Nvidia’s next-gen graphics cards like the RTX 4080 and other ‘Lovelace’ models won’t be all that different in terms of their core architecture compared to current RTX 3000 GPUs, or so the rumor mill reckons.
This is according to hardware leaker Greymon55, one of the regulars on Twitter when it comes to GPUs, who was replying to a tweet querying how AMD’s next-gen RDNA 3 cards will compare to what will presumably be Nvidia’s RTX 4000 range.
Now, while this could be seen as disappointing news on the face of it, before we get carried away here, just because the underlying architecture for Lovelace won’t be all that different to Ampere, doesn’t mean that RTX 4000 products won’t be a chunky leap in performance.
Nvidia is dropping the process from 8nm with current Ampere GPUs to make Lovelace on 5nm, and that should allow Team Green to ramp up clock speeds considerably, upping performance in that way, and also doing the same with core counts (and doubtless ray tracing cores too, as Nvidia won’t want to give up pushing its advantage here).
Plus naturally the architecture, even if it is pretty similar – and that’s assuming this speculation is correct – will still be honed for next-gen on top of that, and we’ll likely see other benefits like improvements with memory bandwidth.
Analysis: Supercharged cards, maybe, but there’s still a big concern here
All of the above refinements could add up to a seriously supercharged fresh take on Ampere, and that’s exactly what we’ve been hearing going by previous chatter on the grapevine. Indeed, rumors have floated the idea in the past that Lovelace could be twice as powerful as RTX 3000 models, but obviously take that with a huge amount of caution.
This could turn out to be a similar scenario to when Nvidia switched from Maxwell to Pascal, a move which saw some major improvements in the frame rate stakes, despite not much in the way of architectural change between generations.
The worry is that this push for better performance and ramping things up in this way could lead to some rather outlandish power demands, certainly at the high-end, for RTX 4000. And again we’ve seen plenty of speculation about how far Team Green might push here, with one rumor suggesting a 1200W PSU for the RTX 4090, possibly requiring buyers of new GPUs like the flagship, and indeed perhaps the RTX 4080, to also be looking at a power supply upgrade. In other words, even more expense on top of what will likely be pricey new graphics cards.
Meanwhile, rival AMD is looking at going a very different route with next-gen GPUs, and again if the rumors are right, we’ll see the launch of Team Red’s first cards to use a multi-chip module (MCM) design.
Via PC Gamer