JD JZao9 1TB
Basic Information
In this report, we review JZao9 from JD.com with Lenovo Y9000 and ZhiTai TiPro7000.
| Model Name | Firmware Version | |
|---|---|---|
| JZao9 | J.ZAO 9 SERIES 1TB SSD | 3.7.2.16 |
| Y9000 | Lenovo Y9000 SSD-M.2-2280s-NVMe-1T | EIFM31.4 |
| TiPro7000 | ZHITAI TiPro7000 THREE-BODY 1TB | ZTA32F46 |
In this report, we use the aged (P/E is about 1000) performance data for comparison.
Latency
We still check the latency first of all.
Max Write Latency
| 10 IOPS (ms) | 4K 1QD (ms) | 512B 1QD 10p filled (ms) | 512B 1QD 50% filled (ms) | 4K mix RW 90% filled (ms) | RW mixed with trim (ms) | Overwrite 50% (ms) | |
|---|---|---|---|---|---|---|---|
| JZao9 | 9.353 | 10.444 | 192.764 | 2815.426 | 399.199 | 389.069 | 359.472 |
| Y9000 | 7.853 | 9.788 | 258.103 | 5.970 | 15.949 | 280.053 | 1303.401 |
| TiPro7000 | 161.036 | 260.417 | 278.812 | 274.341 | 286.072 | 253.195 | 801.613 |
High max latency is the common issue of all these 3 drives. JZao9 even has max latency as high as 3 seconds. It needs a lot of effort to optimize max latency, but most of the simple test tools do not care this value. However, the unpredicted long latency can give user a noticable delay.
Speed
SLC cache
We keep write the drives from the empty state. JZao9 also use the whole capacity as SLC cache, so when the SLC is used up, the write speed will drop significantly.
SLC cache retirement
When the drive is young and utilized space is low, SLC cache gives user the highest performance. But when the drive is aged (e.g. > 2000 P/E cycle), we can find that almost the whole SLC cache is retired, and we cannot observe SLC performance any more in the test diagram below.
Y9000 use a small SLC cache, but it can last longer. In the test below, we can still have the SLC performance when the drive is as old as the JZao9 above.
Full capacity SLC cache can give a beautiful score in FOB state. But when the drive is old and fully used, huge SLC cache would cause huge drop on performance.
5-stage write performance
We test random write performance after filling 10%, 50%, 90% respectively. Furthermore, we will fill the first 50% space for the second time, and repeat test again. And we will also trim the first 50% space, and repeat the same test. Here is the result.
We can find that more space filled, SLC performance can keep for shorter duration. But Trim can recover the performance.
Power
If we use SSD in laptop, the power consumption is also a key consideration.
basic data
We list TMT1/2 setting below.
| TMT1 (℃) | TMT2 (℃) | |
|---|---|---|
| JZao9 | 74 | 84 |
| TiPro7000 | 99 | 109 |
| Y9000 aged2 | 67 | 69 |
Low Power State
| PS4 measured power (mW) | PS4 exit duration (us) | PS3 measured power (mW) | PS3 exit duration (us) | |
|---|---|---|---|---|
| JZao9 | 1764.2 | 58.3 | 1767.5 | 56.3 |
| Y9000 | 2.5 | 155121.4 | 46.5 | 155443.0 |
| TiPro7000 | 9.0 | 46681.9 | 89.7 | 8192.1 |
It seems that JZao9 cannot enter low power state.
Function
JZao9 use the full capacity SLC cache, and is not optimized on Latency and Power, so JZao9 is far from well designed. We also observe functional issues during our tests:
- linux inbox driver cannot detect the drive
- performance test: hang and drive lost
- conformance tests: many failed with error happened. TCG is not supported.
- LED does not work.
