NVIDIA announced new GeForce RTX 40 family video cards based on graphics processors with Ada Lovelace architecture. At the start, the older models of the line are presented – GeForce RTX 4090 and GeForce RTX 4080. As previously predicted, GeForce RTX 4080 will receive two modifications, equipped with 16 GB and 12 GB of memory. However, these adapters will differ not only in the amount of local RAM. We focus on the technical specifications of GeForce RTX 4080 16/12 GB.

The younger model will receive a GPU with activated 7680 computing units instead of 9728 (+27%) in the older one. The number of other functional modules will be proportionally reduced. At the same time, in the case of RTX 4080 12 GB, the graphics processor itself will receive increased operating frequencies – 2.31/2.61 GHz versus 2.21/2.51 GHz.

GeForce RTX 4080 16 GB GeForce RTX 4080 12 GB GeForce RTX 3080 12 GB GeForce RTX 3080 10 GB GeForce RTX 3070 8 GB
CUDA calculators 9728 7680 8960 8704 5888
GPU clock frequencies 2.21/2.51 GHz 2.31/2.61 GHz 1.26/1.71 GHz 1.44/1.71 GHz 1.5/1.75 GHz
Memory size 16 GB GDDR6X 12 GB GDDR6X 12 GB GDDR6X 10 GB GDDR6X 8 GB GDDR6
Memory bus 256 bits 192 bits 384 bits 320 bits 256 bits
Effective memory frequency 23,000 MHz 21,000 MHz 19,000 MHz 19,000 MHz 14,000 MHz
Memory bandwidth 736 GB/c 504 GB/c 912 GB/c 760 GB/c 448 GB/c
Video card power consumption (GCP) 320 W 285 W 350 W 320 W 220 W

NVIDIA has not yet officially confirmed which chips will be used in both cases. There is an assumption that for the older model it is the AD103-300, while the younger one will be based on the more compact AD104-400. It makes sense in theory, especially considering that both versions offer different PCB designs and even layer counts – 12-layer PCBs for the 16GB models and 10-layer PCBs for the 12GB adapters. However, this still requires clarification from the developers.

In addition to the volume of local memory, the bit rate of the bus will also differ. The GeForce RTX 4080 16 GB received a 256-bit data bus and also uses GDDR6X memory chips with an effective frequency of 23,000 MHz. This combination allows you to get a memory bandwidth (BDP) of only 736 GB/s. But the GeForce RTX 4080 12 GB is more modest here. Video cards of this modification are already equipped with a 192-bit bus, and GDDR6X chips work at an effective 21,000 MHz. In this case, the bandwidth is only 504 GB/c (-46%).

We remind that the base GeForce RTX 3080 10 GB offers a 320-bit memory bus, and the GDDR6X with 19,000 MHz is able to provide a RAM of 760 GB/c, while the GeForce RTX 3080 10 GB variant has a 384-bit resolution and bandwidth at the level of 912 GB/c. Even the GeForce RTX 3070 Ti offers 608 GB/c and only the RTX 3070 has 448 GB/c.

Of course, the bit rate of the memory bus and even the total bandwidth are only separate parameters of the video card, which to some extent affect the final performance, but are not decisive. Nevertheless, FPS has always been a rather important indicator that affects the number of fps, especially in modes with increased resolution and maximum graphics quality, when large amounts of memory are used for storing and processing textures.

In the case of the GeForce RTX 4080 16/12 GB, the graphics cards will differ in almost all key characteristics. As you can see, the amount of memory is only one of them and far from the main one. There are so many differences that according to the canons of previous generations, they would be enough to assign similar adapters to different segments, or at least to different levels in the graphic hierarchy.

GeForce RTX 4080 12 GB for $899 received a 192-bit memory bus. Is it definitely not a GeForce RTX 4070?

Why did the developers decide to combine them in one series this time? Probably there were some marketing tricks here. The GeForce RTX 4080 16/12GB will arrive in November with recommended price of $1,199/$899. On the face of it, in all fairness, the 33% difference in price potentially reflects the supposed difference in performance. However, if you look at the cost of the previous generation devices, it is obvious that the models of the GeForce RTX 40 series have significantly increased in price.

We will remind that the GeForce RTX 3080 10 GB had a starting price of $699, and the GeForce RTX 3080 12 GB version – $799. The cost of the latter will subconsciously be compared with that of the GeForce RTX 4080 12 GB, and in this case the price increase does not look so tragic. +$100 for higher performance and new features is quite normal. But the price difference of one and a half times ($1199 vs. $799) with the older RTX 4080 model is perceived quite differently.

Therefore, the GeForce RTX 4080 12 GB belongs specifically to the RTX 4080 series, and not to the RTX 4070, which, taking into account the technical characteristics, would be quite logical. But in this case, the price tag of $899 would already look inadequate against the background of $499 for the GeForce RTX 3070. In the current conditions, the future GeForce RTX 4070 for $650-700 may already look like a completely rational option against the background of older models. In this way, NVIDIA increases the total cost of video cards of the new series, trying to smooth out the perception of the effect of the price increase with various tricks.

In general, the situation with the GeForce RTX 4080 16/12 GB returns to those distant times when some users tried to rank the performance of video cards, focusing solely on the amount of memory. It is not surprising, but after so many years, a similar approach turned out to be working within one family. The older model has 33% more memory, costs as much more and, apparently, will be about a third more productive than the younger one. Everything is coming together.

GeForce RTX 4080 12 GB for $899 received a 192-bit memory bus. Is it definitely not a GeForce RTX 4070?

As for the performance of both GeForce RTX 4080 modifications, the interpretation of the developers should be treated with a certain amount of skepticism. For demonstrations, tests use projects or even pre-release builds with optimizations for the new DLSS and RT algorithms, where the new products show a multiple advantage over the previous generation models. Here it is worth waiting for the results in a wider range of the most popular games. There is still one and a half months left till GeForce RTX 4080 start of sales. Perhaps during this time more information will appear for analysis. But the first estimates can already be made with the start of sales of the older GeForce RTX 4090 24 GB model on October 12.