Complete RTX 4080 SUPER, RTX 4070 Ti SUPER, and RTX 4070 SUPER specifications leak days ahead of official launch

[Redaktion]

Consisting of RTX 4080 SUPER, RTX 4070 Ti SUPER, and RTX 4070 SUPER, the RTX 40 SUPER series is expected to bring some nice, if not revolutionary, upgrades over the vanilla RTX 40 cards with potential price adjustments. Per specifications leaked by VideoCardz, all three GPUs will get increased core counts, base clocks, and 12VHPWR power connectors. In the case of the RTX 4080 SUPER and the RTX 4070 Ti SUPER, NVIDIA is also allegedly improving the memory profile of the boards with more bandwidth.

https://www.notebookcheck.net/Complete-RTX-4080-SUPER-RTX-4070-Ti-SUPER-and-RTX-4070-SUPER-specifications-leak-days-ahead-of-official-launch.788520.0.html

[Jab]

Nvidia cured me from PC gaming. Thanks to their GPU prices, I am now an XSX player mainly (Although I still have older laptop with 1070 on win 10 - because of Skylake processor). I am not going to build a new computer, unless the price to performance ratio gets way way better.

[Amen]

Same lmao. I've been completely sober from my severe pc gaming addiction ever since the AMD's complete and utter incompetence in the laptop market for 4 generations in a row (both in mobile Ryzen APU's & mobile Radeon dGPUs). AMD really should be awarded the Nobel Peace Prize for their services to humanity.

Plan to play older emulated games (since modern gaming sucks in general) on the OnePlus 12R when that gets released. Then get a Switch 2 later on to play the heavier PS5/XSX cross gen ports.

[frankie425]

i wonder why 4090 costs in my nearest store almost 2K. The important is the size of VRAM today, no one cared about games anymore, everything Ai. For stable diffusion even 24Gb VRAM is not enough so they've made driver allowing offloading data to system RAM, but this is not working for any other types of Ai like LLMs, where people forced to make several versions of same LLMs - one for GPU only (4090 is not enough!) and for CPU only (GGUF format with offload to GPU).

[frankie425]

It's interesting how Ai reminds digital music compression, on lesser hardware you run it - worse quality you get, like mp3, hallucinations and errors, better hardware - uncompressed, original and opens new abilities like training your own LLM on your data. Today 4090 is minimum (3090 Ti also great choice but it take 24 hours constant work min to train your smallish LLM) for training your own LLM, profi cards like A100 costs like car and huge amount of VRAM (40-80Gb) there for a reason! Games don't need such because in comparison is very useless task.