Nvidia showed off its new RTX 4000 graphics cards gaming (GPUs) last night. You can buy one as early as October 12. We only possess the RTX 4080 and 4090 right now, but their cheaper siblings will appear later this year.
The RTX 4090 is now Nvidia’s best discrete GPU, replacing the GTX 1080 Ti. Nvidia says that this new card. Which has the same name as its predecessor, the RTX 3090, is twice as fast as the RTX 3090 Ti and up to four times as fast as DLSS. This GPU is huge in every way. It has 16,384 CUDA cores and a maximum boost speed of 2.52GHz. It features a 384-bit memory interface and 24 GB of GDDR6X video memory.
Since it always needs 450W, it makes sense that you’ll need a power generator as a backup in case your computer breaks down. For operation, you need either one 16-pin PCIe Gen 5 cable or three 8-pin PCIe cables. It will go on sale on October 12 for a starting price of $1599.
The RTX 3090 won’t make anymore once the RTX 4090 comes out. But the RTX 3060, RTX 3070, and RTX 3080 are still good choices. We expect Nvidia’s cheaper RTX 4000 gaming GPUs to come out in 2023.
Nvidia has made two versions of the RTX 4080 gaming GPUs. One with 16 GB of GDDR6X RAM and the other with 12 GB. Compared to the second one, which has 7680 CUDA cores, a 2.61GHz boost clock frequency, and a 192-bit memory interface, the first one has 9728 CUDA cores, a 2.51GHz boost clock, and a 256-bit memory interface. Nvidia says that if you are using an RTX 3080 Ti, the 4080 is always two to four times faster.
The 16 GB model uses 320W when running at full speed, while the 12 GB model only uses 285W. The RTX 4080 with 16 GB of memory will cost $1199. While the version with 12 GB of memory will cost $899. These two items are always auctioned off in the month of November.
Both of the new graphics cards use a smaller version of the Ada Lovelace GPU, which became designed by Nvidia and did make by TSMC using 4nm technology. Aside from extra features like DLSS 3 and an AV1 encoder, this GPU also has fourth-generation Tensor Cores.
Deep Learning Super Sampling (DLSS) is a technology from Nvidia that uses artificial intelligence to make video game framerates faster. Deep Learning Super Sampling version 3 (DLSS 3) is already announced and is only available on RTX 4000 GPUs. Together with an Optical Flow Accelerator, data from many consecutive frames are used to make images outside of the game’s engine. If everything goes as planned, the number of frames per second in the game could go up by a factor of four.
As the frame rate goes up, traditional upsampling and picture reconstruction algorithms run into CPU limits. DLSS 3 will fix this problem. This technology can significantly improve frame rates in all games, especially those that put a lot of stress on the central processing unit.
This feature is now available in more than 35 games, including Cyberpunk 2077, Microsoft Flight Simulator, Dying Light 2, and more.