Micron is taking the wraps off their latest data center SSD offering today. The 9400 NVMe Series builds upon Micron's success with their third-generation 9300 series introduced back in Q2 2019. The 9300 series had adopted the U.2 form-factor with a PCIe 3.0 x4 interface, and utilized their 64L 3D TLC NAND. With a maximum capacity of 15.36 TB, the drive matched the highest-capacity HDDs on the storage amount front at that time (obviously with much higher performance numbers). In the past couple of years, the data center has moved towards PCIe 4.0 and U.3 in a bid to keep up with performance requirements and unify NVMe, SAS, and SATA support. Keeping these in mind, Micron is releasing the 9400 NVMe series of U.3 SSDs with a PCIe 4.0 x4 interface using their now-mature 176L 3D TLC NAND. Increased capacity per die is also now enabling Micron to present 2.5" U.3 drives with capacities up to 30.72 TB, effectively doubling capacity per rack over the previous generation.

Similar to the 9300 NVMe series, the 9400 NVMe series is also optimized for data-intensive workloads and comes in two versions - the 9400 PRO and 9400 MAX. The Micron 9400 PRO is optimized for read-intensive workloads (1 DWPD), while the Micron 9400 MAX is meant for mixed use (3 DWPD). The maximum capacity points are 30.72 TB and 25.60 TB respectively. The specifications of the two drive families are summarized in the table below.

Micron 9400 NVMe Enterprise SSDs
  9400 PRO 9400 MAX
Form Factor U.3 2.5" 15mm
Interface PCIe 4.0 NVMe 1.4
Capacities 7.68TB
15.36TB
30.72
6.4TB
12.8TB
25.6TB
NAND Micron 176L 3D TLC
Sequential Read 7000 MBps
Sequential Write 7000 MBps
Random Read (4 KB) 1.6M IOPS (7.68TB and 15.36TB)
1.5M IOPS (30.72TB)
1.6M IOPS (6.4TB and 12.8TB)
1.5M IOPS (25.6TB)
Random Write (4 KB) 300K IOPS 600K IOPS (6.4TB and 12.8TB)
550K IOPS (25.6TB)
Power Operating 14-21 W (7.68TB)
16-25W (15.36TB)
17-25W (30.72TB)
14-21 W (6.40TB)
16-24W (12.8TB)
17-25W (25.6TB)
Idle ? W ? W
Write Endurance 1 DWPD 3 DWPD
Warranty 5 years

The 9400 NVMe SSD series is already in volume production for AI / ML and other HPC workloads. The move to a faster interface, as well as higher-performance NAND enables a 77% improvement in random IOPS per watt over the previous generation. Micron is also claiming better all round performance across a variety of workloads compared to enterprise SSDs from competitors.

The Micron 9400 PRO goes against the Solidigm D7-5520, Samsung PM1733, and the Kioxia CM6-R. The Solidigm D7-5520 is handicapped by lower capacity points (due to its use of 144L TLC), resulting in lower performance against the 9400 PRO in all but the sequential reads numbers. The Samsung PM1733 also tops out at 15.36TB with performance numbers similar to that of the Solidigm model. The Kioxia CM6-R is the only other U.3 SSD with capacities up to 30.72TB. However, its performance numbers across all corners lags well behind the 9400 PRO's.

The Micron 9400 MAX has competition from the Solidigm D7-P5620, Samsung PM1735, and the Kioxia CM6-V. Except for sequential reads, the Solidigm D7-P5620 lags the 9400 MAX in performance as well as capacity points. The PM1735 is only available in an HHHL AIC form-factor and uses PCIe 4.0 x8 interface. So, despite its 8 GBps sequential read performance, it can't be deployed in a manner similar to that of the 9400 MAX. The Kioxia CM6-V tops out at 12.8TB and has lower performance numbers compared to the 9400 MAX.

Despite not being the first to launch 32TB-class SSDs into the data center market, Micron has ensured that their eventual offering provides top-tier performance across a variety of workloads compared to the competition. We hope to present some hands-on performance numbers for the SSD in the coming weeks.

Comments Locked

18 Comments

View All Comments

  • jordanclock - Monday, January 16, 2023 - link

    If you finished reading the rather short article you would know the drives are U.3, not U.2. Also, PCIe4x4 has more bandwidth than the peak read or write speeds of the drives, so I don't see why they would use a higher bandwidth PCIe link. Also, as mentioned in the article, these aren't even the first 30TB drives like this. There is absolutely demand for a drive like this. If 7GBps isn't enough, then you use multiple drives in an appropriate distributed system, like RAID or zpool.
  • Dug - Friday, January 20, 2023 - link

    You don't seem to comprehend how these would be connected, setup, or configured do you? Or even the performance in a zfs array on the correct controller.
  • Silver5urfer - Thursday, January 12, 2023 - link

    It is a DAMN SHAME that Intel beancounters nuked Optane look at P5800X and checkout it's endurance and it's top notch performance it will never choke like the garbage NVMe we have.

