How DRAM-Less SSDs with HMB Perform in Hardware Benchmarks

This article provides a comprehensive overview of how DRAM-less Solid State Drives (SSDs) utilizing Host Memory Buffer (HMB) technology perform in computer hardware benchmarks. It briefly explains the mechanics of HMB, analyzes how these budget-friendly drives compare to traditional DRAM-equipped SSDs in both synthetic and real-world testing environments, and highlights the specific heavy-workload scenarios where this architecture reaches its physical limits.

Understanding HMB Architecture

Traditional SSDs use a dedicated onboard DRAM chip to store the Flash Translation Layer (FTL) mapping table. This table acts as an index, telling the SSD controller exactly where data is physically located on the NAND flash.

DRAM-less SSDs eliminate this chip to reduce manufacturing costs and power consumption. Historically, this caused severe performance degradation. However, modern NVMe SSDs use Host Memory Buffer (HMB) technology to bypass this issue. HMB allows the SSD to borrow a minuscule portion—typically around 64MB—of the computer’s main system RAM via the PCIe interface to store and access the mapping table.

Synthetic Benchmark Performance

Synthetic benchmarks, such as CrystalDiskMark and ATTO Disk Benchmark, test the theoretical maximum speeds of a drive using short, intense bursts of data. In these tests, DRAM-less SSDs with HMB perform exceptionally well.

  • Sequential Read and Write: HMB drives easily match DRAM-equipped drives in sequential tests. Because sequential data is written in a continuous line, the mapping table is barely stressed. A PCIe 4.0 DRAM-less drive can effortlessly hit top-tier speeds of 5,000 MB/s to 7,000 MB/s, relying primarily on its SLC cache and fast NAND flash.
  • Random 4K Read and Write: Random 4K tests measure how fast a drive can locate and process small, scattered files—a task heavily reliant on the mapping table. While system RAM accessed over PCIe has slightly higher latency than dedicated onboard DRAM, HMB still allows these drives to achieve random 4K scores that are within 10% to 15% of premium SSDs.

Real-World and Gaming Benchmarks

In application-focused benchmarks like PCMark 10 and the 3DMark Storage Benchmark, the gap between DRAM-less HMB drives and flagship DRAM drives narrows even further.

  • Gaming and DirectStorage: Game loading relies mostly on reading data rather than writing it. In gaming benchmarks, HMB drives load levels and assets within fractions of a second of high-end SSDs. They also fully support Microsoft’s DirectStorage API, showing no architectural disadvantage in modern gaming tests.
  • System Boot and Productivity: Benchmarks that simulate booting an operating system, launching Adobe applications, or opening large Microsoft Office files show virtually no penalty for using HMB. The tiny latency increase of fetching the mapping table from the PC's RAM is imperceptible in daily computing tasks.

Heavy Workload Benchmarks: Where HMB Struggles

To see a DRAM-less HMB drive fail in benchmarks, the testing must involve sustained, heavy workloads that push the drive past its caching limits.

  • Sustained Write Tests: When benchmarking continuous file transfers (such as moving a 200GB video file), the drive’s fast pseudo-SLC cache eventually fills up. Once full, the SSD must write directly to the slower TLC or QLC NAND while simultaneously folding data and updating the mapping table. Without dedicated DRAM to juggle these tasks, write speeds on HMB drives plummet drastically, sometimes dropping below the speed of a mechanical hard drive.
  • High Queue Depth Multitasking: Workstation benchmarks that bombard the drive with simultaneous read and write requests (high queue depths) expose the bandwidth limits of the HMB allocation. Because the SSD only borrows around 64MB of system RAM, the entire mapping table cannot fit at once. The drive must constantly swap map data in and out of the system RAM, leading to latency spikes and lower IOPS (Input/Output Operations Per Second).
  • USB Enclosure Testing: A critical benchmark limitation occurs when an NVMe HMB drive is placed inside an external USB enclosure. The USB interface does not support the HMB feature. Effectively stripped of its access to system RAM, the drive operates as a purely DRAM-less device, resulting in crippled benchmark scores and sluggish real-world performance.

The Benchmark Verdict

In hardware benchmarking, DRAM-less SSDs with HMB prove that onboard DRAM is no longer mandatory for high performance in consumer workloads. For short-burst synthetic tests, gaming benchmarks, and general productivity suites, HMB architecture delivers 85% to 95% of the performance of flagship drives at a significantly lower cost. They only fall behind in sustained, professional-grade workstation benchmarks where massive, continuous data writes are the norm.