According to the tweet of @momomo_us, four Russian listings have confirmed the specifications of Nvidia’s upcoming GeForce RTX 4060 Ti graphics card, including a whopping 8GB of memory capacity. This list shows Palit’s four video card models, detailing memory configurations and display connection outputs.
The stated specs are in full line with a previously reported leak outlining the full specs of the RTX 4060 Ti. Basically, from these reports, we’re seeing a pretty overwhelming memory configuration surrounding a 128-bit bus and 8 GB memory configuration.
If previous leaks were correct that this memory configuration would run on 18 Gbps GDDR6 modules, the RTX 4060 Ti would have about half the memory bandwidth of the RTX 3060 Ti and RTX 2060 Super, at just 288 GBps. Add in the fact that the RTX 4060 Ti is also reported to have only 8GB of memory, and this GPU is starting to look very underwhelming compared to its predecessors.
The good news is that the RTX 4060 Ti’s Ada Lovelace GPU architecture features significantly more L2 cache than its predecessor (32MB). This should counter many potential shortcomings of the memory configuration. A reported 36% increase in computational performance for TFLOPs makes it significantly faster than its predecessor in non-gaming situations. may become.
But coming back to the gaming side of the card, its slow memory bandwidth combined with just 8GB of memory will make it interesting to see how much of a bottleneck 288 Gbps is.
The problem is that modern video games constantly stream assets and textures on-the-fly into GPU memory as the player roams the game world. For example, the Unreal Engine 5 title utilizes Nanite. This is because the open game world is too large to store all the required textures and assets in video memory. As a result, lower memory bandwidth can lead to lower frame rates when the GPU has to wait for new textures to stream before rendering a completed frame.
Current rumors suggest that the RTX 4060 Ti will be announced at Computex 2023 later this month for $400-$450. So at least you get a cheaper Ada Lovelace GPU, if not significantly faster than its predecessor in memory bound scenarios.