    We consumers get nothing like this tech, look at the capacity. Except that Firecuda NVMe from Seagate which uses top end Micron 176L NAND flash none of the SSDs have good endurance. Samsung dropped ball big time after they killed MLC technology and never cared enough since their brand is earning them money.

    Shame really and we got milked for 2 years straight on PCIe4.0 NVMe SSDs with extreme price and yet trash capacity and insanely poor Endurance rating. Still we do not have any 8TB mainstream high end SSDs. There's a SATA QLC junk from Samsung, check the TLC 870 Evo 4TB it has like 85% of free space man, Samsung can make it 16TB drive without issues and with cool running SATA it would be a boon for many, but they do not want as I think in general people stopped caring about capacities and doing basic research and instead go with the stupid PR and get hooked by numbers and benchmarks.

    PCIe 5.0 NVMe rip off incoming and same 2TB capacity and 4TB top which will be uber expensive because reasons, oh yea theres no viable difference with NVMe vs SATA in general purpose daily compute unless you are running some 4K REMUX on mkvtoolnix on a Threadripper like machine, Optane is still better but dead now.

    Well, fingers crossed for price cuts on Firecuda 4TB for my machine what else can I expect at this point, blowing cash on PCIe5.0 is worthless.

    Also on a final note, the U.3 connector is very interesting once I searched about it, 100% compat with NVMe, SATA and U.2 I wish OEMs offer these as standard and get splitters for SATA. Apart from EVGA no one offers U.2 connectors shame. Enterprise stuff is very good but normies do not care as they only care about MUH Gaming and PR like Direct X Storage gonna change game lol.
  • PeachNCream - Friday, January 13, 2023 - link

    The funny thing is what turned low endurance junk TLC into premium-tier, high-endurance storage was the introduction of even lower endurance QLC. Now, most of the average, uninformed yet self-proclaimed tech savvy that were decrying TLC at introduction because of the decline in endurance from MLC are now defending TLC while using QLC as a point of comparison.

    Granted, TLC is ... marginally acceptable if you're mindful of the endurance limitations and treat it kindly, but I would prefer being able to thrash my storage and not really care. MLC offered that for the most part and struck a good balance between itself and SLC. At any rate if you want capacity, you're going to be stuck with less endurance for the time being until someone finds a cost effective alternative to NAND flash storage. Given how quickly Optane died, I don't think an alternative is coming in the near term.
  • FunBunny2 - Saturday, January 14, 2023 - link

    "At any rate if you want capacity, you're going to be stuck with less endurance for the time being until someone finds a cost effective alternative to NAND flash storage. Given how quickly Optane died, I don't think an alternative is coming in the near term."

    but, but, but... weren't we promised that 3D NAND (TLC or QLC or xLC) would have even better endurance than MLC (may be even SLC?) just because it's fabbed on Really Big Olde Nodes??? weren't we?
  • PeachNCream - Sunday, January 15, 2023 - link

    That is true and I shudder to think what TLC or QLC endurance would look like in a planar NAND cell. It helped, but it clearly didn't totally resolve the problem and I believe modern 3D NAND ended up landing on smaller process nodes anyway in pursuit of lower cost for a given capacity.
  • Silver5urfer - Tuesday, January 17, 2023 - link

    The only reason I accepted TLC is for a single drive, Firecuda because it is very high endurance for a consumer. Samsung 860 Pro was MLC but that was like 5 year old technology I think it might be on 22nm, I do not know rough guess and lazy to search about it. But the newer Firecuda 4TB one NVMe beats it in endurance with TLC.

    Except that instance, none of the TLC drives are replacing MLC, but here if we imagine Samsung making 870 Pro with modern node MLC I bet it would destroy it.

    I do not really expect any Optane successor. Enterprise get the drive like above, meanwhile idiot consumers keep on buying the latest sticker that will give them more E-Points than anything, that is PCIe5 woah and anything that has shiny useless benchmarks and marketing.

    Look at HDD space, still 20TB is uber expensive. $400 that was the discounted price for WD Gold with OptiNAND that I got and that price never stuck even during BF sales. And with HDD shipments slashed by 40% for all the major companies, I do not expect any magic even there like ePMR, Optinand, other types of EAMR technologies for masses. In fact WD removes Helium for all drives below 12TB.
  • FunBunny2 - Monday, January 23, 2023 - link

    "Enterprise get the drive like above"

    the thing about bidnezz and equipment - Uncle Sugar, which is to say you the American Taxpayer, pays them to replace stuff. so they're happy to get the warranty period years out of SSD, and replace at expiry date. they get to deduct the whole cost of each drive. all they care about is that a drive model doesn't die before warranty expires 99.9% of the time.

    99.99% of any SSD will keep going pretty much forever in usual consumer use. I just ordered a Samsung 512 gig drive (damn those 500 gig pieces of shit) so I can clone my MX100, which is a many drive descendent of the first Ubuntu HDD, circa kernel 2.2 (or thereabouts), which has a now way too tiny /boot partition and I need to expand it; potentially destructive process. long since forgotten what was in the MX100. looked it up. turns out MLC. also turns out I've had the drive since 2014! time flies when your having fun.

Log in

Don't have an account? Sign up now