Skip to main content

Micron 9100 Max NVMe 2.4TB SSD Review

Our Verdict

The Micron 9100 Max sets the new benchmark for performance with its single-ASIC architecture. The 9100 easily overpowers more complex and costly designs, and significantly outperforms single-ASIC contenders. The 9100's blend of endurance and cost, along with the advantages of NVMe, assure its success.

For

  • Superior QoS metrics
  • Fastest SSD on the market
  • 1 and 2 OIO performance with mixed workloads
  • Single ASIC design
  • Refined power/thermal envelopes

Against

  • Three year warranty (retail)
  • GUI needs work
  • QoS outliers with 50/50 read/write sequential mix
  • Low 1 and 2 OIO performance with random read/write workloads

Micron 9100 Max Makes An Entrance

Micron, one of the largest NAND fab operators, announced its newest 9100 and 7100 Series high-performance enterprise SSDs at its Storage Business Unit day earlier this year. The company now has a well-rounded range of NVMe offerings that encompass the standard AIC (Add In Card), U.2 and M.2 form factors.

Micron began its PCIe SSD adventures with the high-performance P320h in 2012, which served as the 34nm SLC NAND juggernaut for the most intense write-centric workloads. The company followed with its P420m in 2013, which employed 25nm MLC NAND and brought the same high performance levels to a more read-centric, and thus economical, design.

Micron is one of the founding members of the NVMe consortium, and is even one of the 13 promoter companies (out of 90 total member organizations), but neither of its previous PCIe SSDs employed the new interface. Micron chose to utilize its proprietary drivers instead. The custom drivers bore tremendous performance similarities to NVMe, but they did not feature its broad compatibility, which is one of the primary reasons to employ the efficient interface.

Then, curiously, everything stopped. NAND fabs tend to release a new iteration of each SSD in tandem with each new generation of NAND, but Micron did not release a new PCIe SSD with its 20nm generation of NAND. The delay left a three-year gap between new PCIe SSDs, which is an eternity in the dog years of the enterprise storage industry.

Micron languished with an aging family of PCIe SSDs as its competitors introduced new NVMe products that offered better, and more consistent, performance. Micron's new 16nm MLC-powered Micron 9100 and 7100 Series close the glaring gap and provide the company with a broad range of NVMe solutions that spans all form factors and offers multiple endurance points. The 9100 Series features the AIC and U.2 form factors for mixed and read-centric workloads, and both form factors feature the same capacity points and performance characteristics. The 7100 series (developed in tandem with Seagate) features a mix of U.2 and M.2 form factors for read-centric workloads.

Micron 9100 Series - U.2 and AICMAX 1.2 TB MAX 2.4 TBPRO 800 GBPRO 1.6 TBPRO 3.2 TB
Random Read/Write IOPS700,000/210,000750,000/300,000525,000/50,000700,000/120,000750,000/160,000
Sequential Read/Write GB/s2.9 / 1.33.2 / 2.22.05 / 0.692.8 / 1.33.2 / 2.2
Endurance PB/DWPD (4K)3.5 / 2.76.57 / 2.50.79 / .541.75 / .593.28 / 0.56
Warranty (Years)33333
Power Active/Idle (Watts)7 / 7-217 / 7-307 / 7-167 / 7-217 / 7-30
DRAM2 GB4 GB2 GB2 GB4 GB
Power Loss Protection
Raw Capacity2 TB4 TB1 TB2 TB4 TB

MORE: Best SSDs

MORE: Latest Storage News


MORE: Best Enterprise SSDs

Micron separated its 9100 series into the Max and Pro categories, which address either mixed-use or read-centric workloads, respectively. Like all SSDs, the Micron products offer various levels of performance based on capacity. The 2.4 TB Max we have in the lab represents the best of the 9100 series with up to 750,000/300,000 random read/write IOPS and 3.2/2.2 GB/s of sequential read/write throughput. The random write performance, in particular, is exceedingly impressive; it leads all NVMe SSDs on the market by a large margin.

Random write performance is one of the key performance measurements due to its impact on application performance, and the 9100 Max even outpaces more complicated (and expensive) multi-FPGA and switched architectures, such as the Intel DC P3608 and the Mangstor MX6300. These cards are designed to provide the most intense performance, but the multitude of components increase power consumption while introducing possible additional points of failure. The Micron 9100 offers more performance in several key performance metrics via a streamlined single-ASIC architecture that leverages the friendly price point of 16nm MLC NAND.

The familiar 16-channel Microsemi/PMC-Sierra Flashtec SSD controller is used in the Samsung XS1715, HGST SN150 and the Memblaze PBlaze4 SSDs that we have already reviewed, and it also powers the Techman XC100 we have in the queue. The proven and mature controller is used in many custom designs as well; we even spotted the 32-channel variant in EMC's DSSD D5.

