Next-gen Nvidia GPU leak sparks worries over RTX 4070

Close-up of RTX 3070 Ti Promotional Image
(Image credit: Nvidia)

Nvidia’s next-gen RTX 4000 graphics cards have seen some new info emerge from the rumor mill, which has some folks questioning how the RTX 4070 will shape up.

This comes courtesy of Kopite7kimi, a regular hardware leaker on Twitter, who dropped some new purported spec details pertaining to the top three Lovelace GPUs, as well as a nugget of info about pricing – take all of this with the usual heightened sense of skepticism, mind.

See more

The spillage regarding the flagship lines up with what the leaker already told us just over a week ago, namely that the RTX 4090 will have a slightly increased CUDA Core count of 16,384. Back when that detail was first dropped, the leaker couched it in terms of this “may be” the case, so this would seem to be a suggestion that they’re surer of this configuration now (as opposed to the long-rumored 16,128 cores).

The memory loadout of the flagship Lovelace graphics card will supposedly be 24GB of GDDR6X (21Gbps, with a 384-bit memory bus).

Kopite7kimi also informs us that the RTX 4080 will have 10,240 CUDA Cores with 16GB of VRAM, though it’s not certain whether that will be GDDR6 or GDDR6X (and a 256-bit memory bus). And as for the RTX 4070, that’ll theoretically run with 7,168 CUDA Cores and 10GB of video memory, which will be GDDR6, with a 160-bit bus (we’ll come back to that).

We’re also told in stern terms that we should not be expecting lower asking prices (recommended retail or MSRP) than we’ve seen with current-gen RTX 3000 graphics cards.

Finally, in a follow-up tweet, Kopite7kimi outlines supposed power consumption levels, pegging the RTX 4090 at 450W (as has been the case before), and the RTX 4080 at 420W (with a question mark added here), with the RTX 4070 set to draw 300W.

Analysis: Pricing reality check – and worries over the RTX 4070

Was anybody really expecting lower price tags? We certainly weren’t, as when these graphics cards first launch, if they have anything like the rumored performance bumps, they’ll be much sought after. And just look at how demand is running for Nvidia’s current GPUs: at Newegg, the whole top 30 of best-selling cards is entirely Nvidia models, with not an AMD board in sight, a sad state of affairs when it comes to competitivity around pricing. (And of course that’s why Nvidia’s graphics cards aren’t falling in price as fast as Team Red’s RX 6000 GPUs).

The other point regarding pricing is that the rumor mill believes Nvidia will be first to market with next-gen products too, so given the overall demand situation, and getting its new GPUs out first, it’s hard to imagine a situation where Team Green would be looking at going easy with MSRPs. In fact, if anything, we’d expect a slight price bump for all the above reasons, at least with some models upon launch.

Okay, pricing aside, there are a few big nuggets of spec info here, and that includes the RTX 4070 only having 10GB of GDDR6 VRAM with a 160-bit bus, which has got some folks worried that it looks underpowered in that respect (not to mention anticipating that a 12GB loadout is being reserved for a future 4070 Ti).

Honestly, though, we’d be cautious about putting too much stock in particularly this bit of the rumor-mongering here, as it does seem a bit of an odd route to take for Nvidia with the memory configuration. Time will tell, although the purported power draw of 300W does not point to a weak sauce RTX 4070 of course – that’s almost reaching RTX 3080 levels (320W).

As for the RTX 4090 and 4080 being pitched at 450W and 420W, the former seems to be something of the recent consensus reached on the GPU grapevine, while the latter seems very close to the flagship’s power draw. Again, this may be wrong of course – it’s telling that the leaker marks that 420W figure with a question mark, and that’s the only place one appears with the rumored TBPs – but it is possible. As we’ve discussed before, the RTX 3080 and 3090 had the same 30W difference (350W and 320W respectively), so this isn’t without precedent.

However, what’s interesting to note is that Kopite7kimi recently asserted a 99% confidence level in that 420W prediction for the RTX 4080, and yet there’s a question mark added here – which looks like a step back from that previous confidence level.

All of which just underlines how shaky and shifting these theories on the rumor mill tend to be. To sum up, let’s sit tight for now and not jump to conclusions, because the RTX 4000 launch isn’t too far off, and we may even have an official reveal – or at least firmer rumors – in relatively short order.

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).