As per the earlier rumours and predictions, the RTX 40 Series is based around a brand new GPU architecture, Ada Lovelace, named after the 19th century British mathematician. On paper, Lovelace architecture is based on an all-new 4nm die lithography, courtesy of NVIDIA’s long-time semiconductor partner, TSMC. It still retains the same GDDR6X memory standard, but this time round, the module is packing 76 billion transistors. Lovelace also features new Shaders, 3rd generation Ray-Tracing (RT) Cores, and 4th generation Tensor Cores to drive more intensive deep learning executions, through Optical Flow Accelerators. All features that the RTX 40 Series can and will be made to take full advantage of. Breaking down the features, the new Streaming Multiprocessors are capable of pulling off 90 Shader TFLOPs at twice the power efficiency of its predecessor. Moving on, the 3rd generation RT Cores are capable of doing up to 200 RT TFLOPs of calculations, as well as double the ray-triangle intersection computations. As for its Deep Learning capabilities, we’re looking at 1400 Tensor TFLOPs or 1.4 PFLOPs of calculations. In addition to Lovelace and the RTX 40 Series, NVIDIA also announced its 3rd generation of its AI upscaling technology, otherwise known as DLSS 3. To cut to the chase, the feature is an improvement over the last generation and utilises a new form of neural rendering to really, really push the framerates up. All from lower resolution like Full HD and up into the realm of 4K. New tasks and executions include Real-Time Path Tracing, Shader Execution Reordering, and Opacity Micro-Maps, to name a few. As for what RTX 40 Series cards we can look forward to, NVIDIA announced the RTX 4090 and two flavours of the RTX 4080. For the former, the card ships out with the 24GB GDDR6X and retails at US$1599 (~RM7288). Performance-wise, NVIDIA claims that the card will be anywhere between 2x to 4x faster than the RTX 3090 Ti. The card will be available starting 12 October. As for the other Lovelace card, the RTX 4080 will be available in both 12GB and 16GB flavours. The 12GB SKU will retail from US$899 (~RM4097), while the 16GB model retails from US$1199 (~RM5465). Depending on the SKU, NVIDIA says consumers can look forward to a performance uplift between 2x and 4x of the RTX 3080 Ti. As for avaialability, NVIDIA says that the cards will be available in the month of November.
