Weird micro-stuttering on a server-grade storage setup - HBA/PCI-E issue?

[lacowi4161]

Hey guys, I’ve got a bit of an odd technical hurdle here that I was hoping some of the hardware gurus in the community might have run into. I’m a bit of a digital hoarder when it comes to Doom mods—I’ve got everything from the classic 1994 shovelware CDs to the latest massive TC projects like Ashes: Afterglow. Because my collection has gotten so huge (closing in on 4TB of WADs, PK3s, and dev assets), I recently moved my entire GZDoom installation and my library onto a dedicated storage server I picked up second-hand.
The server is a bit overkill for Doom, honestly. It’s an old dual-Xeon workstation running a dedicated HBA (Host Bus Adapter) linked to a PCI-E expander card https://serverorbit.com/hba-and-control ... er-card-en so I can manage a pretty hefty array of SAS and SATA drives. For the most part, it works perfectly for file transfers and hosting, but I’ve noticed a specific point of frustration when I actually try to play GZDoom directly from the drives attached to that controller. I’m getting these tiny, half-second micro-stutters whenever the engine has to load a new sound effect or a large texture from the PK3 files during gameplay.
I’ve checked the usual suspects—drivers are updated, and I’ve messed with the internal heapsize settings—but I’m starting to wonder if there’s some inherent latency in how an HBA or a PCI-E expander handles the small, rapid file requests that ZDoom-based source ports make. Usually, these HBA setups are optimized for massive sequential data transfers, not the "picky" random access you get when a mod like Brutal Doom is pulling 100 small sprite files in the middle of a chaotic firefight. It’s a bummer because I love the idea of having my "Ultimate Doom Machine" being this industrial-grade tank of a PC, but the stutters are killing the flow when I’m trying to survive a slaughter map.
I know most people just run their games off a standard NVMe or a basic SATA SSD these days, but has anyone here experimented with running GZDoom or similar engines on enterprise-grade storage hardware? I’m curious if there’s a way to force better caching at the OS level or if the PCI-E expander is simply introducing too many "hops" for the engine to handle gracefully.
Do you think the engine’s file-reading logic is just too "old school" to play nice with modern server architecture, or is there a specific controller setting I’m missing that could help smooth out that data pipeline?