- RTX 5070 Ti could have 16GB of VRAM and use 350W of power
- It may run with the same GPU as the RTX 5080, as previously rumored
- Also it might be released before the vanilla RTX 5070, very early next year
Nvidia’s RTX 5070 Ti is the subject of a new leak sharing a bunch of juicy details about the GPU – including the fact that this graphics card is supposedly set to arrive before the vanilla edition of the RTX 5070.
Wccftech has spoken to sources who’ve outlined some fresh specs for the purported RTX 5070 Ti, as well as confirming some of the info divulged by previous leaks (as ever, keep shovelfuls of skepticism on hand).
We’re told that past rumors of the RTX 5070 Ti are correct in asserting that it’ll use the GB203 chip, the same GPU as in the RTX 5080, but obviously it will be a cut-down version. (In theory, GB203-300, with 8,960 CUDA cores, as opposed to the full loadout on GB203-400 with the RTX 5080).
We didn’t hear anything about the VRAM configuration in the last rumor dump, but Wccftech’s sources believe the RTX 5070 Ti will sport 16GB of GDDR7 with a 256-bit memory bus. The video RAM will be 28Gbps giving a total memory bandwidth of 896GB/s, closing in on that 1TB/s mark, which is pretty impressive.
Apparently, power usage will run at 350W, which is 50W more than previously believed. There’s a twist here, though, as leaker Kopite7kimi has chipped in on X to note that the “latest data shows 285W” – while admitting that 350W is one of the possible configurations. In other words, this isn’t yet decided, which is certainly plausible.
Finally, Nvidia is theoretically going to launch this RTX 5070 Ti as the third Blackwell GeForce graphics card, after the RTX 5090 and 5080 – meaning it’ll actually arrive ahead of the RTX 5070 itself.
Analysis: A chip off big brother's block
It’s a bit odd for a Ti version (or a Super, if that’s what it turns out to be) to arrive before the vanilla flavor of a graphics card, although it has happened before occasionally.
Sometimes this might be related to manufacturing nuances and chip yields, and with the RTX 5070 supposedly using that same GB203 GPU as the RTX 5080, that tracks in this respect. (The RTX 5070 is theoretically a different chip entirely, GB205). At any rate, at this point, Nvidia wouldn’t surprise me if it delivered a Ti Super off the bat…
Speculation about current launch plans has the RTX 5090 and 5080 turning up at CES 2025, and the RTX 5070 – or this 5070 Ti, as is the argument here – arriving perhaps at CES as well, or a bit later in January.
Wccftech also claims that Nvidia is going to launch all its next-gen Blackwell graphics cards in the first half of 2025, and most of them in Q1. So that’d suggest a quick follow-up for the 5070 (or Ti, whichever doesn’t come out in January), and the RTX 5060 debuting perhaps a lot sooner than expected (seeing as the rumors around that have been much scarcer – suggesting it’s still a fair way down the line).
It’s possible that Nvidia fears what AMD might bring out in the way of lower-mid-range GPUs with RDNA 4, and could’ve made a decision to push the RTX 5060 through faster in anticipation of competing better in that space.
Release timings aside, the specs of the RTX 5070 Ti are making me wonder if this might be the new GPU for me. That memory bandwidth of almost 900GB/s is a third faster than the current top dog of the RTX 4070 spins, the RTX 4070 Ti Super.
On the other hand, power usage creeping up isn’t so great, of course – the RTX 4070 Ti Super chugs 285W, so 350W is a fair old step up from that. If it happens, of course, because as noted above, Nvidia could still be sticking with 285W. Team Green is likely still working out the efficiency to performance balancing act – and pricing will be key here, too, as ever.
I really hope that Nvidia can keep some semblance of a lid on that price tag, as if so, there’s a strong possibility that this will be my big GPU upgrade in the New Year. (Something I’ll be writing a feature about very soon, as I’ve got a very specific reason for this graphics card upgrade – so stay tuned).