NVIDIA Officially Announces the RTX 40 Series, Here’s What You Need to Know

This morning, NVIDIA officially announced the RTX 40 series and Ada Lovelace graphics card architecture. With that we have three new graphics cards, the RTX 4090, RTX 4080 16GB and RTX 4080 12GB. The flagship GPU in this series is the AD102 Ada Lovelace GPU. This GPU measures approximately 600mm2 and uses the TSMC 4N process node, an optimized version of TSMC’s 5nm (N5) node designed specifically for NVIDIA. The GPU will contain 76 billion transistors.

NVIDIA GeForce RTX 4090

The flagship graphics card at launch will be the GeForce RTX 4090. This card will be 2-4x faster than the current GeForce RTX 3090 Ti and will feature 24 GB of GDDR6X memory.

The card uses 128 SMs out of 144 SMs on the AD102 GPU, so you get a total of 16,384 CUDA cores. The GPU will also feature 96MB of L2 cache and 384 ROPs. The card is boosted to 2520 MHz, giving the card a single-precision performance of 82.6 TFLOPS or 2.3x more than the RTX 3090. The 24 GB GDDR6X memory has a peak bandwidth of 1 TB/s. In terms of power, the card has a standard TDP of 450W.

The Founder’s Edition card is everything we saw in the previous links with a three-slot design and a single 16-pin power connector. The RTX 4090 will launch on October 12 and will cost $1599.

Here are some performance figures from NVIDIA…

RTX 4080 Cards

Just like we saw with the previous generation, there will be two different RTX 4080 models with different memory and GPU specs. So we have an RTX 4080 16GB and RTX 4080 12GB. The RTX 4080 16GB is based on the AD103 GPU and has 9728 CUDA cores, which are boosted to 2505 MHz. The 16 GB GDDR6X is clocked at 22.5 Gbps, giving you a memory bandwidth of 720 GB/s, and the card has a TDP of 320W. This 16GB variant will launch in November and will cost $1199.

The RTX 4080 12GB is based on the AD104 GPU and has 7680 CUDA cores. This makes us wonder why this card wasn’t just called the RTX 4070, but I think we’ll find out when we hear about the RTX 4070. The 12 GB GDDR6X is clocked at 21 Gbps, giving you a memory bandwidth of 504 GB/s. The TDP of the card is set to 285W. This version of the RTX 4080 will also be available in November and will cost $899.

And NVIDIA’s performance figures…

So that’s it for now on the RTX 40 series, guess what? If you’d like to see NVIDIA’s full presentation, you can watch it below.

Leave a Comment