The 9100 leverages the PCIe 3.0 x4 connection to increase performance, but NVMe lets the horses run free with 128 separate submission and completion queues (of a possible 64K/64K), and the expanded queues allow the system to spread I/O requests among more cores in the system to boost performance and reduce latency. In the past, it was impossible to break the single submission/completion-queue barrier without a custom driver. NVMe fixes that and makes its single-queue AHCI/SCSI predecessors look weak in comparison. Further, the standardized NVMe driver will allow software designers to interact with the protocol in a standardized manner, which will increase the number of applications that actually utilize the enhanced parallelism.

During the 9100's unveiling at Micron's SBU event, we noticed that the heatsink and basic design bore a remarkable resemblance to the aforementioned Memblaze PBlaze4. At the time, Micron was not ready to comment on the similarities, but we were able to grab several high-resolution photos of the PCB and compare them to the Memblaze doppelganger we have in the lab. Aside from the NAND, the surface mount components we could view were absolutely identical to the Micron 9100. Micron eventually confirmed that the 9100 is, in fact, a shared design that draws upon a new partnership with Memblaze.

The Memblaze PBlaze4 employs 15nm Toshiba NAND and Micron DRAM, but a large Micron engineering team worked with Memblaze to transition the design to Micron 16nm MLC and validate the new architecture. Micron's 16nm MLC is cheap as chips, and the ability to fab its own NAND provides the company with a pricing crowbar to dislodge other non-fab competitors. It will be interesting to see how this affects Memblaze's sales, but the company is likely in lucrative territory with the Micron partnership. We haven't had the opportunity to compare 15nm Toshiba and 16nm Micron NAND head-to-head, and though we include the Memblaze in the test pool, it isn't a direct comparison due to the incredible amount of extra overprovisioning on the Micron 9100.

The 9100 Max and Pro feature a different amount of spare area, as noted in the raw capacity column in the table. The Max sets aside 40 percent of its capacity for spare area, while the Pro allocates 20 percent. Micron dedicated the lion's share of the spare capacity to overprovisioning, which increases sustained random write performance and endurance. Micron utilizes part of the spare area its RAIN (Redundant Array of Independent NAND) feature, which weaves an extra parity bit into all of the data stored on the drive, which in turn provides recoverability in the event of errors or component failures.

The extra overprovisioning allows the 9100 Max models to provide a higher level of sustained performance and endurance, but it does come at the cost of addressable capacity. The 9100's DWPD (Drive Writes Per Day) endurance metric is close to its competitors, but this is mainly due to how it is calculated. Micron's three-year warranty does not match the typical five-year warranty that most vendors offer, which inflates its DWPD measurement. Both the Max and the Pro have relatively low endurance thresholds in comparison to some competing models, which we dive into on the next page.

The 9100 series features Micron's eXtended Performance and Enhanced Reliability Technology (XPERT), which is a modular suite of techniques that offers enhanced defect and error detection and correction. The suite includes a number of optimizations that protect user data and boost performance, such as power loss protection, data path protection, adaptive read and thermal protection. We have a deeper explanation of the XPERT features here.

Frankly, we expected Micron's newest enterprise SSD to come swinging the 3D NAND hammer. The joint Micron/Intel NAND venture is already pumping out 3D NAND, and Intel has already announced its newest 3D NAND-powered enterprise SSDs. However, Intel is bolting its 3D NAND onto its existing DC P3700 architecture, which eliminates the capacity benefits of 3D NAND and limits the SSDs to 2TB. It appears Micron held out to leverage a new architecture with its 3D NAND, and the 9100 just happens to support up to 4TB of NAND at this point and the Microsemi controller can accommodate up to 8TB.

Micron stated that the overall performance and cost of the solution are more important than the underlying media, and that does bear some merit. Combining the right blend of price, performance and endurance is the perfect trifecta, and from the specs, it appears Micron may have the performance well under control. Let's take a closer look and see if those claims pan out.

  • Flying-Q
    If the flash packages are producing so much heat that they need such a massive heatsink, why is there only one? Surely the flash on the rear of the card would need heatsinking too, even just a flat plate would suffice?
    Reply
  • Unolocogringo
    It appears the heatsink is more for the voltage converters and controller chip to me.
    Reply
  • PaulAlcorn
    18234127 said:
    If the flash packages are producing so much heat that they need such a massive heatsink, why is there only one? Surely the flash on the rear of the card would need heatsinking too, even just a flat plate would suffice?

    This is a standard configuration, though there are a few SSDs that have rear plates. Thermal pads were more common with larger lithography NAND, 20nm, 25nm, etc, because it generated more heat. New smaller NAND, such as the 16nm here, draws less power and generates less heat. In fact, it was very common with old client SSDs to have a thermal pad on the NAND, whereas now they are relatively rare. I think that they may be relying upon reducing the heat enough on one side to help wick heat from the other side, but the heatsink is primarily for the controller and DRAM with the latest SSDs. Also, it may just be convenient to add additional thermal pads to the NAND to keep the spacing for the HS even across the board.
    Reply