Upgrade your system because this model has a bottleneck (AMD x Nvidia).
Upgrade your system because this model has a bottleneck (AMD x Nvidia).
AM5 is only compatible with AMD 7000-series CPUs and those require DDR5 RAM. So you'd have to upgrade CPU and RAM at the least. You could get PCIe 4.0 for GPU and one NVMe by going with B550, but it'll not make much of a difference. I've got a PCIe 4.0 NVMe and the difference to my PCIe 3.0 NVMe, as far as games are concerned, is virtually zero. Very few games show any difference in loading times and where it does it's negligible. For the GPU it's only really needed for low end GPU like the RX 6500 XT, because it only has 4 PCIe lanes. Here's a benchmark for the RX 6600 XT (8 lanes) and the difference is already not that big: https://www.techpowerup.com/review/amd-r...s-scaling/
I was worried because I'm using an RTX 4060 Ti on a 3.0 system, and the 4060 Ti likely runs at PCIe X4 speed.
What is this calculation supposed to do? It needs to be executed at 3.0 times 8.
It operates at PCIe 3.0 x8 speed, matching PCIe 4.0 x4 in bandwidth. Generally this isn't a problem since the linked benchmark demonstrates it. Problems arise when the card supports only 4 lanes, then it would run at PCIe 3.0 x4, equivalent to PCIe 4.0 x2 performance.
I noticed the connection, it appeared on August 24th, 2021. Probably not current after that date, yes?
It hasn't changed much at all since then, particularly with low-end GPUs such as the 4060.
The benchmark outcomes remain consistent. GPUs typically aren't restricted by PCIe bus speed in most cases. This could shift with upcoming hardware, but currently it hasn't. Certain models like the RX 6400 and RX 6500 XT are exceptions due to their limited lane count. Intel may also be an outlier, often needing reBar for peak performance. Performance tends to improve as you move higher in the stack because cards store more data locally and don't rely heavily on bus access. An example of this was tested with the 4090 in late 2022: https://www.techpowerup.com/review/nvidi...s-scaling/