According to a report from ITHome, NVIDIA is prepping the GeForce RTX 3080 Ti for a mid-May launch with an increased price tag of 7,999 Yuan or $1,199 (USD). The die has been updated from the earlier planned GA102-250 to GA102-202/302 with a shader count of 10,240 cores and a bus width of 384-bits. The GPU will be paired with 12GB of GDDR6X memory, just 2 GB more than the RTX 3080 and half as much compared to the RTX 3090:
As already reported earlier, the new die has been designed to reduce the Ether mining performance with a combined firmware-hardware block in place. The new firmware for the GPU runs graphics workloads but CUDA and ether-mining don’t work with the new beta driver. On the other hand, if the old firmware is flashed to a new chip, the device won’t boot. Strangely, the device ID of the older and newer chips is the same which allows flashing the new BIOS to older QS, but even that doesn’t work in all cases. If the GPUs have the same PCI Device ID, then the override command can be used to flash a custom BIOS as many enthusiasts do at present. However, similar to the A and non-A Turing dies, the device ID is different for the new dies, preventing cross-flashing of the firmware from the older to the newer GPUs.