NVIDIA is expected to present three graphics cards of the new GeForce RTX 50 (Blackwell) series at the upcoming CES 2025. These are the RTX 5090, RTX 5080, and RTX 5070 models. CEO Jensen Huang will present the next-generation technologies for gamers and consumers.
There are already some details about the upcoming graphics cards on the web. Although the presentation of all three new products will take place simultaneously, each device will be launched separately.
NVIDIA GeForce RTX 5090
The flagship NVIDIA GeForce RTX 5090 is based on the GB202-300-A1 GPU and PG144/145-SKU30 PCB. The GPU contains 170 active streaming multiprocessors (out of a total of 192 SMs) and 21760 cores (out of a total of 24576 cores). This means that 11.4% of the function blocks are not working (compared to 11.1% for the RTX 4090 with AD102 GPU).
NVIDIA GeForce RTX 5090 has 32 GB of GDDR7 memory with a 512-bit access bus. The memory operates at 28 Gbps and provides 1792 GB/s of bandwidth. The total power consumption of the card is an impressive 600 watts. Previous rumors claimed that the Founders Edition would have a dual-slot cooler.
NVIDIA GeForce RTX 5080
The NVIDIA GeForce RTX 5080 is based on the PG144/147-SKU45 board and uses the GB203-400-A1 GPU. In this case, all 84 streaming multiprocessors and 10752 cores available in the GPU are used, but this is 51% less than the flagship RTX 5090. For comparison, the RTX 4080 had 40% fewer function units than the RTX 4090.
The GeForce RTX 5080 graphics card was allocated only half the memory of the older model – 16 GB. The access bus was also halved to 256 bits. However, it uses faster GDDR7 memory (32 Gbps), which provides a total bandwidth of 1024 GB/s. Power consumption will be 400 watts, which is 25% more than its predecessor.
NVIDIA GeForce RTX 5070
The NVIDIA GeForce RTX 5070 graphics card will offer a PG147 board and a GB205 GPU. The number of functional units in this model is currently unknown. Previous rumors claimed that the device would have 12 GB of memory with a 192-bit access bus (like the RTX 4070). However, thanks to the faster memory (28 Gbps), the total bandwidth will be increased by 33% compared to its predecessor – up to 672 GB/s. The GeForce RTX 5070 graphics card will require 250 watts of power (compared to 220 watts for the RTX 4070 SUPER, which is 14% more).
In addition to «brute force» in the form of more functional blocks, the new GeForce RTX 50 series graphics cards will have a number of other optimizations. The entire line will feature 14-layer PCBs, which will provide improved signal integrity between the GPU and GDDR7 video memory. All cards should have a single 12V-2×6 power connector interface. Only some top-end RTX 5090 or RTX 5080 models will have a dual 16-pin connector. In addition, the cards must be fully PCIe 5.0 compliant and support DP2.1a with full 80Gbps bandwidth support. Larger L3 cache and new memory compression techniques are designed to increase the overall bandwidth available to the GPU.
Source: wccftech
Spelling error report
The following text will be sent to our editors: