When will RTX 4000 series come out?

NVIDIA’s new graphics cards Ada Lovelace are coming in 2022, while Hopper will be delayed to 2024 so the RTX 4080 and 4090 could arrive almost 100% by September 2022. Various sources point to the same launch date, and even the RTX 4070 is predicted to come first.

The RTX 4000 series is coming, but when? The truth is, there are many rumors, but few leaks regarding Ada Lovelace’s prototypes or performances. It looks like they will consume more energy but their power will be enormous, with the RTX 4070 being said to arrive first in August or September.

RTX 4080 and 4090 arriving in September, Hopper arriving in 2024

All eyes are on Twitter user @greymon55, who revealed and predicted all of NVIDIA’s plans for the RTX 4000. The other day he confirmed that the RTX 4080 and 4090 will arrive in September-October after the previous versions, releasing the RTX 4060 to December (as in the RTX 3000).

Yes, it’s true that NVIDIA intends to release the NVIDIA RTX 4070 first in August-September, and that comment doesn’t seem quite right to me. It looks like NVIDIA Hopper’s ambitious project will arrive in 2024 with a 5nm node from TSMC, just like the RTX 4000.

We will see a significant increase in the CUDA cores of the RTX 4080 and 4090 compared to the RTX 3080 and 3090, most pointing to 71% more performance. Personally, it looks like a percentage provided via a technical sheet that is on paper, which is not at all realistic because we already know that technical sheets are just numbers.

In fact, they serve to give us an idea of ​​​​a performance, but in practice, we have seen similar technical sheets later on, which were disappointing. If we get into the technical details, we will see the GDDR6X at the top end and with the same memory bus widths (384 and 320 bits).

Rumors suggest that the RTX 4080 will have around 16,000 cores and the RTX 4090 will have around 18,000 cores. As for the possible MCM architecture, it looks like AMD’s products will come before NVIDIA as the red team plans to run it on their Radeon Instincts.

According to Greymon55, these GPUs will have up to 512MB of 3D Infinity Cache and 32GB of GDDR6, all powered by a 5nm process made by TSMC. We didn’t say but the NVIDIA RTX 4000 was implemented in this 5nm process to be manufactured by TSMC, so we’ll forget about Samsung.

As for efficiency, the 5nm process is expected to bring 30% more power efficiency compared to Ampere’s 8nm. However, they say that NVIDIA RTX 40 will consume 30-50% more than the previous generation.