Nvidia RTX 4070 may not show up this year – but the replacement GPU could surprise you

RTX 3070 Ti Close Up Showing Fan and Logo
(Image credit: Nvidia)

Nvidia’s RTX 4070 may not be among the first batch of next-gen GPUs released, and if the rumor mill is right, we’ll get two different spins on the RTX 4080 arriving as well as the flagship RTX 4090.

The latest chatter from the grapevine on Lovelace comes from multiple sources who VideoCardz spoke to, combined with info drawn from regular hardware leaker Kopite7kimi on Twitter (who recently summarized all their latest findings), and Chiphell (Chinese forums), all of which corroborates VideoCardz’s findings in some places.

There’s some fresh info on clock speeds and power usage, but the biggest revelation here is that Nvidia seems to be planning to release two versions of the RTX 4080, one with 16GB of VRAM, and one with 12GB – but there will be other major differences with the spec (take all this with large heaps of seasoning, naturally, as with any rumor).

The RTX 4080 16GB is the one said to be built on the AD103 GPU with 9,728 CUDA Cores, with the 12GB spin on the 4080 supposedly dropping down to use the lesser AD104 GPU, and it’ll sport 7,680 CUDA Cores – quite a difference.

As VideoCardz observes, what appears to be the case is that the RTX 4080 12GB is what was previously identified on the rumor mill as the RTX 4070 (or 4070 Ti) – which kind of explains why it seemed so beefy, something we’ve commented on in recent times.

VideoCardz informs us that Nvidia has made a “last-minute name change” for this card to turn it into a lesser spin of the RTX 4080, and for the moment, there is no RTX 4070 – with graphics card manufacturers not expecting this to change (with the Lovelace launch now so close: it’s expected on September 20, at GTC).

Other spec nuggets shared in this fresh rumor dump include that the RTX 4080 12GB will have a power usage of 285W (TGP or Total Graphics Power), which can be pushed up to 366W. The boost clock will purportedly be 2610MHz.

The stronger version of the RTX 4080 with 16GB of VRAM will supposedly boost to 2505MHz, with a TGP of 340W (able to be pushed up to 516W maximum). The GDDR6X memory is said to run at 23Gbps, notably, and this will be the only GPU to have that speedy VRAM.

As for the RTX 4090, that boosts to 2520MHz, and the rumor mill again points to that long-held TGP of 450W, but at max it can be pushed to 660W. The 24GB of video RAM will be 21Gbps modules, the same as the lesser RTX 4080 12GB.

The tech site tells us that the RTX 4090 should be out in the first half of October, and the RTX 4080 models are apparently set to debut in the initial two weeks of November.

Nvidia RTX 3070, RTX 3080, And RTX 3090 Lined Up In A Promotional Image From Nvidia

(Image credit: Nvidia)

Analysis: RTX 4070 ETA?

So, what about the RTX 4070 then? Well, we had heard mention a couple of weeks back – again from Kopite7kimi on Twitter – that there were two possible specs for the 4070 graphics card. The mentioned higher-end possibility is now the RTX 4080 12GB, in theory, so that would leave the other configuration of an AD104 chip with 10GB of VRAM and 7,168 CUDA Cores to be the RTX 4070. Maybe. (Let’s face it, the Lovelace rumor space is becoming increasingly confusing).

If this is what’s happening, it’ll be disappointing for a couple of reasons. Firstly, the RTX 4070 only having 10GB of video memory won’t go down well, and secondly, the prospect of it being late to the table is disheartening, too. When will it turn up? If  there’s no sight of it yet, at all, is this a GPU that Nvidia is planning to delay until early 2023, as some rumors have suggested?

Before we get too downhearted, though, what we don’t yet have is any indication of pricing for Lovelace outside of the RTX 4090 (we’ve just heard some hints as to the flagship’s price tag, and sadly, the news isn’t what we wanted to hear).

If we aren’t getting an RTX 4070 initially, but we are getting an RTX 4080 12GB which is priced relatively okay for its performance – considering it does drop down a fair way from the 4080 16GB in terms of a lesser GPU and considerably fewer CUDA Cores – then maybe it won’t matter. Which is to say that it’ll all be about the price/performance ratio of these GPUs, whatever they end up being called.

That Nvidia has purportedly decided to notch up the name to an RTX 4080 from the 4070, however, leaves a worrying feeling about pricing in our gut. Particularly given the above mentioned RTX 4090 pricing leak. Time will tell, as ever, and with less than a week to go before the Lovelace launch, we won’t have to wait long to discover more about Team Green’s next-gen graphics cards.

All this does leave us wondering, however, whether the tactic might be to charge a premium for the new products – pretty much as expected for high-end GPUs fresh off the assembly lines – and further knock down pricing on that excess RTX 3000 stock as the alternative option for those who don’t want to splash out too much.

Darren is a freelancer writing news and features for TechRadar (and occasionally T3) across a broad range of computing topics including CPUs, GPUs, various other hardware, VPNs, antivirus and more. He has written about tech for the best part of three decades, and writes books in his spare time (his debut novel - 'I Know What You Did Last Supper' - was published by Hachette UK in 2013).