My ADATA RAM throughput is all over the place in Starfield benchmarks. Should I quantify the bottleneck and export a baseline curve?

I ran a scenario where memory frequencies were jumping wildly between 5800-6200MHz during heavy exploration. I thought adjusting the disk read queue depth would help, and while the raw numbers went up, the throughput 'sawtooth' caused by CPU scheduling stayed. I realized I had to go into the BIOS. I enabled the Fast I/O Channel resource allocation and used a rendering benchmark tool to pinpoint the system bottleneck. The path was: Queue Depth Adjustment → Enable Resource Allocation → Quantify Render Test → Export Baseline. I could still feel the heat radiating off the RAM heatsinks during high-frequency bursts, but the exported data showed throughput fluctuations dropped by about 15%. This proved that optimizing the low-level bus protocol is way more effective than just messing with software settings.