With the RTX 40 series launching less than a month from now, let’s take a look at the specs, look at the highlights, and see how the top GPUs in the RTX 40 family stack up against the RTX 3090 Ti and RTX 3080 Ti.
RTX 40 vs. RTX 3090 Ti and 3080 Ti: Brief Overview
As the above graph indicates, many changes are coming from the RTX 30 series to the RTX 40 series. However, not all of the specs differ from the Ti variants of the RTX 3090 and 3080, and all of the RTX 40 series use GDDR6X memory.
Nvidia claims that the RTX 4080 variants will provide two to four times the performance of the RTX 3080 Ti. Meanwhile, the RTX 4090 Ti is claimed to be two to four times faster than the RTX 3090 Ti.
Other noteworthy similarities include GPUs that support hardware-based ray tracing as well as Nvidia’s superior sampling technology, DLSS. While the RTX 30 series supports DLSS, there are a few differences that we’ll discuss in the next section.
Another interesting note is that there are two variants of the RTX 4080: one with 16GB of GDDR6X memory, which will retail for $1,199, and one with 12GB of GDDR6X memory that will cost $899. It wouldn’t be the first time that Nvidia had done this, as the company released two different versions of its RTX 3080 graphics cards – the launch model included 10GB of memory, while another version released nearly two years later included 12GB of memory.
But memory (and pricing) isn’t the only difference between the two RTX 4080 variants. The 16GB model has more RT, Tensor, and CUDA cores and requires at least 750 watts of power. In contrast, the 12GB model has a higher base and booster clock and requires at least 700W of power.
RTX 40 vs. RTX 3090 Ti and 3080 Ti: Architecture
The RTX 30 series is based on Nvidia’s Ampere graphics architecture, which uses Samsung’s 8nm technology. Meanwhile, the RTX 40 series is built on the new Ada Lovelace architecture, which uses TSMC’s 4nm N4 technology.
Ampere’s architecture is the second generation of RTX, which uses third generation Tensor and second generation RT cores. Ada Lovelace, named after the mathematician and writer, is engineering a 3rd generation RTX, using 4th generation Tensor cores and 3rd generation RT cores. And as you’d expect with a new generation of microarchitecture, every core of the Lovelace RT and Tensor is more powerful than every core of the Ampere RT and Tensor.
Lovelace architecture is also the only graphic architecture that supports the new 3rd generation of DLSS. Nvidia noted in a blog post that DLSS 3 is “powered by the new 4th generation Tensor Cores and Optical Stream Accelerator,” found exclusively in the RTX 40 series.
Of course, these comparisons only tell the story on paper, and we won’t know how the GPUs really compare until we’ve had a chance to rate the RTX 40 series ourselves. However, the new manufacturing process and graphic architecture are showing promising signs that this will likely be a decent leap in performance between two generations of GeForce graphics cards.
Taylor is the Associate Technology Editor at IGN. You can follow her on Twitter Tweet embed.