Cheap NVMe SSDs look like a deal on paper. Higher sequential read and write numbers than the drive you already own, lower price, what is not to like? Then you move a large file and the drive stalls halfway through.
According to makeuseof.com, the author bought a Lexar NQ780 1TB NVMe drive for a home server and hit unexpected slowdowns during a 145 GB Proxmox VM transfer. The reference is truncated, but there is enough to extract useful lessons.
- Drive tested: Lexar NQ780 1TB NVMe (PCIe 4.0)
- Workload: ~145 GB Proxmox VM transfer between two NVMe drives
- Initial speed: Fast for the first few seconds
- Bottleneck: SSD dropped to very un-SSD-like speeds mid-transfer
- Reference hardware: HP Z6 G4 workstation used as a server, with PCIe 3.0 M.2 slots
What happened during the transfer
The transfer started fast for a few seconds, then the Lexar drive dropped to unexpectedly low speeds. The job took far longer than the on-paper numbers suggested it should. The author was working over an unstable SSH session, which made the wait feel worse and the risk of an interrupted job more annoying.
The HP Z6 G4 workstation has PCIe 3.0 M.2 slots, so the drive was bottlenecked by the slot, not by the drive's PCIe 4.0 ceiling. Read speeds would have topped out around 3.5 GB/s in practice. The author expected writes to fit comfortably within that limit, since PCIe 3.0 x4 lanes are still generous for sequential work.
What the box does not tell you
Capacity and sequential read/write numbers are not the whole story. For light workloads, almost any modern SSD will do. Once you push the drive with mixed workloads, the components under the heatsink start to matter.
DRAM is one of those components. Higher-end SSDs include a small DRAM chip on the drive itself to keep a live map of where every piece of data physically lives. When the system asks for a file, the controller checks that map, finds the right NAND cells, and fetches the data. Budget drives often skip the DRAM chip and instead borrow a small slice of the host's RAM, using a feature called Host Memory Buffer, or HMB.
HMB drives are close to DRAM drives for many tasks. The gap shows up with random reads and writes spread across the drive, and with mixed workloads like several VMs running at once. For huge sequential transfers, DRAM matters less, but it is not the only thing that does.
NAND type also plays a role. Manufacturers squeeze more bits into each cell to raise capacity and lower cost, moving from SLC (1 bit per cell) through MLC (2 bits) and TLC (3 bits) to QLC (4 bits). More bits per cell means more voltage states the controller has to distinguish, which slows writes and shortens endurance. The NQ780 sits in this cost-cut segment of the market.
Our analysis: Why budget drives can disappoint
Sequential numbers look great on the box because the controller's cache is full, the NAND is fresh, and the workload is simple. Once the cache empties, the drive has to write directly to the NAND at its real sustained speed. That is where the gap between a budget drive and a flagship drive opens up.
The honest answer is that the cheapest NVMe drives cut costs on components that do not show up in marketing materials. DRAM, controller quality, and NAND type are common targets. Endurance ratings (TBW) are often lower too, which matters for server workloads.
Worth noting is that not every workload exposes the gap. If you mostly boot Windows, launch apps, and copy files occasionally, you may never notice. If you run VMs, databases, or large media projects, the gap shows up fast. Multi-tenant or self-hosted servers tend to expose it first.
Who should care and who can relax
| Use case | Will a cheap NVMe SSD feel slow? |
|---|---|
| Web browsing, office work, light file copies | Rarely |
| Single VM, occasional large transfers | Sometimes |
| Multiple VMs, databases, sustained mixed I/O | Often |
If your work involves sustained writes or random I/O across a full drive, the cost savings shrink fast when you factor in time. If your work fits in a few hours a week, a budget drive is a sensible call.
What to watch next
Watch for SSD reviews that include sustained write speeds, not just peak numbers. Watch for NAND type, DRAM presence, and TBW endurance ratings. If the spec sheet is missing those details, assume the drive has been cut to a price.
Bottom line: cheap NVMe SSDs are not bad, but they are not interchangeable with flagship models. Match the drive to the workload, not to the headline sequential speeds, and the budget choice stops feeling like a compromise.
Key takeaways
- Sequential read/write numbers do not capture sustained performance on cheap NVMe drives
- DRAM-less drives using Host Memory Buffer can lag under random and mixed I/O
- PCIe generation can bottleneck even fast PCIe 4.0 drives on older motherboards
- Endurance ratings and controller design matter more than headline sequential speeds for servers
Frequently asked questions
Why did the budget NVMe SSD slow down during a large transfer?
Cheap NVMe SSDs often rely on a small write cache. Once that cache fills, the drive must write directly to the NAND at its real sustained speed, which is much slower than the peak number on the box.
Does DRAM matter on an NVMe drive?
DRAM helps the controller keep a fast map of where data lives. Drives without DRAM use Host Memory Buffer, which is fine for light work but can lag under random and mixed I/O.
Are QLC drives slower than TLC drives?
Generally yes for sustained writes, because QLC stores more bits per cell and takes longer to program. The gap is smaller for reads and for short bursts.
Will a budget NVMe drive slow down my gaming PC?
Usually not. Games load in short bursts that fit inside the cache. Slowdowns are more common in servers, VMs, and large file work.
How do I avoid this when shopping?
Look for sustained write benchmarks, TBW endurance ratings, DRAM presence, and NAND type. If the spec sheet hides those details, the drive was probably cut to a price.