Biwin Black Opal X570 PRO Review – A new player on the market with an interesting product and minor weaknesses

1 - Introduction and Technical Data 2 - Teardown with Component and Material Analysis 3 - Streaming and Root Cause Analysis of Drops 4 - Synthetic Benchmarks 5 - Workstation Benchmarks 6 - Temperatures, Power Consumption and Conclusion

The two diagrams below from the SPECwpc workstation test series show impressively how clearly real workloads differ from synthetic benchmarks such as CrystalDiskMark or ATTO, especially with an SSD that is not empty but has already been used intensively and written to with a real system image. The focus here is on the Biwin X570 PRO 2TB with PCIe 5.0 connection based on the SM2508G controller, which is pitted against a range of other current NVMe SSDs under real-world conditions.

Write performance in real applications (PECwpc Write)

In the six workstation workloads, there are clear differences depending on the application. In the CCX and Maya 2017 scenarios, the Biwin X570 PRO is in the upper third of the test field in each case, but cannot set itself apart from other Gen5 SSDs with Phison E26 or E27. Values of 134.64 MB/s (CCX) and 156.78 MB/s (Maya 2017) are solid, but not significantly better than competing products such as the MP700 Pro (Corsair) or the Spatium M580 models from MSI. More interestingly, however, PCIe 4.0 models like the MSI M480 2TB are not far behind, suggesting that the application profiles in these scenarios are heavily controller or latency limited, not bandwidth limited.

In the Maya Venice test, the Biwin SSD does not fall behind with an average value of 54.22 MB/s, but is on a par with PCIe 4.0 competitors. In icePak, on the other hand, it clearly positions itself in the top group with 383.27 MB/s, which indicates efficient burst write behavior and functioning caching – a positive signal. In the prodDEV workload, which stands for highly fragmented and unpredictable accesses, the result of 3.66 MB/s is weak, but is within the range of the comparison devices – this test is generally very controller-intensive and restricts almost all participants to the same extent.

The results in NAMD (193.45 MB/s), MCAT (187.32 MB/s) and Handbrake (357.93 MB/s) are outstanding. In 3ds Max, the Biwin X570 PRO falls slightly behind the fastest PCIe 4.0 models with 205.45 MB/s, but still remains in the upper range. These results indicate that the Biwin SSD performs well in real-world workloads with high data rates and even distribution, but does not offer any unique selling points compared to well-optimized Gen4 drives.

Read performance in real applications (SPECwpc Read)

The read picture shows a similar trend, but with slightly better relative results overall for the Biwin SSD. In CCX, Maya 2017 and Maya Venice, the X570 PRO is consistently above 290 MB/s, which puts it on par with the best Gen5 drives. The performance is stable, the differences to models such as the MP700 Elite or MSI M580 are only a few percentage points.

In icePak (214.92 MB/s), it also performs well in the field, although not quite at the top. The throughput in prodDEV is strikingly strong at 190.56 MB/s – here the Biwin SSD is one of the few that can run this critical application at this level. In NAMD (776.71 MB/s), MCAT (596.21 MB/s) and Handbrake (975.44 MB/s) it is in the top group or even the leader. The results in 3ds Max (449.22 MB/s) are clearly above all Gen4 SSDs and confirm that the full potential of the SM2508G controller is utilized, especially when reading large amounts of data from sequential or semi-structured sources.

Comparison with synthetic benchmarks and interim conclusion

If you compare these practical values with the CrystalDiskMark and ATTO results of the same SSD, a central problem of synthetic tests becomes clear: The measurements in CrystalDiskMark consistently showed over 14 GB/s read throughput and over 13 GB/s write performance, for practically every test size, regardless of whether 1, 4, 16 or 32 GiB. ATTO also showed consistently identical maximum values, regardless of the file size and duration of use of the SSD. This stability suggests a performance potential that is not reflected in reality.

In fact, the SPECwpc test series shows that under real load, i.e. in professional applications with a high proportion of metadata, variable access depth and realistic access patterns, the transfer rates collapse to below 400 MB/s and sometimes significantly below. These values are not comparable with the alleged 14,000 MB/s from synthetic benchmarks. Even in the best cases, the real write and read values are 5 to 10 times lower than the laboratory values. This shows: The synthetic peak rates are little more than a maximum figure for marketing purposes. They do not reflect thermal behavior, reorganization processes or access reality.

It is also obvious that the firmware of many SSDs, including the Biwin X570 PRO, is specifically optimized for benchmarks such as CrystalDiskMark. The SM2508G reacts in these tests with maximum parallelization and buffer strategy, while under real workloads a different prioritization obviously takes place. In particular, the varying results with prodDEV and Maya Venice indicate typical effects of mapping and garbage collection logic under real conditions.

In real workstation applications, the Biwin X570 PRO proves to be a powerful SSD that can keep up with other high-end models in many scenarios. It achieves good to very good values for sequential access in real applications and is strong in data-intensive, read-heavy workloads. However, its performance in real-world tests is far from the maximum values shown synthetically. Those who base their choice of SSD exclusively on benchmarks that run under ideal conditions may be disappointed in productive environments with intensive workloads. The SPECwpc results provide a much more honest picture in this respect.