Although NVIDIA officially announced the stunning GeForce RTX 4090 flagship graphics card during the GTC 2022 keynote, the RTX 4080 product line with 16/12GB two-speed memory bit width still puzzled many people – especially after considering The video memory bit width is only 192-bit, which is only half of the previous generation RTX 3080 12GB graphics card (384-bit).
So when equipped with 21Gbps GDDR6X memory, the bandwidth of the RTX 4080 12GB graphics card is only 504 GB/s.
In contrast, the RTX 3080 12GB graphics card with 19Gbps GDDR6X memory has a bandwidth of up to 912 GB/s.
Even the RTX 3080 10GB model, which was cut to 320-bit @ 10GB of VRAM, still has 760 GB/s of bandwidth.
(via TechPowerUp)
Meanwhile, the RTX 4080 16GB graphics card has a 256-bit bit width paired with faster 23Gbps GDDR6X memory. But only 736 GB/s of bandwidth, which is also inferior to the RTX 3080 10GB model.
In fact, even the RTX 4090, which is positioned as Ada Lovelace’s flagship SKU, has only the same memory bandwidth as the previous-generation RTX 3090 Ti (1008 GB/s) and slightly higher than the non-Ti-suffixed RTX 3090.
Of course, there are many factors that affect the final performance of the GPU – at least the memory capacity of the RTX 40 series starting lineup has been greatly improved.
For example, the RTX 4080 12GB provides 20% more memory than the RTX 3080 10GB, and the RTX 4080 16GB is also 33% higher than the RTX 3080 12GB.
It is a pity that in the environment of cryptocurrency mining, the previous generation RTX 30 series still has a considerable backlog of inventory, and the buying momentum of the new card market is bound to be hit by a large number of selling mining cards.
In this case, the first lineup of the RTX 40 series will inevitably be criticized for deliberately cutting specifications and raising prices to set off the so-called “cost-effectiveness” of the RTX 30 series, and the time to market of the mainstream graphics card models of the RTX 40 series may also be further delayed.
