We divided our performance testing into two sections. On this page, we use a single drive in a desktop system to measure baseline performance. NAS-focused disk drives have become very popular in desktop systems as secondary drives. The increased vibration resistance increases the performance consistency in systems with several fans and other cooling components, like liquid pumps. Vibration will slow all hard disk drives, even those we're testing today, but the additional sensors and firmware optimizations help ensure the best performance.
All the drives in our tests are built for NAS environments. We included the two latest Seagate IronWolf products from the Guardian Series and Toshiba's NAS-focused 8TB N300. We also included the Red Pro 6TB and HGST Deskstar NAS 6TB, which are two older 7,200-RPM drives from Western Digital.
Sequential Read Performance
To read about our storage tests in-depth, please check out How We Test HDDs And SSDs. We cover four-corner testing on page six of our How We Test guide.
HDDs do not scale with increased sequential workloads because queue depth (QD) 1 to 32 performance is nearly identical. Increasing the queue depth can actually decrease performance slightly because it causes additional arm movement rather than just reading the tracks sequentially (if the data is defragmented and lined up optimally).
In this section, the first two charts have the most value. The first chart shows the sequential read performance across the entire platter. Drive makers have learned to use the fastest (outer) portion of the platter as a cache to aid the DRAM. Performance is much slower on the inner tracks of the spinning disks. The two Helium-filled Red HDDs are on the lower end of the performance chart, but this is a basic test. You can use this chart as a guide for sequential performance as you fill the drive. The 10TB Red's performance drops from 220 MB/s to ~180 MB/s once the drive is filled halfway with data. The 8TB Red drops from 200 MB/s to roughly 150 MB/s.
The second chart shows performance on the outer tracks with all of the drives under the same conditions. The third chart breaks the results down into QD2 performance. The 8TB Red is the slowest with 200 MB/s, and the Red 10TB slots in the middle with 217 MB/s.
Sequential Write Performance
Sequential write throughput is nearly identical to read performance. The drives do not use the DRAM buffers to accelerate sequential write performance, so sequential reads and writes are almost always identical with consumer-focused products. This is the performance you can expect if you are simply transferring large multimedia files between the drives and not mixing reads and writes.
Random Read Performance
Random performance scales with additional queue depth. It's much easier to reach elevated queue depths with HDDs during real-world workloads because they are slower than SSDs. Still, disk manufacturers have found ways to increase random performance via increased buffer capacity and using the outer edges of the platters. The Helium-filled Red HDDs have strong random read performance. The new 12TB Seagate IronWolf drive is slightly faster, though.
Random Write Performance
The 10TB Red has a 256MB DRAM buffer and the 8TB only has a 128MB buffer, so we expected a bigger performance delta between the two dives during the random write workload.
80% Mixed Sequential Workload
We describe our mixed workload testing in detail here and describe our steady state tests here.
The Seagate IronWolf 12TB put on an impressive show in the four-corner tests. The Reds with Helium perform better at lower queue depths with a synthetic application (80% read) workload. The newer Reds also outperform the older 6TB Red Pro. The older 6TB HGST Deskstar NAS and new 8TB Toshiba N300 trail by a substantial margin.
80% Mixed Random Workload
The mixed random workload is the most likely to benefit from the additional DRAM cache in the Red 10TB, and we see it move well past the 8TB model during the test. Again, we're using a workload mixture commonly seen in desktop environments.
Sustained Mixed Sequential Workload
The chart shows us performance with 100% reads on the left, and 100% writes (or 0% reads) on the right. A 50% mix would be roughly equivalent to reading one file from the drive while writing another. For instance, in a perfect world, reading two media files while writing another would give you a 66% read workload.
Writing 4KB random data to the full LBA range does not benefit from a performance increase due to the extra buffer capacity. That really doesn't matter because the new Helium-filled Reds join their kin at the top of this chart. They are well above the competition.
Here is the wacky thing: Modern disk drives deliver higher random steady-state performance than entry-level planar TLC NAND SSDs.
PCMark 8 Real-World Software Performance
For details on our real-world software performance testing, please click here.
But don't get any ideas about disks making a comeback for your system drive, not without Optane Memory in front of it. None of the products in the charts were designed to run as your main system drive, so these tests are outside of the recommended use case.
Application Storage Bandwidth
We still like to run the test to gauge system performance and to compare to other products. The two new Red products fall to the bottom of our chart behind the rest of the drives.
MORE: Best SSDs
MORE: How We Test HDDs And SSDs
MORE: All SSD Content
They sell for less here, and is remarkable quiet (about same, or lower, as my current WD green 5400rpm, that they replaced, vs the 7200rpm).
They use quite low power as well.
And yes, i would else agree with 5400rpm is better for consumers for archive usage.