For a few months we have known that the new NVIDIA graphics cards will arrive in a few months, since they will be presented during CES while they would have a launch during the first months of the year, that is why we know more details about them. In this case, we now know that the company which created the RTX once again resorted to Samsung for the memory who will join the series 50leaving aside those of Micron that we have seen over the last two generations to resort again to those offered by the South Korean company.
There are three key companies currently developing video memories, as we well know, they are SK Hynix, Samsung and Micron, which offer solutions of all kinds, making it difficult for some brands to choose between the one and the other since each has its strong points. After using Micron memories for two entire generations, Nvidia wants to strike again Samsung a big role, since the desktop versions of its new graphics cards will have the memory GDDR7 of this company, while in laptop companies they will use all three, but they will also prioritize them over the other two companies.
One of Samsung's biggest wins this year, getting NVIDIA to use its memories
As we know well, Samsung has not gone through its best moment, the semiconductor manufacturer has quite a few problems in offering advances in the race for sub-3nm nodes, while in other aspects like the AI, he can't keep up. too much. That's why the fact that NVIDIA chose to use the video memories they developed is a victory, especially since they hadn't collaborated with them on this in some time because there was another company that had a lot more. priority.
As reported by Korean media Greened.kr, NVIDIA has chosen to use Samsung's GDDR7 memories due to 28 Gbps and 32 Gbps in its next generation, offering quite a significant improvement over previous series. One of the most important data that we need to know about how RTX graphics will integrate this VRAM memory is the fact that the RTX 5080 will be the one that will integrate the 32 Gbps one, while the rest will use the 28 Gbps one, some thing this in A beginning may seem strange, but it has certain reasons.
In this case, NVIDIA has chosen to use 28 Gbit/s memory in the RTX5090 because the interface it uses is the widest 512-bit, which, by using many more memory chips, could directly affect the power consumption of the entire board. Despite this, the increase in bandwidth that this GPU will have is quite significant, and we are talking about being able to reach up to 1792 GB/s, 78% more compared to the 1008 GB/s of which it is capable of reaching . the RTX 4090, so the fact that it integrates “inferior” memory does not mean that it will have worse performance.
