In a brief post published on NVIDIA’s website today, the company announced that it will no longer produce the GeForce RTX 4080 12GB graphics card. The RTX 4080 12GB, the lowest-end of the initially announced RTX 40 series cards, drew significant criticism for splitting the 4080 tier between two cards that did not even share a common GPU. NVIDIA has yielded to the pressure of these complaints by removing the card from their RTX 40 series lineup and canceling its November launch.
The brief message from NVIDIA reads as follows:
The RTX 4080 12GB is an excellent graphics card, but its name is incorrect. Two GPUs with the designation 4080 are confusing. Therefore, we press the “cancel launch” button on the 4080 12GB. The RTX 4080 16GB is a phenomenal graphics card expected to excite gamers worldwide on November 16. If the long lines and enthusiasm for the 4090 are any indication, the 4080’s reception will be phenomenal.
NVIDIA provides no additional information regarding their plans for the AD104-based video card at this time. Nevertheless, given the circumstances, it is reasonable to assume that NVIDIA will launch it at a later date with a different part number.
|NVIDIA GeForce Specification Comparison|
|RTX 4090||RTX 4080 16GB||RTX 4080 12GB
|Memory Clock||21Gbps GDDR6X||22.4Gbps GDDR6X||21Gbps GDDR6X|
|Memory Bus Width||384-bit||256-bit||192-bit|
|Single Precision Perf.||82.6 TFLOPS||48.7 TFLOPS||40.1 TFLOPS|
|Tensor Perf. (FP16)||330 TFLOPS||195 TFLOPS||160 TFLOPS|
|Tensor Perf. (FP8)||660 TFLOPS||390 TFLOPS||321 TFLOPS|
|Architecture||Ada Lovelace||Ada Lovelace||Ada Lovelace|
|Manufacturing Process||TSMC 4N||TSMC 4N||TSMC 4N|
|Launch Price||MSRP: $1599||MSRP: $1199||Was: $899|
Examining the cards’ specifications makes it clear why NVIDIA’s core base of enthusiast gamers was not pleased. While both RTX 4080 components shared the same architecture, they did not share a GPU. Alternatively, ordinary performance.
As it stood, the RTX 4080 12GB would have utilized the smaller AD104 GPU instead of the AD103 GPU used in the 16GB model. In practice, the 12GB model would have delivered only 82% of the shader/tensor throughput and 70% of the memory bandwidth of the 16GB model. A significant performance gap that NVIDIA’s figures before the product’s release have confirmed.
NVIDIA, on the other hand, is not unfamiliar with overloading a product line in this manner, with similarly-named components delivering unequal performance and the difference being denoted solely by their VRAM capacity. This practice began with the GTX 1060 series and persisted through the RTX 3080 series. However, the performance gap between the RTX 4080 parts was significantly larger than anything NVIDIA has ever done before, bringing a great deal more attention to the issues associated with having such dissimilar details share the same product name.
Equally criticized was NVIDIA’s initial decision to market an AD104 component as an RTX 4080 card. Traditionally, the next card below the xx80 card in NVIDIA’s product stack is an xx70 card. And while video card names and GPU identifiers are essentially arbitrary, NVIDIA’s early performance numbers painted a picture of a card that would have performed similarly to what most people would expect from the RTX 4070 – delivering performance up to 20% (or more) behind the superior RTX 4080, and on par with the previous-generation flagship, the RTX 3090 Ti. In other words, there has been a great deal of suspicion within the enthusiast community that NVIDIA was attempting to market the RTX 4070 as the RTX 4080, along with a price increase to match.
Regardless, these plans have been officially abandoned. What plans NVIDIA has for their AD104-based RTX 40 series card is currently known only to the company. On the 16th of November, when the RTX 4080 series launches, only 16GB AD103-based cards will be available, with prices beginning at $1199.