Nvidia RTX 4070 leak is an eye-opener, but let’s not dismiss this GPU yet

Nvidia’s RTX 4070 graphics card has been flying low under the radar of late, with all attention focused on the 4070 Ti which is supposedly about to launch, but now we’ve seen leaked specs for the lower-tier model.

Sadly, it’s not great news on the face of it, as according to leaker Kopite7kimi - a regular presence on Twitter when it comes to GPU spillage - the RTX 4070 will be considerably cut-down compared to the RTX 4070 Ti.

At least when it comes to the CUDA cores anyway, because as you can see, the RTX 4070 is set to purportedly run with 5,888 cores, a fair old dip from the max loadout of the AD104 GPU the 4070 cards are based on, which is 7,680 cores (the configuration the rumor mill believes the 4070 Ti will tout).

Also, don’t forget that this figure for the RTX 4070 matches the RTX 3070 which has 5,888 cores – though remember, architectural advances and other specs (like clock speeds) means a 4070 configured as this leak reckons will still be a good deal faster (there’s no chance Nvidia would have it otherwise, of course).

The better news is that while past speculation pointed to an RTX 4070 with 10GB of VRAM – a prospect that prompted a lot of bitterness in GPU-related online discussions, for sure – this leak pins a 12GB loadout on the 4070. It also allegedly has a 192-bit memory bus, which is good to hear since lesser bandwidths have been, er, bandied about in the past.

Analysis: Weaker or stronger, pricing will be key

What we’ve also heard in the past is that Nvidia has been having difficulty in trying to determine where to pitch the RTX 4070, and whether to push out a weaker or stronger offering in this mid-range category (or upper-mid-range, we suppose).

The theory advanced by Moore’s Law is Dead a couple of months back is that Team Green was waiting to see how powerful RDNA 3 cards are before deciding on a final spec, and indeed, that wait may continue – although there’s a pretty firm indication that AMD is squarely targeting the RTX 4080 with its flagship RX 7900 XTX.

In other words, maybe Nvidia now feels confident enough to make the decision to go with a slightly weaker RTX 4070, with a good deal fewer CUDA Cores than the 4070 Ti. It’s a leap, for sure, and we shouldn’t get carried away here, but this rumor does suggest Nvidia isn’t going all out to make the RTX 4070 a graphics card that comes that close to the Ti variant.

Even if this is true, as we touched upon just now, the other specs this RTX 4070 runs with, 12GB VRAM and a 192-bit bus, do match the 4070 Ti (leaked specs, that is). So, even if cores are cut quite brutally, the jury is still out performance-wise, to an extent. And also there’s the possibility that a somewhat lesser-performing graphics card could be in line for a much cheaper price tag than the RTX 4070 Ti, so could end up a good relative value proposition.

Given Nvidia’s current form, though, we find competitive pricing an unlikely prospect, what with all those RTX 3000 cards still to sell through. This rumor has ushered in a good deal of pessimism online regarding the RTX 4070, and we completely understand why, but the key will be price/performance really, and how much Nvidia charges for however powerful this GPU ends up being. Sadly, we’re leaning towards the pessimist’s point of view here, and aren’t holding our breath for a good result on that score.

Perhaps the only thing that might change this is if the pricing on the RTX 4070 Ti comes good, following a price reduction on the RTX 4080 which is rumored to be a possibility. In that case, maybe the RTX 4070 could have a relatively palatable price tag…

Watch this space, and of course the other obvious question is: how much of a delay will there be between the 4070 Ti and 4070? With the former supposedly launching in January, will the base RTX 4070 also be pushed out next month? Given how under the radar it’s been, that perhaps seems unlikely, and previously rumored release dates of the likes of March 2023 might be more on the money for the RTX 4070 (again, particularly if Nvidia needs to make more time to sell through remaining RTX 3000 stock).

Via VideoCardz