Dell PERC and HPE Smart Array RAID controllers: how to read the name, cache, cables and backplanes
Updated
In brief. The controller name encodes its class and installation method. For Dell: H3xx — entry level without cache, H7xx — mid-range with cache and battery, H8xx and H9xx — high-end with external ports. For HPE: E208 — without cache, P408 and P816 — with cache, MR416 — the new MegaRAID line. The letters at the end matter more than the digits: i — internal ports, e — external, a — a module in the dedicated motherboard connector, p — a standard PCIe card. In the Servix catalog Dell PERC H345 costs 4,312,941 UZS, Dell PERC H755 with battery — 8,625,882 UZS.
Select a controller for your server → — tell us the server model and number of drives, and we will check it against the backplane and cables.
Quick links: how to read a Dell PERC name · HPE Smart Array and MegaRAID · cache and battery · cables and backplanes · RAID or HBA · what is in the catalog
How to read a Dell PERC name
PERC stands for PowerEdge RAID Controller, and its numbering runs continuously across generations. The hundreds indicate the class, not the speed.
Designation | Class | Cache | Typical use |
|---|---|---|---|
H3xx (H330, H345, H355) | entry level | no | mirrored boot pair, simple RAID 1 and 10 arrays |
H7xx (H730, H740P, H745, H755) | mid-range, the workhorse | 2–8 GB + battery | main data array, RAID 5, 6 and 10 for virtualization and databases |
H8xx (H830, H840) | high-end with external ports | yes | connecting external JBOD disk shelves |
H9xx (H965) | current PERC 12 generation | 8 GB | 24G SAS and NVMe, new PowerEdge |
HBA (HBA330, HBA355i) | no RAID | no | drive passthrough for ZFS, vSAN, Ceph and Storage Spaces |
The letters after the number matter more than the digits themselves. P at the end (H730P, H740P) means a larger cache. i — a controller with internal ports, e — with external ports, for shelves. The form factor is indicated separately: Mini Mono and Mini Monolithic go into a dedicated slot on the board and do not take up PCIe, Adapter — a standard card. So “H730” without the form factor specified does not yet answer whether it will fit a particular server.
Generations correspond to server generations: PERC 9 (H330, H730, H830) is the 13th generation of PowerEdge (the H330 was also installed in the 14th), PERC 10 (H740P, H840) is the 14th; in the 15th, late PERC 10 models — H345 and H745 — were added, and PERC 11 (H355, H750, H755) became the main one; in the 16th — PERC 11 (H355, H755) and PERC 12 (H965i). A previous-generation controller as a rule does not fit a new server: both the cable connectors and the backplane change.
HPE: Smart Array, MegaRAID and boot modules
Family | What it is | Cache | Catalog example |
|---|---|---|---|
E208 | entry-level SR Gen10, essentially an HBA with simple RAID levels | no | HPE Smart Array E208i-p — 5,750,588 UZS |
P408 | mid-range SR Gen10 with cache and battery | 2 GB | HPE Smart Array P408i-p — 7,188,235 UZS |
P816 | sixteen internal lanes for dense drive cages | 4 GB | HPE Smart Array P816i-a — price on request |
MR416 | new MegaRAID line for Gen10 Plus and Gen11, supports NVMe | 4 GB | HPE MegaRAID MR416i-p — 9,344,705 UZS |
NS204i | boot module: two mirrored NVMe M.2 drives for the hypervisor | — | HPE NS204i-p — 10,063,529 UZS |
Suffixes are read the same way as for Dell: i — internal ports, e — external, a — a modular controller in the dedicated board connector (AROC), p — a PCIe card. The pair P408i-a and HPE Smart Array P408i-p are the same controller in two versions, and the choice between them is determined not by performance but by whether the server has a free slot and the dedicated connector.
Set apart is the NS204i. It is not a RAID controller for data but a boot device: two M.2 drives in a mirror so that the hypervisor does not take up bays in the main drive cage. The practice is simple — the system on the NS204i, the data on the main controller.
Cache and battery: why they are not optional
A write cache turns small random writes into sequential ones and makes a several-fold difference on RAID 5 and 6. But it is enabled only with a healthy battery or supercapacitor: without backup power the controller switches to write-through mode and writes directly to the drives, losing the entire gain.
- The battery is a consumable. After three to four years it degrades, the controller disables write-back on its own, and the server “suddenly slows down” without a single error in the drive logs. This is the most common cause of complaints about a slow array after several years of operation.
