HGST 0B26312 300GB SAS 15K Ultrastar HDD: Spindle Count Explained
Three hundred gigabytes on a drive that costs more than a two-terabyte one looks absurd until you understand what you are buying. You are not buying storage. You are buying a spindle, and in performance-tier arrays spindles are the unit that matters.
The HGST Ultrastar 0B26312 is a 300GB SAS 6Gb/s 15K 2.5-inch drive, and its small capacity is a consequence of its speed rather than a limitation someone forgot to fix.
Why 15K drives have small capacities
Physics sets the ceiling here, not marketing segmentation.
A platter spinning at 15,000 RPM experiences considerably more mechanical stress than one at 7200 RPM. Keeping the head flying stably over the surface at that speed requires tighter tolerances, and the practical result is smaller platters and fewer of them. Smaller platters mean less surface area, and less surface area means less capacity.
Heat and power follow the same curve. A 15K drive draws more power and produces more heat than a nearline drive, which constrains how densely you can populate a chassis with them.
So the trade is explicit: you give up capacity per drive and accept higher power draw, and in return you get the lowest seek latency available from mechanical storage. For workloads where a transaction waits on a disk read, that trade has been worth making for two decades.
0B26312 specifications
| Part Number | 0B26312 |
| Manufacturer | HGST / Western Digital |
| Product Family | Ultrastar |
| Capacity | 300GB |
| Interface | SAS 6Gb/s |
| Rotational Speed | 15,000 RPM |
| Form Factor | 2.5 inch SFF |
| Drive Class | Enterprise Performance |
| Condition | Refurbished, function-tested |
| Warranty | 30-day TechSellerUSA warranty |
Spindle count is the design principle
This is the concept that makes small 15K drives make sense, and it is the one most often missed when someone tries to "upgrade" an array.
Every drive has one set of heads. A drive can service one request at a time from its physical mechanism, so the number of concurrent operations an array can handle scales with the number of drives, not with total capacity.
Sixteen 300GB drives and four 1.2TB drives both give you roughly 4.8TB of raw capacity. The sixteen-drive array delivers approximately four times the IOPS, because it has four times the heads working in parallel.
For a database, a virtualisation datastore or a busy application server, that difference is the entire reason the array was built the way it was. Consolidating onto fewer larger drives to "simplify" it quietly removes three quarters of its performance.
This is also why replacing a failed 300GB drive with a 1.2TB one is not an upgrade. The array uses 300GB and strands the rest, and you have gained nothing while spending more.
SKU: 0B26312 | Refurbished & function-tested | 30-day warranty
View Details & Request a Quote →15K against flash, honestly
SSDs beat 15K drives on latency by an enormous margin. That is not in dispute, and anyone specifying a new latency-critical tier today should be looking at flash.
Where 15K still holds ground is in two situations.
Existing arrays. A working array built on 15K spindles does not become faster if you drop one SSD into it. Mixed drive types within a RAID group degrade to the slowest member, so a partial swap achieves nothing. Replacing a failed member like for like is the correct action, and a full flash migration is a separate project with its own budget.
Sustained write workloads. Enterprise SSDs are rated by drive writes per day, and write-heavy roles need mixed-use or write-intensive models at a significant premium. A mechanical drive has no equivalent wear ceiling — it fails over time rather than wearing out through writes. For heavy sustained writes that do not need flash latency, that removes a whole cost dimension.
Bare drive, no carrier
This is the manufacturer part with no server carrier and no OEM firmware coding. It suits Supermicro, white-box and generic-tray chassis, and costs less than the OEM-branded equivalent of the same mechanism.
It is the wrong choice where an HPE Smart Array, Dell PERC or IBM ServeRAID controller expects vendor firmware, or where you need the tray supplied. Our compatibility guide lists which part numbers apply to which platform.
Ordering
We supply the 0B26312 refurbished and function-tested with a 30-day warranty. Performance arrays are usually populated in sets, so tell us how many bays you are filling and we will quote accordingly. Browse the full Western Digital and HGST range or our complete enterprise hard drives catalogue.
Common questions
Why do 15K drives have such small capacities?
Physics. A platter spinning at 15,000 RPM experiences far more mechanical stress, so keeping the head flying stably requires smaller platters and fewer of them. Less surface area means less capacity, and heat and power draw rise at the same time.
Why not use fewer, larger drives instead?
Because IOPS scale with spindle count, not capacity. Sixteen 300GB drives and four 1.2TB drives give similar raw capacity, but the sixteen-drive array delivers roughly four times the concurrent operations because it has four times the heads.
Can I put one SSD into a 15K array to speed it up?
No. Mixed drive types within a RAID group degrade to the slowest member, so a partial swap achieves nothing. Replace failed members like for like, and treat a flash migration as a separate project.
Do mechanical drives wear out from writes like SSDs?
No. Enterprise SSDs are rated by drive writes per day and write-heavy workloads need costlier mixed-use or write-intensive models. A mechanical drive fails over time rather than wearing out through writes, which removes a cost dimension on sustained write workloads.
Will a 6Gb/s drive work in a 12Gb/s backplane?
Yes, at 6Gb/s. SAS negotiates down to the lower of the two speeds, and on a mechanical drive the spindle is the bottleneck long before the interface is, so the link speed is not the constraint.
Populating a performance array? Tell us the bay count and we will quote the full set plus a spare.
