Right now, we’re waiting for the next generation of graphics cards from NVIDIA (and AMD). For its part, the Chameleon is preparing its launch and has already stopped production of its top-of-the-range models. The aim behind all this is to clear stocks of older cards to avoid internal competition. In any case, there are new speculations surrounding the RTX 5070 Ti, and we know more about its memory and power consumption.
RTX 5070 Ti: what about the configuration?
It’s now an open secret, but we have a very strong suspicion that NVIDIA will be announcing its next-generation graphics card at CES in early 2025. We’re expecting the company to present at least three references: the RTX 5090, the RTX 5080 and the RTX 5070 Ti.
With regard to the latter, an initial rumour suggested its configuration, as we mentioned in this article. However, new information has surfaced to clarify the memory part of the card in particular.
So, with this new RTX xx70 Ti, we can expect a larger VRAM capacity than the current generation. Rumours suggest 16GB of GDDR7 on a 256-bit memory bus. This would mean memory operating at 28 Gbps, giving us a total bandwidth of 896 GB/s! Compared with the RTX 4070 Ti, this represents a 33% increase in VRAM capacity and a 77.7% increase in bandwidth, which is quite a gain. Compared with the Super, there’s no increase in memory, but a 33.2% gain in bandwidth.
However, this rumour lends further credence to the fact that the RTX 5070 Ti does indeed feature a neutered GB203. This would benefit from a total of 8960 cuda cores, i.e. 6% more than an RTX 4070 Ti Super. Nevertheless, with the improvements made at Blackwell, we’re expecting performance gains of over 6%!
All that remains is to talk about power consumption, as this is a subject of much debate. Some people were talking about a TBP of 350W, but Kopite7Kimi says that the official TBP is actually 285W. The 350W would correspond to the power consumption of certain customised versions.