NVIDIA Readies GeForce GTX 1660 Ti Based on TU106, Sans RTX
It looks like RTX abilities received’t accomplish it to sub-$250 market segments as the GPUs don’t seem to be swiftly sufficient to address staunch-time raytracing, and it makes cramped financial sense for NVIDIA so that you just would possibly add billions of additional transistors for RT cores and tensor cores. The firm is hence carving out a sub-class of “Turing” GPUs beneath the TU11x ASIC series, which is able to strength contemporary GeForce GTX family SKUs, similar to the GeForce GTX 1660 Ti, and diversified GTX 1000-series SKUs. These chips provide “Turing Shaders,” which are typically “Turing” CUDA cores which have the IPC and clock-speeds rivaling existing “Turing” GPUs, however no RTX or AI capabilities. To sweeten the deal, NVIDIA will equip these playing cards with GDDR6 memory.
The GeForce GTX 1660 Ti will absolute confidence be slower than the RTX 2060, and be primarily based on a contemporary ASIC codenamed TU116. In accordance with a VideoCardz file, this 12 nm chip packs 1,536 CUDA cores primarily based on the “Turing” architecture, and the same precise memory setup as the RTX 2060, with 6 GB of GDDR6 memory for the duration of a 192-bit wide memory interface. The dearth of RT cores, tensor cores, and a decrease CUDA core depend would possibly accomplish the TU116 a enormously smaller chip than the TU106, and one thing NVIDIA can have the funds for to promote at sub-$300 mark-factors similar to $250. The GTX 1060 6 GB is conserving the fortress for NVIDIA on this section, moreover diversified GTX eleven-series SKUs similar to the GTX 1070 once in a whereas shedding below the $300 at outlets’ mercy. AMD no longer too long previously improved its sub-$300 portfolio with the Radeon RX 590, which convincingly outperforms the GTX 1060 6 GB.