GeForce RTX 4090 arrives with 24 GB of memory and RTX 4080 has two versions – APK Games

This Tuesday (20) marks the official announcement of the long-awaited video cards GeForce RTX 4000. The line was revealed during the conference GTC 2022, by Jensen Huang, CEO of Nvidia. As usual, the promise of more graphics performance is present here. In this regard, the highlight goes to the powerful GeForce RTX 4090. She arrives with the models GeForce RTX 4080.

GeForce RTX 4090 (image: publicity/Nvidia)

Nvidia GeForce RTX 4090: 24GB memory

The GeForce RTX 4090 arrives with “only” 24 GB of GDDR6X memory. The datasheet also includes a frequency of 2.23 GHz and 16,384 Cuda cores to do what we most expect from a high-end video card: impressive graphics in games.

According to Nvidia, the GeForce RTX 4090 can be two to four times faster than the GeForce RTX 3090 Ti in gaming.

Of course, that promise comes with some, shall we say, side effects. For starters, the card is generously sized, requiring the user to have a large case. A source of respect is also a requirement, after all, the consumption standard here is 450 W. Well, none of this is surprising for gamers, right?

GeForce RTX 4090 (image: publicity/Nvidia)
GeForce RTX 4090 (image: publicity/Nvidia)

Nvidia GeForce RTX 4080 in two versions

But the GeForce RTX 4090 doesn’t come alone. Nvidia also announced the GeForce RTX 4080 in versions with 12GB and 16GB of GDDR6X memory.

Again, Nvidia makes a comparison to the previous generation. The 16GB GeForce RTX 4080 can be two to four times faster than the RTX 3080 Ti.

Here, we find 9,728 Cuda cores and a base frequency of 2.21 GHz, in addition to a base consumption of 320 W. This is for the version with 16 GB of memory.

The version with 12 GB of memory comes with 7,680 Cuda cores, base frequency of 2.31 GHz and standard consumption of 285 W.

They are configurations so different that the last model could have received another name, do you agree? Well, it seems that, originally, the 12 GB RTX 4080 would be called the RTX 4070.

GeForce RTX 4080 (image: publicity/Nvidia)
GeForce RTX 4080 (image: publicity/Nvidia)

Ada Lovelace Architecture and DLSS 3

Powerful numbers are only part of this story. The other serves for ada lovelace. Nvidia paid homage to renowned 19th century math by placing its name on the third-generation RTX architecture.

Well, the architecture is based on a new 5nm manufacturing process from TSMC called 4N (yes, it’s just to confuse us). Up to 76 billion transistors can be found here, up from up to 28 billion in the previous generation. In the words of Nvidia, “Ada is the world’s most advanced GPU architecture”.

The attributes of the new architecture also include fourth-generation Tensor cores (for artificial intelligence tasks). These have up to five times the data transfer rate compared to the RTX 3000 series cores.

The Ada Lovelace architecture also makes room for the also new technology DLSS 3, which uses deep learning to improve graphics rendering. One of the promises of the new version is to combat performance issues caused by CPU bottlenecks, explains Huang.

In the RTX 4000 series, the DLSS 3 technology is exclusive, not least because it requires the Flow Engine, a hardware component that is only available for new cards. The technology should reach other lines in the not too distant future, but via software implementation.

The ray tracing standard, in version 3, also appears here, as it should be.

YouTube video

Prices and availability

In the United States, the new RTX GPUs officially go on sale on October 12. The suggested prices are these:

  • GeForce RTX 4090: US$ 1,599
  • GeForce RTX 4080 (16GB): US$ 1,199
  • GeForce RTX 4080 (12GB): $899

There is still no release information in Brazil. But we can expect that, at the latest, the first units will arrive in the first quarter of 2023. That’s enough time for interested parties to sell their cars.

GeForce RTX 4090 and RTX 4080 — data sheet

RTX 4090 RTX 4080 (16GB) RTX 4080 (12GB)
Cuda Nuclei 16,384 9,728 7,680
Frequency 2.23 GHz 2.21 GHz 2.31 GHz
boost 2.52 GHz 2.51 GHz 2.61 GHz
Memory 24GB (GDDR6X) 16GB (GDDR6X) 12GB (GDDR6X)
memory interface 384 bit 256 bits 192 bits
Consumption 450 W 320 W 285 W
technologies HDMI, DisplayPort (3), 7680×4320 pixel resolution, PCIe 4, 3rd generation ray tracing, DLSS 3 HDMI, DisplayPort (3), 7680×4320 pixel resolution, PCIe 4, 3rd generation ray tracing, DLSS 3 HDMI, DisplayPort (3), 7680×4320 pixel resolution, PCIe 4, 3rd generation ray tracing, DLSS 3

Leave a Comment