Western Digital has officially introduced its first data center-class SSDs using 3D NAND flash. The new HGST Ultrastar SS300 SSDs are designed for SAS applications with mixed workloads that demand drives with endurance but are not the highest performance possible. The new drives were co-developed with Intel, so they do not have various proprietary SanDisk technologies, such as the Guardian (as some previous-gen Ultrastar SS), but support numerous new features instead.

The Ultrastar SS300 drives use Intel’s 3D NAND memory - which we suspect is their recent 64-layer TLC & MLC - as well as a proprietary controller that was presumably co-designed with Intel. HGST does not disclose many details regarding the chip, but it might be the same controller that powers the recently launched Intel SSD DC P4500 and P4600 drives, albeit featuring a different configuration and a different firmware. From the feature set, the Ultrastar SS300 controller supports extended error correction code (ECC with a 1x10^-17 bit error rate) to enable higher performance and data integrity, exclusive-OR (XOR) parity (in case a whole NAND die fails) as well as parity-checked internal data paths. In addition, the Ultrastar SS300 complies with the T10 Data Integrity Field (DIF) standard, which requires all interconnect buses to have parity protection (on the system level), as well as a special power loss data management feature that does not use supercapacitors. Western Digital will offer various Ultrastar SS300 models with features like TCG encryption, TCG+FIPS, and instant secure erase to comply with different security requirements.

The HGST Ultrastar SS300 SSDs are aimed at various market segments, including financial transactions, e-commerce, virtualization, database analytics, etc. that generate read-intensive and mixed-use workloads and thus need both performance and reliability. The new SSDs come in a 2.5”/15 mm form-factor with SAS 12 Gb/s interface and are drop-in compatible with existing servers. Meanwhile, when compared to direct predecessors, the Ultrastar SS200, the new SS300 offer similar capacities — from 400 GB to 7.68 TB, but the new drives excel the ancestors in most possible metrics (endurance, performance, power, etc.).

When it comes to performance, the HGST Ultrastar SS300 features sequential read speeds of up to 2100 MB/s (+20% vs. the SS200) as well as sequential write speeds of up to 2050 MB/s (3D MLC, +105%) or 1250 MB/s (3D TLC, +25%). Random read performance of the Ultrastar SS300 is up to 400K (+60%), while random write performance is rated at 170-200K for 3D MLC-based performance/endurance-optimized models as well as 115K-130K for capacity-optimized 3D TLC-powered drives. Interestingly, in addition to 9 W and 11 W power envelopes supported by the predecessor, the SS300 also supports a 14 W mode that unlocks “additional performance capabilities” (i.e., IOPS) as well as an ultra-low power setting.

HGST Ultrastar SS300 Series Specifications
  3D MLC NAND-Based 3D TLC NAND-Based
Capacities 400 GB
800 GB
1,600 GB
3,200 GB
480 GB
960 GB
1,920 GB
3,840 GB
7,680 GB

1,920 GB
3,840 GB
7,680 GB
Form Factor 2.5"/15mm
Interface SAS 6/12 Gb/s, dual port for 12 Gb/s
Controller Proprietary
NAND 3D MLC NAND 64-layer 512 Gb
3D TLC NAND
Sequential Read 2100 MB/s
Sequential Write 2050 MB/s 1250 MB/s 1200 MB/s
Random Read (4 KB) IOPS 400,000
Random Write (4 KB) IOPS 200,000 170,000 120,000 80,000
Mixed Random R/W (70:30 R:W, 4KB)
Max IOPS
285,000 265,000 130,000 115,000
Write Latency 512 B 85 ms unknown
Power Idle 3.7 W (<3.2 TB) - 4.7 W (>3.2 TB)
Operating 9 W, 11 W, 14 W (configurable)
Endurance DPWD 10 3 ~1 ~0.5
Max. PB 400GB: 7PB
800GB: 15PB
1.6TB: 29PB
3.2TB: 59PB
400GB: 2PB
800GB: 4.5PB
1.6TB: 8.5PB
3.2TB: 17.5PB
480GB: 0.9PB
960GB: 1.75PB
1.92TB: 3.5PB
3.84TB: 7PB
7.68TB: 14PB
1.92TB: 1.8PB
3.84TB: 3.5PB
7.68TB: 7PB
Encryption AES-256 (?)
Power Loss Protection Yes
MTBF 2.5 million hours
Warranty Five years or max PB written (whichever occurs first)
Models HUSMM3232ASS20x
HUSMM3216ASS20x
HUSMM3280ASS20x
HUSMM3240ASS20x
HUSMR3232ASS20x
HUSMR3216ASS20x
HUSMR3280ASS20x
HUSMR3240ASS20x
HUSTR7676ASS20x
HUSTR7638ASS20x
HUSTR7619ASS20x
HUSTR7696ASS20x
HUSTR7648ASS20x
HUSTV1576ASS20x
HUSTV1538ASS20x
HUSTV1519ASS20x
Legend for Model Numbers H = HGST
U = Ultrastar
S = Standard
MM = NAND type/endurance
(MM=MLC/mainstream endurance, MR=MLC/read-intensive,
TR=TLC/read intensive TV=TLC very read intensive)
32 = Full capacity (3.2TB)
32 = Capacity of this model
(76=7.6TB, 38=3.84TB 32=3.2TB, 19=1.92TB, 16=1.2TB,
96=960GB, 80=800GB, 48=480GB, 40=400GB)
A = Generation code
S = Small form factor (2.5" SFF)
S2 = Interface, SAS 12Gb/s

x in Model Number denotes Encryption level:
0 = Instant Secure Erase
1 = TCG Encryption
4 = No Encryption, Secure Erase
5 = TCG + FIPS

As for endurance, various versions of the HGST Ultrastar SS300 support 0.5, 1, 3 or 10 writes per day, thus extending addressable markets of the Ultrastar SS-series to applications that require extra endurance at 10 DPWD. All the Ultrastar SS300 SSDs are rated for 2.5 million hours MTBF and are covered with a five-year limited warranty (or the max PB written, whichever occurs first).

Western Digital has already started to ship its SS300 SSDs to select OEM partners, including Oracle. The company does not announce pricing of the drives since they depend on configurations, volume purchasing and other factors.

Related Reading:

Source: Western Digital

POST A COMMENT

3 Comments

View All Comments

  • Sivar - Monday, May 22, 2017 - link

    "...as well as a special power loss data management feature that does not use supercapacitors."
    Why disallow super-capacitors?
    Reply
  • DanNeely - Monday, May 22, 2017 - link

    The linked spec doesn't talk about power loss, and I don't have time to read it all. But the way this article is written I read it as the spec requires handling it somewhere other than on the drive. Or phrased differently, the drive doesn't need caps to meet the standard; not that having caps is forbidden by it. That said I am curious what the standard is myself; is it just some sort of super journaling/etc, or moving power loss control up from individual drives to the drive array (eg the enclosure) as a whole. Reply
  • neo_1221 - Tuesday, May 23, 2017 - link

    If you read the datasheet in the link, there's a section that reads "an exclusive power loss data management feature that does not require supercapacitors." Not really any information that isn't already in the AT article. I'm also curious what they're using instead of supercaps. Reply

Log in

Don't have an account? Sign up now