NVIDIA is set to deviate from its tradition of high PCIe lane counts for mid-range GPUs with the forthcoming launch of the GeForce RTX 40-series, as revealed by VideoCardz. Employing the AD106 silicon, the eagerly awaited GeForce RTX 4060 Ti will be equipped with a PCI-Express 4.0 x8 interface.
This decrease in PCIe lane count aims to simplify board design for AIC partners by reducing PCB traces and related SMDs for each PCIe lane. NVIDIA documents shared with board partners reveal that the RTX 4060 Ti graphics card, featuring an increased memory buffer, will use a different GPU variant and demand higher power. However, these specifications are based on proposed SKUs and may change since the GPUs have not been supplied yet.Although the VRAM capacity differs, with the 16GB variant having SKU 363 and the 8GB model SKU 361, the board design (PG190) will stay the same. NVIDIA found it necessary to utilize a slightly modified GPU variant for the RTX 4060 Ti, indicating minor design adjustments while maintaining the same CUDA core count.
The RTX 4060 Ti's power consumption is expected to rise with the doubled memory size, but only by 5W compared to the 160W specification of the 8GB model. The memory clock for the 16GB variant remains unconfirmed, but it is reasonable to assume that it will stay at 18 Gbps at this stage.
After encountering launch issues with the RTX 4080 12GB, NVIDIA is exercising caution when designing different cards under the same name. They aim to preserve similar specifications between both SKUs, with current discussions centering on capacity, GPU variant, and TDP.
Based on available information, the RTX 4060 Ti is projected to launch in the second half of July, while the 8GB model is expected to debut on May 24th. NVIDIA intends to release both 8GB and 16GB versions of the GeForce RTX 4060 Ti graphics card, both relying on the 5nm "AD106" silicon. The 8GB model will use the ASIC code "AD106-350," whereas the 16GB model will employ the ASIC code "AD106-351."
It's important to note that this TDP refers to the silicon itself and not the TGP (typical graphics power), which accounts for the power consumption of the entire board. The higher-density memory on the 16GB model will likely result in a higher TGP. NVIDIA is anticipated to use four 32 Gbit (4GB) GDDR6 memory chips to achieve the 16GB capacity, as opposed to eight 16GB chips.