Nvidia today announced its upcoming RTX 50 series (Blackwell) GPUs at its CES 2025 event. The company announced four SKUs: RTX 5090, 5080, 5070 Ti and 5070. The specifications of these GPUs are largely consistent previous leaks meaning that reports of the RTX 5080 and 5070 having the same amount of memory as their predecessors were true.
Thus, the RTX 5080 and 5070 respectively have 16 GB and 12 GB VRAM buffers, despite all the criticism from the tech press, since modern high-quality textures, as well as ray tracing effects, eat up most of the frame buffer.
However, Nvidia says that running out of VRAM shouldn’t be an issue this time, as the company is introducing a new range of technologies to address such issues under “RTX Neural Shaders.”
RTX Neural Shaders are essentially programmable shaders infused with the power of neural networks to help with various aspects of efficiently rendering game assets. Texture compression is one such area, and Nvidia says its new RTX Neural Texture Compression (RNTC) technology is capable of compressing “thousands of textures” in a minute, thereby saving up to seven times more GPU or system memory than traditional ones. ways.
On their Nvidia blog writes:
RTX Neural Texture Compression uses artificial intelligence 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. Neurally compressed textures save up to 7 times more VRAM or system memory than traditional block-compressed textures at the same visual quality.
In addition to RNTC, the new RTX Neural Shader also introduces RTX Neural Materials (RNM) for high-quality in-game material processing and RTX Neural Radiance Cache (RNRC) for faster path tracing.
Here are the specs for the Nvidia RTX 5090, 5080, 5070 Ti and 5070:
| Model | GPU | CUDA colors | Memory | Memory interface | Memory speed (Gbps) | Memory Bandwidth (GB/s) | Total Graphics Power (TGP) |
|---|---|---|---|---|---|---|---|
| PTX 5090 | GB202-300 | 21760 | 32 GB GDDR7 | 512-bit | 28 | 1792 | 575 W |
| PTX 5080 | GB203-400 | 10752 | 16 GB GDDR7 memory | 256-bit | 30 | 960 | 360 W |
| RTX 5070 Ti | GB203-300 | 8960 | 16 GB GDDR7 memory | 256-bit | 28 | 896 | 300 W |
| PTX 5070 | GB205 | 6144 | 12 GB GDDR7 | 192-bit | 28 | 672 | 250 W |
In terms of performance, Nvidia speaks that the RTX 5090 can deliver twice the performance of the RTX 4090, and the former received a price increase as its price was pegged at $1,999 compared to the 4090’s launch MSRP of $1,599.
|
|
|
That’s not the case for the RTX 5080, however, as it’s half the price of the 5090 and therefore $200 cheaper than the 4080 debuted. They’ll be available starting January 30th.
|
|
|
Meanwhile, the RTX 5070 Ti and 5070 retail for $749 and $549 respectively and will go on sale starting in February. Here’s what’s new from AMD RX 9070 XT and 9070 will most likely compete.