Lexar NM1090 PRO 1TB | Test Report
Test Platform
| Item | Test configuration |
|---|---|
| Drive | Lexar NM1090 PRO 1TB; M.2 2280, PCIe Gen5 ×4 |
| Sample | One sample; serial number QDV975W100012P3200 |
| Firmware | Y0527A00 |
| Platform | NUC, Core Ultra 9 285H, 16 GB RAM; Ubuntu 26.04.1, Linux 7.0.0-34 |
| Tools | PyNVMe3 26.11; Quarch PPM for power measurement |
| PCIe Link | 32 GT/s ×4 |
| HMB | Not enabled |
Test Procedure
PyNVMe3 was used to test protocol, performance, power, stress, power cycling, and model reading.
| Category | Test ID | Scope |
|---|---|---|
| Protocol and reset | TP1 | Management, I/O, reset, power management, register and security functions; some tests were skipped or had anomalies. |
| Low-power states | TP2.a | Power consumption and read/write response in low-power states. |
| Performance and power | TP2.b | Read/write performance, response time and power at different write-workflow checkpoints. |
| Stress and data verification | TP2.c | Completed approximately six hours of stress testing and full-drive readback verification. |
| Sudden power loss and recovery | TP2.d | One test timed out in cycle 3 during recovery and did not complete; another independent run completed 100 cycles. |
| Normal power cycles | TP2.e | Completed 100 normal power cycles. |
| Advertised performance | TP2.f | Peak sequential and random read/write performance in a small region. |
| Link speed switching | TP2.g | Fixed and dynamic link-speed switching under sustained reads. |
| Model loading | TP2.h | Read synthetic images of different sizes at various queue counts and queue depths. |
Product Radar

This chart shows five aspects of the tested sample. Protocol covers functions related to NVMe and PCIe specifications; performance covers read and write capability; latency concerns response time; power covers electricity consumption; and other includes stress, power cycling and recovery. Each dimension is scored from 1 to 5, with a higher score farther from the center.
Performance Positioning

Each black dot places one drive according to two tests. Farther right means higher sequential write speed; higher up means stronger random read performance. The width and height of the rectangle represent longer response times in two workloads. Color represents power consumption during testing: blue is lower and red is higher.
Model Loading Time
This test simulates reading model images from an SSD and compares loading times at different queue counts and queue depths. The two drives were tested with different logical block formats and prefilled region sizes.

The four panels show small (4.7 GiB) and large (40 GiB) model images, each with a single queue and multiple queues. Queue depth on the horizontal axis is the number of read requests that each queue can issue at the same time; the vertical axis shows average read time. A lower line means the read completed sooner under those test conditions. Each point summarizes multiple measurements; the vertical bars show the observed range.
Key Issues
| ID | Test ID | Description |
|---|---|---|
| NM1090-01 | TP2.d | In one sudden power-loss run, model identification and first I/O succeeded after the third power-on, but temporary I/O queue deletion timed out; admin commands in the same test process subsequently continued to time out. An independent run completed 100 cycles. |
| NM1090-02 | TP1.d | After Boot Partition writes, readback was all zero or differed from written content; 20 of 24 cases failed. |
| NM1090-03 | TP1.i | After the second incorrect authentication in TCG testing, transaction rollback reduced the authentication attempt counter from 2 to 1. |
| NM1090-04 | TP2.b | During half-drive TRIM, P99.9 response time was 218.116 ms and the maximum was 602.481 ms. |




