Nvidia is quietly preparing to launch its next-generation Quadro RTX GPUs with the Ampere architecture for the professional sector, GPUs due to launch in October. Now it has been discovered that the top of the range model will use GA102 silicon to the fullest, although luckily that translates to just 10752 CUDA Cores, which now seems like a very small thing thanks to the existence of the Nvidia GeForce RTX 3090 has 10496 CUDA Cores, which is the top of the range GPU for the gaming segment.
In this way, the top-of-the-range Quadro RTX barely has an extra 256 CUDA Core, although the biggest difference is that it doubles the amount of memory it offers, albeit slower, that is, 48 GB GDDR6 (vs GDDR6X) at a speed of 16 GHz (vs 19.5 GHz). Due to the use of a 384-bit memory interface, this GPU achieves a bandwidth of 768 GB / s, which translates to offering 8 GB / s more than the GeForce RTX 3080 and 168 GB / s less than the GeForce RTX 3090.
The Nvidia Quadro RTX 8000, the top of the range GPU for the professional field with the Turing architecture, went on sale at a price of around 7,000 euros, so we will have to wait and see if this range of GPUs will also see its price drastically reduced.