PyNVMe3 Test Report of TiPlus7100
Read latency is too high in some workload (>300ms).
| 10 IOPS 99% (ms) | 4K 1QD 99.9% (ms) | RW mixed with trim 99.9% (ms) | |
|---|---|---|---|
| TiPlus7100 | 895.650 | 0.118 | 372.113 |
| SN770 | 102.830 | 0.179 | 16.043 |
| S50Pro | 1.824 | 0.122 | 78.861 |
Sequential read performance is better than competitors. But the time-0 speed is slow.
Trim performance is not good.
| IOPS (K) | Average Latency (ms) | performance before trim (MB/s) | performance after trim (MB/s) | |
|---|---|---|---|---|
| TiPlus7100 | 0.982 | 16.282 | 1181.302 | 1861.139 |
| SN770 | 4.494 | 3.554 | 899.344 | 1818.023 |
| S50Pro | 2.039 | 7.842 | 193.388 | 162.009 |
SLC cache reclaim is not aggressive.
Random read performance is low.
| 10% filled (K IOPS) | 50% filled (K IOPS) | 90% filled (K IOPS) | |
|---|---|---|---|
| TiPlus7100 | 80.877 | 67.528 | 66.457 |
| SN770 | 110.502 | 103.483 | 100.875 |
| S50Pro | 149.866 | 140.465 | 136.020 |
There is no significant decrease in read and write performance within PE1200.
| sequential read MB/s | sequential write MB/s | random read K IOPS | random write K IOPS | |
|---|---|---|---|---|
| TiPlus7100 | 5812.128 | 2671.073 | 80.877 | 20.379 |
| PE300 | 5825.692 | 2671.769 | 81.010 | 20.384 |
| PE600 | 5836.068 | 2661.653 | 81.234 | 20.307 |
| PE900 | 5817.895 | 2665.166 | 80.856 | 20.334 |
| PE1200 | 5823.219 | 2686.926 | 71.782 | 20.500 |
After two months of being idle without power, the read performance noticeably declines, with an evident increase in read latency for some operations. However, there have been no issues of data errors.
