NVIDIA’s GeForce RTX 4080 graphics cards aren’t expected to make their debut for another month. Yet, for some unknown reason, benchmarks allegedly belonging to the card have made their way online.
It should be made clear that even by our standards, this leak about the RTX 4080 16GB is highly suspect for a couple of reasons. Firstly, it comes at a time when the performance details of NVIDIA’s flagship, the RTX 4090, won’t officially be appearing until next week. Second, and in conjunction with the first point, we are wondering how the parties in question managed to get their hands on a working driver for the card, especially since it is NVIDIA alone that provides the official and unreleased drivers of a new product to reviewers, in order to get their reviews out in time. With that, we’ll also be telling you here and now to take the next part with a very, very tight pinch of salt.
As per the results posted on the Chiphell forums, the alleged RTX 4080 16GB was detailed in an AIDA64 posting, where the details point to the GPU’s alleged specs sheet. Specifically, it has a boost clock of 2505MHz and 76 Compute Units (CUs) or in the case of NVIDIA’s labelling conventions, Stream Multiprocessors (SMs). In terms of performance, the GPU’s single-precision FLOPs are listed as 53620 GFLOPs, which is supposedly just a little higher than what NVIDIA says the card will do at launch. Further, the card seems to have been tested on an AMD X570 system, seeing how the leakster’s CPU is a Ryzen 7 5800X3D.
Moving on, the leaker also published the alleged 3DMark performance of the card, specifically the Fire Strike Ultra, Time Spy Extreme, and Port Royal test. The card scored 17465 points, 13977 points, and 17607 points respectively, all of which are scores that surpass NVIDIA’s previous generation flagship, the RTX 3090 Ti. In addition, there is also a screenshot of the card’s performance running Shadow of the Tomb Raider, where it maintained an average frame count of 128 fps at 4K resolution.
The NVIDIA GeForce RTX 4080 is set to launch in November, but the GPU brand has failed to set an exact date for it. The card will be available in two SKUs: one with 12GB GDDR6X and the other with 16GB GDDR6X graphics memory. The 12GB model will have 7680 CUDA cores plus a boost clock of 2.61GHz and 192-bit memory bus, while the 16GB variant will have 9728 CUDA cores, a boost clock of 2.51GHz, and a 256-bit memory bus. Price-wise, the 12GB model will start from RM4730 and the 16GB will start from RM6300.
(Source: Chiphell via Videocardz)