CrystalDiskMark results
CrystalDiskMark offers a very comparable and practical assessment of sequential performance and is particularly suitable for verifying manufacturer specifications. The measurements of the WD Red Pro 26 TB are exactly at the level of the data sheet specifications (up to 272 MB/s) and thus confirm the high sequential performance of this hard disk. At the same time, one should be aware of the limitations of this benchmark. CDM tests defined access patterns under idealized conditions with an empty drive, typically in the initial area of the volume. In real applications, fragmentation, parallelized accesses, thermal effects and physical scattering on the disk surface occur, which are not mapped in CDM. With SSDs in particular, cache strategies can have an excessive influence on the result, but this hardly plays a role with mechanical hard disks such as the WD Red Pro.
Truly practical statements on video stream performance or long-term stability cannot be derived from CrystalDiskMark alone. Tools such as AJA System Test or long-term logging under real workload are much better suited for this. Nevertheless, the consistent CDM results across different file sizes show that the WD Red Pro 26 TB maintains its transfer rates regardless of the test load. This is a strong quality feature for applications such as video recording, data archiving or NAS use. The four CrystalDiskMark results shown document the sequential and random read and write performance of the WD Red Pro 26 TB using different test file sizes: 1 GiB, 4 GiB, 8 GiB and 32 GiB. This gradation not only allows a statement to be made about the peak values, but also gives an indication of the behavior with increasing buffer fill level and thermal and logical load.
 |
 |
 |
 |
Sequential performance (Q32T1)
The sequential transfer rates remain remarkably constant across all tests. When reading, the values fluctuate between 267.8 and 277.2 MB/s, when writing between 259.1 and 267.0 MB/s. The differences are small and fall within the range of typical measurement tolerances. The decisive factor is that even with the largest test file with 32 GiB, no dips are recognizable. This indicates a consistently high buffer size, efficient cache management and robust disk mechanics. For a hard disk with CMR recording and ten platters, this behaviour is a clear indicator of an architecture designed for continuous load without an intermediate, volume-limited flash buffer, as used by many inexpensive SSDs.
Random 4K performance
There is a clear trend for random accesses in the 4 KiB range. The read rates (especially Q1T1) decrease slightly with increasing test size. While 0.934 MB/s is still achieved in the 1-GiB test, this value drops to 0.697 MB/s in the 32-GiB test. This reduction is typical for mechanical drives, as the servomechanics have to be repositioned more frequently with larger test windows, which slows down the IOPS. Write operations, on the other hand, are much more robust and remain at a stable high level of 2.669 to 4.280 MB/s. This is due to the fact that write operations are often buffered and optimized in the HDD firmware design, while random read accesses are directly limited by the mechanics.
The CrystalDiskMark evaluation confirms the high sequential transfer rate of the WD Red Pro 26 TB with stable values of up to just under 270 MB/s for read and write, regardless of the test size. The performance for random accesses is, as expected, limited, but consistent. The benchmark results thus support the previous AJA and thermal measurements: This hard disk was clearly developed for sequential continuous loads and performs this task reliably and without thermally or mechanically induced performance drops. For workloads with high IOPS requirements, however, it is only suitable to a limited extent, which is due to the mechanical nature of the medium.
ATTO Benchmark
ATTO provides significantly more insight into the behavior of hard drives with block-by-block load distribution than other benchmarks. The WD Red Pro 26 TB shows consistent, stable behavior here, both when reading and writing. Full performance is achieved from block sizes of 128 to 256 KB, which is typical for applications that move large files, such as video recording, data archiving or backup streams. The combination of consistently high performance from 128 KB, no dips at 32 GB test size and high synchronization between read and write makes it clear that the firmware and hardware are designed for sequential workloads. Only enterprise models with a similar number of platters and CMR recording deliver comparable behavior in this class. For scenarios with many small files (under 16 KB block size), however, the performance is limited, which is not surprising for a NAS HDD in this class.
