RTX 50 may be the first graphic card with GDDR7 memory

RTX 50 may be the first graphic card with GDDR7 memory.

GDDR7 video memory will be available next year! RTX 50 may be the first one.

News on June 29th, while announcing its quarterly results, Micron confirmed for the first time that it will soon launch the next-generation GDDR7 memory.

Micron revealed that their GDDR7 memory products will be officially released in the first half of 2024, using a 1βnm manufacturing process.

However, more details are lacking, such as key specifications such as capacity and frequency.

The current mainstream video memory specification is GDDR6, with a maximum frequency of 20GHz, from the AMD RX 7900 series.

NVIDIA’s mid-to-high-end cards have popularized GDDR6X, and the frequency on the RTX 4080 has reached 22.4GHz.

GDDR7 will inevitably further expand capacity and increase frequency.

Samsung has disclosed that it plans to achieve an astonishing 36GHz, which is a full 60% faster than now , but Samsung has not given a timetable.

If such a frequency is matched with a 384-bit bit width, it can bring an ultra-high bandwidth of 1.7TB/s, which is 70% higher than the current RTX 4090.

The NVIDIA RTX 50 series has basically been locked in 2025. By then, the GDDR7 video memory should be basically mature, and there will be no big problem with the first launch.

RTX50: Have to wait two years for the 3nm process?

News on June 28, whether it is a desktop or a notebook, the Ada Lovelace RTX 40 series has basically been laid out, so it is time to talk about the next generation, but the RTX 50 series will not come soon, and it will take two years.

According to a roadmap exclusively learned by HarwareLuxx, NVIDIA will launch the next-generation gaming GPU in 2025, but there is no clear code name yet, let’s call it “Ada Lovelace Next” for the time being.

This means that there will be no new graphics card next year, maybe there will be another Super series upgrade?

The RTX 50 series is set! Have to wait two years for the 3nm process?

In recent years, the NVIDIA GPU architecture has maintained a two-year upgrade rhythm, which will be extended to three years for the first time.

Fermi GeForce 400/200: 2010
Kepler GeForce 600/700: 2012
Maxwell GeForce 900: 2014
Pascal GeForce 10: 2016
Turing GeForce 20: 2018
Ampere GeForce 30: 2020
Ada Lovelace GeForce 40: 2022
Ada Lovelace Next GeForce 50?: 2025

According to previous reports, the codename of NVIDIA’s next-generation gaming GPU is “Blackwell” – David Blackwell, the first black academician of the American Academy of Sciences.

It is expected to be manufactured using TSMC’s 3nm process. For the first time, MCM multi-chip packaging will be introduced. The SM unit structure will undergo a new revolution.

The RT optical tracking unit will introduce a noise reduction accelerator to greatly improve performance. The video memory may be expected to be upgraded to GDDR7.

For AMD, the next-generation RDNA4 architecture is said to continue the Chiplet small chip design, and the scale has been greatly increased.

It is expected to reach 144 sets of CU units, 18432 stream processors, 32/48GB video memory, and 3.5GHz frequency, but it is too early.

In addition, NVIDIA’s top super chip will be the existing Grace Hopper until 2026. After all, it has been announced for a long time, but it has just been put into commercial use.

The CPU processor will also be upgraded in 2025, temporarily called “Grace Next”.