The second solution is the residual blood AD104 core, 7168 CUDA cores, 160-bit 10GB video memory, 250W power consumption, and a 3DMark TSE score of about 10,000 points.
The RTX 4080 is relatively simple. The difference between the two solutions is mainly in the PCB and video memory capacity.One is a 12-layer, 256-bit 16GB GDDR6X, and the other is a 10-layer 192-bit 12GB GDDR6X.
The other specifications are the same,The rumors are all 9728 CUDA cores, 23GHz memory frequency, 340W power consumption, and 3DMark TSE about 15,000 points.
As for how to choose in the end, it may depend on product positioning, competition situation, or how Lao Huang swings his knife, and even does not rule out simultaneous release.
In addition, the RTX 4080 FE public version of the real card has been leaked a few days ago, indicating that the overall design has been basically finalized, and only the final fine-tuning is left.
In addition,The RTX 4090 will use a 14-layer PCB design, and the video memory is determined to be 384-bit 24GB GDDR6X, this will not change.