The two results shown from the ATTO Disk Benchmark document the sequential read and write performance of the WD Red Pro 26 TB over a wide range of block sizes (512 B to 64 MB), each with two different test data sizes: once 1 GB, once 32 GB. In contrast to CrystalDiskMark, ATTO works in a more deterministic, block-oriented manner and allows a very fine resolution of the I/O behavior depending on the block size. For mechanical hard disks in particular, this is an important addition to the classic sequential measurement.
 |
 |
Sequential throughput with 1 GB file
Stable sequential transfer rates are already achieved from a block size of 64 KB. From 256 KB, the write rate value is consistently around 258 to 261 MB/s, the read rate value is 264 to 266 MB/s. There are no individual outliers, which indicates a well-tuned firmware and consistent servomechanics. The performance peaks are in the range of the maximum values determined in earlier AJA and CDM measurements and confirm the results obtained there. As expected, the lower block size range (below 16 KB) shows severe limitations: The read rate at 512 bytes, for example, is only 38.21 MB/s, while the write rate is 34.30 MB/s. This range is not relevant in practice for typical video and NAS workloads, but it is for many small I/O operations in file systems with a lot of metadata (such as when unpacking large directory trees or email archives). The slow rates are due to the mechanical structure, as no significant optimization through caching or prefetching can take effect.
Sequential throughput with 32 GB file
The second test with a 32 GB test file size shows practically identical results. Here too, a plateau is reached from 256 KB block size. The write values are between 254 and 258 MB/s, the read rates between 260 and 266 MB/s. There is no noticeable drop due to thermal throttling, buffer overflow or other limitations at higher test loads. This is by no means a matter of course with hard disks, as many models show a slight loss of performance with larger transfers due to servo readjustment or head switching. The fact that this is not the case here speaks for the robust CMR architecture of the WD Red Pro 26 TB. As expected, the values in the area of smaller blocks are slightly weaker than in the 1 GB test, which can be attributed to the longer test duration and greater mechanical scattering. Nevertheless, the differences remain within reasonable limits, which also speaks for consistent performance across the entire disk surface.
Summary and conclusion
The WD Red Pro 26 TB has shown a consistently stable and expected behavior in all measurements carried out, at least as far as one can judge with a single unit. Because one thing is also a fact: at over 600 euros, this hard disk is so expensive that apparently not even the manufacturer itself could afford two of them to enable a RAID scenario as a press sample. Accordingly, as the lowest link in the hardware supply chain, I had to make do with exactly one hard disk, which effectively ruled out comparisons with RAID 1, RAID 5 or even the effects of vibration in multi-HDD environments, but was more in line with my financial constraints.
Nevertheless, there is a lot to prove with the available data. According to AJA, ATTO and CrystalDiskMark, the sequential transfer rates are stable in the range between 250 and 270 MB/s for read and write operations, without any dips with longer test durations or larger block sizes. The ATTO measurements show a very early plateau from a block size of 256 KB, which indicates good buffer management and an efficient access structure. Even with a 32 GB test block, the performance remains consistently high, which confirms its practical use for long-term streaming and backup.
The power consumption under various load scenarios, from idle to random access read, also shows a good-natured picture. At idle, the consumption is around 6 W, with sequential video streaming at around 8.5 W and with random IO accesses at around 13 W. Critical peaks of up to 30 W were only briefly reached during spindle startup, which can be quite relevant in multi-bay NAS with an insufficient power supply. Thermographically, the disk remained below 50 °C even after an hour of synthetic full load; in productive OBS server use with two active streams, the housing temperature leveled off at around 36 °C. As expected, the rear panel showed hotspots at the typical controller and VRM positions, but without any alarming signs of thermal saturation.
In terms of random IO performance, mechanical limitations naturally remain. 4 KiB read rates below 1 MB/s at Q1T1 are not unexpected, but irrelevant in classic NAS or streaming workflows anyway. The decisive factor is the ability to continuously and sequentially retrieve data, and here the WD Red Pro impresses with consistency, not spectacle. Precisely because the WD Red Pro 26 TB is technically anything but new and has been on the market for some time, it was important to me to shed light on aspects that were rarely found in previous reviews. Long-term current measurement, real thermography during stream recording, power spikes during startup or the scaling in ATTO and AJA tests over large test windows; all this provides additional insights that you often look for in vain in the usual data sheet-based short tests.
Even though I had to make do with a single disk, the goal was clear: to get as close as possible to a practical overall picture. After all, if a hard disk costs well over 600 euros, you can at least expect that it can be evaluated transparently in every respect, even if it doesn’t end up being a benchmark orgy in a RAID array. But others are responsible for that.
The Western Digital WD Red Pro 26 TB was provided on loan for this test without any obligation or preconditions. There was no obligation to publish.
Kommentieren