- Controllers without cache (H345, E208) perform poorly on RAID 5 and 6 — they belong in mirrors or in drive passthrough mode.
- Battery status should be checked through monitoring, not once a year: the status is visible in iLO and iDRAC and should be added to the alerting system.
How RAID levels relate to drive capacity and rebuild time is covered in the article RAID and the server disk subsystem.
Cables and backplanes: why you can't just move a controller and expect it to work
The most common question about PERC is whether a cable from one server will fit another. The short answer: the cable is chosen not by the controller but by the “server + backplane + controller” combination, and it has its own part number.
- Connectors differ by generation. On 12-gigabit SAS controllers these are internal Mini-SAS HD (SFF-8643) and external SFF-8644; on new platforms, SlimSAS (SFF-8654), and Gen11 servers also have MCIO. They are mechanically incompatible.
- Cable length and routing are tied to the chassis. 1U and 2U have different geometry; a cable from an 8SFF cage will not reach in a 24-drive cage, and vice versa.
- Backplanes come in two types. With an expander — one connector serves all drives in the cage; with direct attach — each group of four drives has its own lane, and the controller must have enough ports. This is exactly where questions about “cascading” come from: a backplane with an expander can be cascaded, a direct-attach one cannot, and then you need a controller with more lanes, for example P816 instead of P408.
- The number of drives is limited not only by ports. An eight-lane controller without an expander will physically serve eight drives, no matter how many bays the chassis has.
The practical order is this: first the drive cage is determined (how many drives and of what type), then the backplane, and only then the controller and its cables. The reverse order — buying a controller and then looking for a cable for it — usually ends in an unnecessary purchase.
RAID or HBA: when the controller only gets in the way
If there is software-defined storage on top — ZFS, vSAN, Ceph, Storage Spaces — hardware RAID is unnecessary and harmful: it hides the drives behind its own abstraction, interferes with integrity checks and complicates diagnostics. For such setups an HBA is used or the controller is switched to passthrough mode (HBA mode, non-RAID).
The opposite case is classic virtualization on local drives with a single host: there, hardware RAID with cache and battery is easier to maintain and more predictable. An in-between option is a boot module for the hypervisor plus an HBA for data.
What is in the catalog
Controller | Class | Price |
|---|---|---|
Dell, entry level, no cache | 4,312,941 UZS | |
Dell PERC 9, cache and battery | 5,750,588 UZS | |
Dell PERC 11, cache and battery | 7,188,235 UZS | |
Dell PERC 11, high-end | 8,625,882 UZS | |
HPE, no cache | 5,750,588 UZS | |
HPE SR Gen10, 2 GB cache | 7,188,235 UZS | |
HPE SR Gen10, 16 lanes | price on request | |
HPE MegaRAID, NVMe and SAS | 9,344,705 UZS | |
boot module, 2 × 480 GB NVMe | 10,063,529 UZS | |
LSI MegaRAID, universal card | 5,750,588 UZS |
The full list is in the section RAID controllers, drives for them are in the section SSD and HDD drives, and ready-made “drives plus controller” combinations with price-per-terabyte calculations are in the selection drives and RAID controllers.
What to check before ordering
- The server model and its generation — a previous-generation controller usually does not fit a new platform.
- Backplane type: with an expander or direct attach, and how many lanes the controller needs.
- Form factor: a modular connector on the board or a PCIe slot, and whether it is free.
- Whether drive passthrough mode is needed — if ZFS, vSAN or Ceph runs on top.
- Whether there is a battery and how old it is, if the controller is bought second-hand.
- A separate boot device for the hypervisor so as not to take up drive cage bays.
Tell us the server model and the number and type of drives — we will select the controller, cables and backplane as one kit and check them against the vendor's compatibility list. Write to us.
Sources
- Dell PowerEdge RAID Controller 11 User’s Guide (PERC H755, H750, H355, H350) — Dell
- HPE Smart Array SR Controllers QuickSpecs — HPE
RAID calculator → — usable capacity, fault tolerance and rebuild time.
Other Servix collections
- Server configurations and prices
- Server for 1C: configurations and prices
- How to choose a Dell PowerEdge
- HPE ProLiant Gen11: configurations and prices
- FortiGate by number of employees
- Dell PowerVault ME5 storage systems
- HPE MSA 2060 and 2062 storage systems
- Aruba Instant On for the office
- Network adapters, transceivers and DAC
- Xeon and EPYC server processors
- Drives and RAID controllers
- Server for video surveillance
- Dell PowerVault reference guide
- HPE MSA reference guide
- Server for virtualization
- File server and backup