Nvidia GeForce RTX 5090, RTX 5080, RTX 5070 Ti and RTX 5070 desktop GPU unveiled with surprisingly decent prices

[Redaktion]

Nvidia has shown off its GeForce RTX 50 series desktop graphics cards at CES 2025. Four SKUs have been shown off, namely the GeForce RTX 5090, RTX 5080, RTX 5070 Ti and RTX 5070.

https://www.notebookcheck.net/Nvidia-GeForce-RTX-5090-RTX-5080-RTX-5070-Ti-and-RTX-5070-desktop-GPU-unveiled-with-surprisingly-decent-prices.942871.0.html

[rouzman]

The RTX 5070, positioned in NVIDIA's next-generation graphics card lineup, offers an excellent price-to-performance ratio, delivering high performance for gaming and creative applications at a competitive price point.

[RobertJasiek]

The RTX 5070, positioned in NVIDIA's next-generation graphics card lineup, offers an excellent price-to-performance ratio, delivering high performance for gaming and creative applications at a competitive price point.

Are you a Nvidia bot? Since you already know all the performance, tell us about Time Spy, Blender, OpenCL etc. benchmarks! Almost the only thing known so far is DLSS 4 software in only those few games hand-picked by Nvidia, which says precisely nothing about creative applications, let along serious AI training or inferencing.

Nvidia keeps numbers of Tensor cores secret. CUDA cores are a weaker indication; judging only by the increase of their number, 5070 is a tiny improvement, which even might decrease in efficiency given the 50W (or 25%) increment of the wattage.

The gap between 5080 and 5090 is more than double in CUDA cores so that 5080, 5070TI, 5070 etc. are gimped dramatically, turning 80 tier into 50 or 60 tier and 70 tier into something even much below that.

New core type, new VRAM type and bandwidth might be the saviour but this is very unlikely for almost all applications.

The most likely, RTX 5000 ought to be skipped by everybody. However, before final judgement, we should await proper tests to see whether anything remains after removing Jensen's gross lies.

[GeorgeS]

New core type, new VRAM type and bandwidth might be the saviour but this is very unlikely for almost all applications.

The most likely, RTX 5000 ought to be skipped by everybody. However, before final judgement, we should await proper tests to see whether anything remains after removing Jensen's gross lies.

Indeed.

I'll await seeing benchmarks and detailed reviews of actual consumer products before passing judgement.

Granted, YES it is expected that each sub-model should perform better than the prior generation model.

Exactly how much that might actually be in the real world has yet to be seen.