Nvidia at its CES 2025 event today announced the anticipated RTX 50 series (Blackwell) GPUs. The company announced four SKUs, the RTX 5090, the 5080, the 5070 Ti, and the 5070. Specifications of these GPUs are pretty much in line with the previous leaks and that means the reports of the RTX 5080 and 5070 having the same amount of memory capacity as their predecessors have turned out to be true.
Thus the RTX 5080 and 5070 respectively have 16GB and 12GB VRAM buffers despite all the criticisms from the tech press as modern high quality textures as well as ray tracing effects eat up a lot of the frame buffer.
However, Nvidia says lack of VRAM should not be a problem this time around as the company is debuting a new range of technologies to solve such issues under the ambit of "RTX Neural Shaders".
RTX Neural Shaders are essentially programmable shaders super-charged with the power of neural networks to help in various aspects of efficient game asset rendering. Texture compression is one such area and according to Nvidia, its new RTX Neural Texture Compression (RNTC) technology has the capability to compress "thousands of textures" within a minute thus saving up to seven times more GPU memory or system memory than traditional ways.
In its blog post, Nvidia writes:
RTX Neural Texture Compression uses AI to compress thousands of textures in less than a minute. Their neural representations are stored or accessed in real time or loaded directly into memory without further modification. The neurally compressed textures save up to 7x more VRAM or system memory than traditional block compressed textures at the same visual quality.
Aside from RNTC, the new RTX Neural Shader also introduces RTX Neural Materials (RNM) for in-game high-quality material processing and RTX Neural Radiance Cache (RNRC) for faster path tracing.
Here are the specs of the Nvidia RTX 5090, 5080, 5070 Ti and 5070:
| Model | GPU | CUDA Cores | Memory | Memory Interface | Memory Speed (Gbps) | Memory Bandwidth (GB/s) | Total Graphics Power (TGP) |
|---|---|---|---|---|---|---|---|
| RTX 5090 | GB202-300 | 21760 | 32GB GDDR7 | 512-bit | 28 | 1792 | 575 W |
| RTX 5080 | GB203-400 | 10752 | 16GB GDDR7 | 256-bit | 30 | 960 | 360 W |
| RTX 5070 Ti | GB203-300 | 8960 | 16GB GDDR7 | 256-bit | 28 | 896 | 300 W |
| RTX 5070 | GB205 | 6144 | 12GB GDDR7 | 192-bit | 28 | 672 | 250 W |
In terms of performance, Nvidia says that the RTX 5090 can deliver twice the output of the RTX 4090 and the former has received a price bump since it has been pegged at $1999 compared to the 4090"s launch MSRP of $1599.
| | |
However, that isn"t the case with the RTX 5080 as it costs half that of the 5090 and thus is $200 cheaper than what the 4080 debuted at. These will be available starting January 30.
| | |
Meanwhile, the RTX 5070 Ti and 5070 are priced at $749 and $549 respectively and will be on shelves starting in February. This is what AMD"s new RX 9070 XT and 9070 will likely be competing against.