The NVIDIA GeForce RTX 4090 graphics card, released in October 2022, will soon end its life cycle, according to Chinese sources. Production of this flagship model, along with the China-exclusive GeForce RTX 4090 D version, will end in October 2024. This decision is related to the preparation for the launch of the NVIDIA RTX 5090, which, according to rumors, will appear in early 2025.

 

Despite the high cost, which in Ukraine starts from ₴63 thousand, the video card was popular among gamers and specialists in the field of artificial intelligence. However, the RTX 4090 had its share of issues – most notably excessive power consumption (TDP 450W) and incidents of melting power connectors that in some cases crashed the card. However, the RTX 4090 became the benchmark for gaming performance, and demand for it remained high.

 

With the end of production of the NVIDIA RTX 4090 and RTX 4090 D, there may be a temporary shortage, which is likely to lead to a temporary increase in prices, especially in the segment of artificial intelligence, where this model is also actively used.

 

The new NVIDIA RTX 5090 promises to be 48% faster in games and, according to preliminary data, will be equipped with 28 GB of GDDR7 with a bandwidth of 28 Gbps. The question of energy consumption still remains open — various sources provide conflicting information: some say it is increasing, others are saying it is decreasing.

 

The GeForce RTX 4090 model uses the AD102 crystal in a configuration with 16384 CUDA cores, which provide almost 83 Tflops of computing power, and a 384-bit memory bus. The video buffer is filled with GDDR6X chips with a total volume of 24 GB, the bandwidth is 1008 GB/s. The standard TDP level is 450 W. This is 100 W higher than GeForce. RTX 3090 on the Ampere architecture.

 

Formerly USA banned the export of consumer and professional video cards to China. So far, the decision applies only to the NVIDIA GeForce RTX 4090. Also, in the fall of 2022, the United States banned the export of chips to China with a certain data transfer rate. The limitations affected the A100 and H100 computing accelerators. To solve the problem, NVIDIA created their simplified versions – A800 and H800, respectively.