According to the latest rumors, the Nvidia GeForce RTX 5090 will be significantly quicker than its current flagship gaming GPU, with a performance boost of up to 70% compared to the RTX 4090. Nvidia’s new gaming graphics card, based on the forthcoming Blackwell architecture, is expected to launch at the end of 2024, and many people are predicting that AMD won’t be releasing a high-end GPU to counter it.
This performance prediction marries up with the rumored RTX 5090 specs, and would make this GPU likely to go straight onto our best graphics card guide if you want the very best gaming performance. However, if there is indeed no competition at that level of performance, Nvidia will be able to charge whatever it likes for the RTX 5090, and some people are expecting it to cost up to $2,500.
The latest speculation comes from regular PC hardware gossiper Tom at Moore’s Law is Dead, who has an interesting discussion about Nvidia’s next-gen GPUs with his co-host Dan in the video below. “From what I’ve heard it could be 60 or 70% faster,” says Tom when discussing the GeForce RTX 5090 in comparison to the RTX 4090, and the duo predict that it’s very unlikely to have a fully-enabled die, as Nvidia will want to reserve its very best chips for the booming AI market.
“They’re not going to put the full die on the 5090, because the full die is too valuable in other markets,” says Tom’s co-host Dan. “It’s going to be cut down however much they need to so they can have a decent amount of products binned into the 5090 category… It’s more a question of what they can get away with more than what it needs to be.”
This wouldn’t be a new strategy for Nvidia, as not every chip it makes in the early days of production will be stable enough to have all of its parts enabled. Neither the RTX 4090 or the 3090 had fully-enabled GPUs. Tom from Moore’s Law is Dead speculates that, if he was Nvidia, none of the GPUs in the Blackwell lineup would have fully enabled chips.
“I would just do like one $2,000 5090, a $1,000 5080, a $700 5070, $400 5060, and then maybe super long term some 5050 Ti for $300 or something, and that would be five cards,” he says. “Maybe all of them are cut down 5%, none of them get the full die, none of them are cut down a lot though, and everything else is just sold as laptop chips, AI, or data center. I don’t know why they wouldn’t do that next gen – just keep it simple.”
What immediately stood out to me here was the $2,000 price for the RTX 5090 which, given the predicted lack of high-end competition from AMD, could be what Nvidia ends up charging for its flagship gaming GPU. In fact, Tom from Moore’s Law is Dead, even speculates that it could end up costing up to $2,500 if it has the right performance, saying Nvidia would need to “just make it command that $2,000-2,500 price point.”
In the meantime, if you’re looking to build a gaming PC for a more palatable price, check out our GeForce RTX 4070 Super review, where we put Nvidia’s latest mid-range GPU through its paces.