In a short post published on NVIDIA’s website today, the company announced that it is “releasing” its planned 12GB GeForce RTX 4080 card. The RTX 4080 12GB card, the lowest end of the initially announced RTX 40 series cards, has drawn heavy criticism since it announced the splitting of the 4080 layer between two cards that don’t even share a common GPU. Seemingly bowing to the pressure of these complaints, NVIDIA has removed the card from its RTX 40 series series, as well as canceling its November launch.
NVIDIA Briefing Message It reads as follows:
NVIDIA is not providing any further details about its future plans for the AD104-based video card at this time. However, given the circumstances, it is plausible at the moment that NVIDIA intends to release it at a later date, with a different part number.
|Compare NVIDIA GeForce Specifications|
|RTX 4090||RTX 4080 16GB||RTX 4080 12GB
|Royal Oman Police||176||112||80|
|increase the clock||2520MHz||2505MHz||2610MHz|
|memory clock||21 Gbps GDDR6X||22.4 Gbps GDDR6X||21 Gbps GDDR6X|
|Memory bus display||384 bit||256 bit||192 bit|
|VRAM||24 GB||16 GB||12 GB|
|One accurate performance.||82.6 TFLOPS||48.7 TFLOPS||40.1 TFLOPS|
|Tensor Perv. (FP16)||330 TFLOPS||195 TFLOPS||160 TFLOPS|
|Tensor Perv. (FP8)||660 TFLOPS||390 TFLOPS||321 TFLOPS|
|TDP||450 watts||320 watts||285 W|
|L2 cache||72 MB||64 MB||48 MB|
|GPU||M 102||M 103||M 104|
|number of transistor||76.3 b||45.9 B||35.8 b|
|general engineering||Ada Lovelace||Ada Lovelace||Ada Lovelace|
|manufacturing process||TSMC 4N||TSMC 4N||TSMC 4N|
|launch price||MSRP: $1,599||MSRP: $1199||Was: $899|
By taking a look at the cards’ specs, it’s easy to see why NVIDIA’s base base isn’t for enthusiastic gamers. While both RTX 4080 parts share a common architecture, they do not share a common GPU. Or, for that matter, joint performance.
The 12GB RTX 4080 would have been based, as it were, on the smaller AD104 GPU, rather than the AD103 GPU used on the 16GB model. In practice, this would have caused the 12GB model to provide only about 82% of the previous shader/tensioner throughput, and only 70% of memory bandwidth. Significant performance gap confirmed by NVIDIA pre-launch numbers.
NVIDIA, for its part, is no stranger to overloading a production line in this way, with parts similar to them offering uneven performance and the difference only indicated by their VRAM capacity. This was a practice that began with GTX 1060 . Seriescontinued with RTX 3080 series. However, the performance gap between the RTX 4080’s parts was much larger than anything NVIDIA had done previously, which has drawn more attention to the issues that come from having such disparate parts that share a common product name.
An equal criticism is NVIDIA’s decision to sell the AD104 part as an RTX 4080 card as a start. Traditionally in the NVIDIA product stack, the next card under the xx80 card is some form of the xx70 card. And while the video card names and GPU IDs are essentially random, NVIDIA’s early performance numbers painted a picture of a card that could have performed pretty much as the type of card most people would expect from an RTX 4070 — delivering over 20% higher performance (or More) behind the best RTX 4080, and on par with the flagship of the last generation, the RTX 3090 Ti. In other words, there was a great deal of skepticism within the enthusiastic community that NVIDIA was trying to sell what would have been the RTX 4070 as the RTX 4080, while holding a higher price tag to match.
Anyway, those plans are now officially scrapped. Everything NVIDIA is planning for the AD104-based RTX 40 series card is something only the company knows at this time. Meanwhile, November 16thThe tenth At the launch of the RTX 4080 series, 16GB AD103-based cards will be the only offerings available, with prices starting at $1,199.
“Lifelong beer expert. General travel enthusiast. Social media buff. Zombie maven. Communicator.”