There's every reason to believe that the future GeForce RTX 5090 will surpass anything we've ever imagined when it comes to graphics cards.
NVIDIA is all about artificial intelligence and the gas pedals it designs for specialized companies. The new chips presented at the latest Computex show overshadow all communication about the next generation of graphics cards, which should nevertheless arrive before the end of 2024, according to insistent rumors. The latest rumors surrounding the spearhead of this new range - the GeForce RTX 5090 - even point to completely insane power, far, far from what the GeForce RTX 4090 can offer.
© Wccftech
Quoting Panzerlied, an often well-informed user of the Chiphell forums, the Wccftech site reports on this GeForce RTX 5090, which could be released in December, a few weeks after the launch of the GeForce RTX 5080. We already know that the GPU planned for this RTX 5090 - the GB202 - will feature many more Cuda cores than that of the RTX 4090: rumors speak of over 22,000 cores, compared with 16,384 on the former card. We also know that video memory will be in GDDR7 (vs. GDDR6X), offering a total of 28 GB (vs. 24 GB) with a speed of 28 Gbps (vs. 21 Gbps) and a 448-bit interface bus (vs. 384-bit). In other words, on a strictly technical level, there's no contest between the two cards.
But the differences aren't limited to this data alone, as NVIDIA is set to take advantage of the superior engraving finesse used on the GB202 to enable it to reach frequencies never seen before on a GPU. We're talking about a base frequency of 2.9 GHz, whereas the AD102 in the GeForce RTX 4090 was limited to 2.23 GHz. These are only base frequencies, bearing in mind that the boost provided by NVIDIA always goes well beyond this (+200/300 MHz), and that graphics card manufacturers even push the cursors a little further: the RTX 5090 should therefore exceed 3 GHz under load. We can't wait to see this racing beast in action... and also to see how many watts it will gobble up!
