We have a date for Nvidia’s next major GPU dump. It’s 3rd of January next year and a GeForce Special Event keynote for the CES tech jamboree in Las Vegas (opens in new tab), albeit Nvidia will be conducting the event remotely.
Beyond describing the event as detailing “the latest technologies, products and software from GeForce”, Nvidia isn’t getting more specific. However, we’re expecting to see two main announcements at the event.
On the desktop side, very likely we’ll finally see what Nvidia has done with the cancelled RTX 4080 12GB model (opens in new tab). Odds are it will be rebranded RTX 4070 Ti and be announced on the 3rd at that CES event.
Given the RTX 4080 12GB boards were already in mass production when Nvidia had a change of heart, we’re not expecting much by way of a change in specifications in the transition to RTX 4070 Ti branding.
So, that’s 7,680 shaders, aka CUDA cores in Nvidia parlance, 240 texture units, 80 ROPs and 12GB of memory running over a 196-bit bus. Roughly speaking, those numbers are around 40 to 45 percent of the flagship RTX 4090 GPU (opens in new tab). While it’s possible Nvidia may tweak some of those specs for the purported RTX 4070 Ti via a video BIOS update, we’re not expecting a dramatic deviation.
Instead, the big unknown is pricing. The RTX 4080 12GB was priced at $899 and the assumption is that the RTX 4070 Ti will be cheaper. But how much cheaper? Odds are, we’ll know on January 3.
The other major announcement at the CES event is expected to be mobile variants of Nvidia’s RTX 40-series GPUs, otherwise known as Ada Lovelace. We’re expecting to see RTX 4080 and 4070 branded mobile GPUs announced at minimum, with the possible addition of a mobile RTX 4090 and 4060 mobile chips to top and tail the range.
Based on the best currently available information, the mobile lineup would therefore look something like this. At the very top, the RTX 4090 will use the same AD103 GPU as the desktop 4080 card and pack 9,728 shaders, 304 texture units and 112 ROPs, all running over a 256-bit memory bus. Clock speeds would be significantly lower than the desktop 4080 at approximately 1.4GHz.
The RTX 4080 and 4080 Ti mobile will step down to the AD104 GPU briefly seen in that cancelled 4080 12GB desktop board, with the Ti model offering much the same hardware features as the 4080 12GB desktop but running at lower clock speeds. Subtract roughly 15 percent from the shader, texture and ROP counts to arrive at the non-Ti model. The smaller AD104 chip should clock a little higher in a mobile implementation and might flirt with as much as 2GHz in higher performing installations. However, the usual lower-power Max-Q variants for thin-and-light will inevitably be much slower.
As for the RTX 4070 mobile, that’s likely to be based on the as-yet unannounced AD106 chip. Specs will likely fall around 4,608 shaders, 144 textures and 48 ROPs, plus a 128-bit memory bus. Finally, the RTX 4060 could use another unannounced chip, AD107, with further cut-down specs, likely to be around 2,560 to 3,072 shaders, 80 to 96 texture units and around 32 ROPs. Mercifully, AD107 is currently expected to stick with a 128-bit memory bus rather than stepping down even further to a 96-bit bus.
So, there you have it: what to expect on January 3rd. Let us know below whether that sounds exciting or if it just adds to the frustration at the latest GPU launches.