Nvidia announced GeForce RTX 3060 for $329 and the Resizable-Bar technology.
Nvidia announced its new mid-range graphics card, the GeForce RTX 3060, built to Samsung’s 8nm manufacturing process and powered by GA106-300 silicon, comprising 3584 CUDA Cores, 112 TMUs / Tensor Cores and 28 RT Cores at 1.32/1.78 GHz Base/Turbo frequency along with 12 GB of GDDR6 memory @ 16 GHz along with a 192-bit memory interface and 170W power consumption.
Nvidia announced GeForce RTX 3060 and Resizable-Bar technology
All this translates into an FP32 performance of 13 TFLOPs, a RayTracing performance of up to 35 TFLOPs, and a performance of up to 101 Tensor-FLOPs in A.I. related operations. This graphic card will be on sale at the end of February for $329. Regarding images, we only have that one, and that is that there will be no Founders Edition models, only Custom models.
Additionally, it was announced that the Nvidia GeForce RTX 30 Series will be upgraded to receive a performance improvement through the Resizable-BAR technology, which means that the CPU can access the GPU’s ultra-fast VRAM (GDDR6) offering greater performance. For now, only the desktop and mobile GeForce RTX 30 Series will have access to this technology regardless of whether the platform is Intel or AMD.
Resizable BAR is an optional PCI Express interface technology. As you move through the world in a game, the GPU’s memory (VRAM) constantly transfers textures, shaders, and geometry through many small data transfers from the CPU to the GPU.
With the increasing size of assets/assets in modern games, this results in many data transfers. By using the Resizable BAR, the assets can be requested when needed and sent in full, so that the CPU can efficiently access all the frame memory. And if multiple requests are made, transfers can occur simultaneously, instead of being queued, thus improving performance.