While everyone is expecting (and hoping) an adequate response from AMD to the GeForce RTX 30 in the consumer market, the first details on what could be one of the new NVIDIA proposals for workstation / professional users based on the newborn Ampere architecture reach us from the network. .
When we talk about this market segment we are obviously referring to the next generation of NVIDIA Quadro cards, a line that had already shown a marked improvement with the first generation Quadro RTX based on Turing GPU architecture.
As for the GeForce RTX 3000 models, the Quadro RTX series will also benefit from the improvements made in terms of efficiency by the new 8nm Samsung production node, as well as the renewed Ampere architecture in terms of performance.
According to reports Moore's Law is Dead, the new flagship NVIDIA Quadro RTX with Ampere architecture will be a monster of power, using a 10,752 Cuda Core GPU and 48GB of VRAM. For a yardstick, the GeForce RTX 3090 counts 10496 Cuda Core and 24GB of GDDR6X memory on 384-bit bus.
From the screenshot above, probably taken from the NVIDIA control panel, we note that this new card will use 16Gbps GDDR6 memories instead of the faster GDDR6X seen on RTX 3080 and RTX 3090. Considering that the memory bus is 384 bit, the width of band reaches 768 GB / s, definitely lower than the approximately 936GB / s of which the top of the range GeForce RTX 3090 (GDDR6X on 384 bit bus) is capable.
The choice of NVIDIA could be dictated by a purely economic issue, however – despite the clear gap between the two types of memory (GDDR6 vs GDDR6X) in terms of performance – it should not be decisive on a type of card like the Quadro RTX.
The data, which partly confirms what emerged this winter, also report a Boost Clock for the GPU of 1860 MHz, a value higher than that of the GeForce RTX 3090 Founders Edition which stops at a Boost Clock of 1695 MHz.