At the time of writing, the current stable release of unRAID is unRAID 4.7. While 4.7 is a great product, one of its limitations is that it uses MBRs and not GPTs. In layman’s terms, that means that it is incompatible with any hard drive larger than 2.2 TB. As 2.5 TB and 3 TB drives are on the market today at attractive price points, many unRAID users have switched to using the latest unRAID beta (currently unRAID 5.0beta11), which has both MBR and GPT support. This means that any capacity drive can be used as a parity, data, or cache drive in the latest unRAID beta. I recently got my hands on 15 of the fabulous 3 TB Hitachi DeskStar 5K3000 CoolSpin hard drives. These are green drives that spin at 5400 RPM and use a SATA III (6.0 Gb/s) interface. I took the opportunity to test as much hardware as I had available to me for 3 TB hard drive compatibility. Here is the hardware I tested:
Supermicro X7SLA-H with a built-in Intel Atom CPU and 2 GB of DDR2 533 RAM (2 x 1GB)
ZOTAC GF6100-E-E with an AMD Sempron 140 CPU and 2 GB of DDR2 800 RAM (1 x 2GB)
Biostar A880G+ with an AMD Sempron 140 CPU and 2 GB of DDR3 1333 RAM (1 x 2GB)
Supermicro X8SIL-F-O with an Intel i3-540 CPU and 4 GB of DDR3 1333 RAM (2 x 2GB)
Asus M4A78LT-M with an AMD Sempron 140 CPU and 2 GB of DDR3 1333 RAM (1 x 2GB)
I also tested the 2 port PCIe x1 SIL3132 card that we use in certain GreenLeaf builds. I decided not to run any thorough tests on the Supermicro AOC-SASLP-MV8 controller as it has already been well established as being fully compatible with 3 TB drive through tests conducted by other members of the unRAID community.
All tests were performed using unRAID 5.0beta10 (which was the latest beta available at the time) and Joe L.’s preclear script 1.12beta (which is the only version of preclear currently available that supports 3 TB drives). A hardware component passed the test if it precleared successfully and was recognized by unRAID as an array drive. Here are the results:
| Motherboard | Backplane | SATA Controller | Result | Duration (HH:MM:SS) |
|---|---|---|---|---|
| Supermicro X7SLA-H | Norco SS-500 | Onboard | PASS | 48:49:35 |
| Supermicro X7SLA-H | Norco SS-500 | Onboard | PASS | 47:41:13 |
| Supermicro X7SLA-H | Norco SS-500 | Onboard | PASS | 48:28:24 |
| Supermicro X7SLA-H | Norco SS-500 | Onboard | PASS | 48:55:51 |
| Supermicro X7SLA-H | Norco SS-500 | SIL3132 | PASS | 54:47:52 |
| Supermicro X7SLA-H | Norco SS-500 | SIL3132 | PASS | 53:24:05 |
| ZOTAC GF6100-E-E | Kingwin 3-in-2 | Onboard | PASS | 41:29:50 |
| ZOTAC GF6100-E-E | Kingwin 3-in-2 | Onboard | PASS | 42:06:13 |
| ZOTAC GF6100-E-E | Kingwin 3-in-2 | Onboard | PASS | 40:28:49 |
| ZOTAC GF6100-E-E | Kingwin 3-in-2 | Onboard | PASS | 43:02:11 |
| ZOTAC GF6100-E-E | Kingwin 3-in-2 | SIL3132 | PASS | 46:49:55 |
| ZOTAC GF6100-E-E | Kingwin 3-in-2 | SIL3132 | PASS | 47:30:24 |
| Biostar A880G+ | Top Dock | Onboard | PASS | 41:58:44 |
| Biostar A880G+ | None | Onboard | PASS | 42:25:21 |
| Biostar A880G+ | None | Onboard | PASS | 41:36:09 |
| Biostar A880G+ | None | Onboard | PASS | 42:44:09 |
| Supermicro X8SIL-F-O | None | Onboard | PASS | 42:12:56 |
| Supermicro X8SIL-F-O | None | Onboard | PASS | 42:09:42 |
| Supermicro X8SIL-F-O | None | Onboard | PASS | 41:22:09 |
| Supermicro X8SIL-F-O | None | Onboard | PASS | 37:45:11 |
| Supermicro X8SIL-F-O | None | Onboard | PASS | 38:29:49 |
| Supermicro X8SIL-F-O | None | Onboard | PASS | 41:11:57 |
| Asus M4A78LT-M | None | Onboard | PASS | 49:01:51 |
| Asus M4A78LT-M | None | Onboard | PASS | 48:44:29 |
| Asus M4A78LT-M | None | Onboard | PASS | 48:38:10 |
| Asus M4A78LT-M | None | Onboard | PASS | 48:01:27 |
| Asus M4A78LT-M | None | Onboard | PASS | 47:44:05 |
| Asus M4A78LT-M | None | Onboard | PASS | 45:58:00 |
The good news is that every single piece of hardware I tested is fully compatible with the Hitachi 3 TB drives. However, as you can see from the duration results above, some of the preclear cycles were slower than others. At first I thought that certain SATA controllers were slower than others. Here’s a quick analysis of that hypothesis:
| Motherboard | SATA Controller | Average Duration (hours) |
|---|---|---|
| Supermicro X7SLA-H | Onboard | 47.75 |
| Supermicro X7SLA-H | SIL3132 | 53.50 |
| ZOTAC GF6100-E-E | Onboard | 41.50 |
| ZOTAC GF6100-E-E | SIL3132 | 46.50 |
| Biostar A880G+ | Onboard | 41.50 |
| Supermicro X8SIL-F-O | Onboard | 40.17 |
| Asus M4A78LT-M | Onboard | 47.50 |
The slowest set of hardware was the Supermicro X7SLA-H and the SIL3132 controller with an average duration of 47.5 hours. This happens to the be motherboard with the slowest CPU and slowest RAM as well. The fastest set of hardware was the Supermicro X8SIL-F-O with an average duration of 40.17 hours. This also happened to be the motherboard with the fastest CPU and the most RAM. I believe these test results show that when preclearing 3 TB drives, the speed of the CPU and the RAM installed matters more than the SATA controller being used. Given this revised hypothesis, here’s the take-home analysis of this data:
| CPU | Amount of RAM | RAM Speed | Channels | Average Duration (hours) |
|---|---|---|---|---|
| Atom | 2GB | DDR2 533 | Dual | 50.63 |
| Sempron 140 | 2GB | DDR2 800 | Single | 44 |
| Sempron 140 | 2GB | DDR3 1333 | Single | 44.5 |
| i3-540 | 4GB | DDR3 1333 | Single | 40.17 |
If you plan to preclear a lot of 3 TB drives, more RAM and a faster processor can help speed up the process by as much 10 hours.