The RTX4090 brings again an innovation in the line of RTX cards. The gpu itself is dubbed “Ada Lovelace’ from the famous mathematician.
Looking at the specs we can see that the card shines in compute, especially FP32 by reaching up to 82.58 Tflops, extending another time the raster potentiality of the RTX series by 2.35x compared to the RTX3090.
We have the resume in this table comparing the specifics of the two flagships RTX4090 – RTX4080:
The improvements in capabilities were necessary especially for games like Cyberpunk which needs a much robust implementation of ray tracing that now avoids disorganized workloads but processes data more rapidly.
The light also shines on new generation of tensor cores, the 4th, which now brings full support for DLSS3. Also to mention these cores are entirely derived from the Hopper chip which uses FP8 transformer engine.
The Ray Tracing units are now able to process double amount of ray-triangle intersection, while the OMM and DMM engines will let the rtx units process the data more efficiently in some particular difficult situation like foliage, particles and fences.
Finally the frequencies are up to 2520 Mhz for the core, while the memory of type GDDR6X stand at 1325 Mhz for an effective ~1 TB/s of bandwidth. Board’s TDP max is 450W which is not very bad for a gpu of this class also thanks to the fabrication process of 4Nm TSMC.
Prices are $1599 for the 4090, $1199 for the 4080. Link to buy directly from Nvidia in the links.
Links
RTX40 series
https://www.nvidia.com/en-us/geforce/buy/
