Nvidia just revealed the full potential of Ada Lovelace, and the RTX 4090 Ti GPU could boast a 70% higher CUDA core count than RTX 3090 Ti. That same increase applies to the AD102 chip’s ray tracing and Tensor cores, but its boost clock speeds match the recently announced GeForce RTX 4090.
During an Nvidia virtual press event, the company revealed max RTX 4000 GPU specs, highlighting the differences between Lovelace and Ampere. While the RTX 4090 isn’t equipped with a full AD102 chip, souped-up variants like the RTX 4090 Ti might pack all the available 18,434 CUDA cores, 144 third-gen RT cores, and 576 fourth-generation Tensor cores.
| Nvidia AD102 GPU | |
| VRAM (GDDR6X) | TBC |
|---|---|
| Boost clock | 2.52GHz |
| Base clock | 2.23GHz |
| CUDA cores | 18,434 |
| Ray tracing cores | 144 (Third gen) |
| Tensor Cores | 576 (fourth gen) |
| Bus Width | 384-bit |
Comparatively, the full Ada Lovelace GPU features 71% more cores than its Ampere GA102 predecessor. However, opting for an RTX 4090 Ti seemingly won’t get you faster speeds over and above the RTX 4090, as both come with a 2.5GHz boost clock ceiling.
While the RTX 4090 will pack 24GB memory, earlier rumours suggest the RTX 4090 Ti could pack 48GB VRAM. We’ll need to wait and see whether that particular SKU actually makes it to market, but it may save enthusiasts splashing out on workstation cards like the RTX 6000.
On Tuesday, Nvidia officially announced the RTX 4080 and RTX 4090, two next-gen graphics cards that wield the new Ada Lovelace architecture. In addition, the green team unveiled a new version of its AI upscaler, but you’ll only be able to use DLSS 3.0 on RTX 4000 GPUs.
Nvidia hasn’t announced a best graphics card contender kitted out with a full AD102 GPU, but it’ll potentially arrive later in the RTX 4000 series life cycle. There’s also still no sign of either the RTX 4060 or RTX 4070, but two RTX 4080 models are set to arrive in November. The 12GB VRAM flavour is already causing a stir among enthusiasts, as it falls behind its 16GB counterpart in terms of specs and performance.