubt that subsequent RTX 50 series Founders Edition cards will receive a similar treatment. Who knows.

I don't really want to think about the price of this thing. The RTX 4090 is a relative bargain compared with the price increases across the RTX 40 series, but I find it hard to believe that Nvidia will not price its flagship GPU below $2,000 in a world without any competition. If we go by the 33% figures, then $2,100 is a reasonable price.

It's not surprising that the RTX 4080 Super will be priced at the same as the RTX 5080.

Yes, it is the RTX-5080. The rumours and the reactions to its specs have not been kind. This new card doesn't look very exciting, with a reported 5% increase in CUDA Cores over the RTX 4008 Super. The same 256-bit interface and 16 GB GDDR7 VRAM are the only things that get my techie heart pounding.

It appears that it is relying on a massive increase in memory bandwidth, and a lot more power to be able gain the performance advantage over the RTX 4080 Super to deliver a really tantalising prospect for an upgrade. The rumours suggest that it should be able to beat the gaming performance the RTX 4090 just a little bit--though I'm struggling with this given these specs--but I still think it's going to be a hard sell because the gap between the RTX 5080 ad RTX 5090 are so wide you could throw an asteroid.

As always, we have to take these specs lightly until we receive official confirmation from Nvidia about its new cards. It keeps them close to the chest up until launch. But they do feel right. In a sense, it feels wrong. No one wants the RTX 5080 looking so gimped. The only exception is AMD, who may be tempted to come out of retirement of their high-end GPUs if the target can be achieved.

We don't yet know when the cards will be released. People I've talked to have long believed that it would launch this year, in some form. But Nvidia doesn't need to release the RTX-50-series until 2024. It has no competitors, and the RTX 40 series are still performing well right now. Even if the RTX 4090 and RTX 3080 launch in the next 3 months, there's a possibility that these numbers will not be final. Especially those power figures. There was a lot of TGP speculation before the final 450 W reveal for the RTX 4090. So, even if the leaked specs are accurate as of now, there could be some movement.