Today we have to review the most powerful graphics card released so far by Nvidia under the GeForce line: The incredible GeForce RTX 3090 . Equipped with more than 10,000 Cuda Cores and a total of 24GB of GDDR6X ram memory, this card is not only ideal for destroying any game we throw at it, but also for productivity tasks thanks to its enormous amount of VRAM and its enormous amount by Cuda Cores that make it the new Titan.
In case the power of the GeForce RTX 3090 was not enough, the particular version that we are going to show you today is the FTW3 ULTRA from EVGA, with factory overclock to 1800MHz, and iCX3 cooling with triple-fan, 9 sensors. of temperature for monitoring and dynamic control of the fans, and an impressive backplate with heatpipes to remove the heat from the memory chips located in the back, being the definitive graphic for enthusiasts.
Before starting to show the graphics and its performance, we want to make a clarification regarding some comments that we saw on social networks in recent days: Although the RTX 3090 is part of the GeForce line, its main orientation is not to run video games, but offer the best possible performance for professional users who need to optimize their workflow and reduce execution times to the minimum possible. It does not come to replace the RTX 2080 Ti, but the Titan RTX, and that is why it has its impressive 24GB of VRAM, which obviously will not be used when playing games, but when working. The star for video games is the GeForce RTX 3080, whose review can be found by clicking here.
Clarified the segmentation of the graph, let’s see its specifications:
RTX 3090 FTW3 ULTRA
|RTX 3090||RTX TITAN|
|Tensor Cores||328 (3rd Gen)||576 (2nd Gen)|
|RT Cores||82 (2nd Gen)||72 (1st Gen)|
|VRAM memory||24GB GDDR6X||24GB GDDR6|
|Memory bus||384 bits||384 bits|
|Memory Frequency||19.5 Gbps||14 Gbps|
|Bandwidth||936GB / s||672GB / s|
|TFLOPs FP32 Pico||37.8||35.6||16.3|
|TFLOPs FP32 Peak Tensioner||284||130.5|
|Number of Transistors||28.3 Billion||13.6 Billion|
|Fabrication process||Samsung optimized process at 8nm||TSMC Nvidia FinFET 12nm|