Toshiba XG6 M.2 NVMe SSD Review: Higher Density and Improved Efficiency

SSDs are steadily displacing HDDs in the consumer market, but progress is slow in some areas. Yes, more notebooks are coming with SSDs by default, but adoption in OEM desktop PCs has been slower. Today we take a close look at Toshiba’s latest OEM-oriented SSD. Toshiba designed the XG6 for desktop PCs, mobile systems, embedded systems, and even data centers.

That means the XG6 could power your next laptop or pre-built PC. More importantly for us, the XG6 comes with Toshiba's new 96-Layer TLC BiCS flash that will soon power the newest enthusiast-class SSDs from multiple vendors. The new flash promises more performance and efficiency at a lower price point, and based on our testing, it delivers.

As mentioned, the Toshiba XG6 is the industry’s first SSD with Toshiba’s latest 96-Layer 3D TLC flash. Toshiba’s new 512Gbit die provides a 40% capacity increase over the company's previous-gen 64-layer flash. The increased density results in lower manufacturing costs, which in turn allows Toshiba to be more aggressive with its SSD pricing.

Toshiba's new flash supports Toggle 3.0 speeds (667-800 MT/s), which means its program and read latency is lower than the company's previous-gen Toggle 2.0 flash (400 MT/s). Toshiba's new flash is not quite as fast as Samsung’s latest 256Gbit 3D flash that supports Toggle 4.0, but Toshiba's increased capacity is the big attraction. 

We normally expect new flash to be paired with a new SSD controller, but the XG6 uses the same controller as the XG5. Toshiba's firmware and flash improvements, update to the NVMe 1.3a spec, and focus on improved power efficiency have yielded tremendous improvements, though.

Toshiba’s XG6 uses an enhanced SLC write buffer to provide higher sequential write speeds than the XG5 and up to twice the performance in random workloads. That's a surprisingly big improvement considering the SSD uses the same controller, but it's clear the advantages of the much faster 96-layer flash are shining through.

Specifications

ProductToshiba XG6 256GBToshiba XG6 512GBToshiba XG6 1TB
PricingN/AN/AN/A
Capacity
(User / Raw)
256GB / 256GB512GB / 512GB1024GB / 1024GB
Form FactorM.2 2280 S2 (Single-sided)M.2 2280 S2 (Single-sided)M.2 2280 S2 (Single-sided)
Interface / ProtocolPCIe 3.1 x4 / NVMe 1.3aPCIe 3.1 x4 / NVMe 1.3aPCIe 3.1 x4 / NVMe 1.3a
ControllerTC58NCP090GSDTC58NCP090GSDTC58NCP090GSD
DRAMNANYA LPDDR3NANYA LPDDR3NANYA LPDDR3
MemoryToshiba BiCS FLASH 96-layer 3D TLCToshiba BiCS FLASH 96-layer 3D TLCToshiba BiCS FLASH 96-layer 3D TLC
Sequential ReadUp to 3,180 MB/sUp to 3,180 MB/sUp to 3,180 MB/s
Sequential WriteUp to 2,960 MB/sUp to 2,960 MB/sUp to 2,960 MB/s
Random ReadUp to 355,000 IOPSUp to 355,000 IOPSUp to 355,000 IOPS
Random WriteUp to 365,000 IOPSUp to 365,000 IOPSUp to 365,000 IOPS
EncryptionTCG Pyrite and OPAL 2.01 as an option (TCG Pyrite: x = 0, TCG OPAL: x = A)
EnduranceN/AN/AN/A
Part NumberKXG6xZNV256GKXG6xZNV512GKXG6xZNV1T02
Warranty5-Years5-Years5-Years

The Toshiba XG6 comes in three capacities of 256GB, 512GB, and 1TB. The drive offers up to 3.18/2.96 GB/s of sequential read/write throughput and up to 355,000/365,000 random read/write IOPS. Toshiba measures performance when the data lands in the SLC cache, a fast buffer that absorbs incoming data, which means these measurements are peak values. We recorded ~1.6 GB/s of throughput when data spilled over to the native TLC flash. That's actually very competitive compared to other SSDs.

Power consumption is one of the most important considerations for OEM devices. Toshiba's new flash features a 1.2V I/O rating. That's 33% lower than the previous gen’s 1.8V, which equates to big power savings. The XG5 is rated for 4.2W of active power consumption and falls to just 3mW in its lowest power state (L1.2). We'll put those claims to the test. 

The XG5 is an OEM-oriented product, so pricing varies based on the customer. Toshiba also offers models that support either TCG Pyrite or TCG OPAL 2.01 encryption. Toshiba backs the XG6 with a five-year warranty, but OEMs have wide latitude in how they pass that warranty down to their customers. Toshiba didn't provide us with an endurance rating because the OEM also dictates it.

A Closer Look

The XG6 comes in an M.2 2280 S2 single-sided form factor with the Toshiba TC58NCP090GSD SSD controller. This controller communicates with the host over a PCIe 3.1a x4 interface with the NVMe 1.3a protocol.

The 1TB XG6 has a 512MB Nanya DRAM buffer. Most SSDs have a 1MB-to-1GB DRAM-to-NAND ratio, but recent SSD designs appear to signal a move to lower ratios. For instance, the 1TB Intel 660p only comes with 256MB of DRAM.

The XG5 packs 1TB of storage into just two flash packages. That equates to 953GB of usable space after you format the drive in Windows. Laptop makers are intensely focused on reducing weight, and the XG6's 7.3 grams is much more palatable than the typical 40-70 grams for a SATA SSD.

MORE: Best SSDs

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Create a new thread in the Reviews comments forum about this subject
This thread is closed for comments
16 comments
Comment from the forums
    Your comment
  • AgentLozen
    Seems pretty solid. Thanks for the review.
  • philipemaciel
    Cons: "seperately"
  • rantoc
    Seems like a solid laptop ssd
  • Brian_R170
    Why bother to review this model if consumers can't even buy it except in a pre-built system? How many people make their system purchase decisions based on the SSD inside? Outside of boutique stuff, how many OEMs actually list the SSD model in their specifications?

    Plus, any future SSD design based off of this model won't likely have identical specs, so you'd have to retest it anyway.
  • stdragon
    I've had three OEM Toshiba NVMe drives fail in less than a year in a Dell.
  • 2Be_or_Not2Be
    Why is the Intel Optane 905P included on some benchmarks and not on others? It would be nice to have consistency throughout the review. Either have it in all of the benchmarks, or just don't include it at all.

    I understand you may be including results from a previous test on the 905P that may not have had all of the same tests as this XG6. However, it's frustrating to me when I see it one benchmark (I do like that it's included), but then when I want to compare against it on another one, it's not there.

    So I would say if you don't have it for all of the tests, just exclude it from all of them.
  • PaulAlcorn
    Anonymous said:
    Cons: "seperately"


    Good eye, fixed!
  • PaulAlcorn
    Anonymous said:
    Why is the Intel Optane 905P included on some benchmarks and not on others? It would be nice to have consistency throughout the review. Either have it in all of the benchmarks, or just don't include it at all.

    I understand you may be including results from a previous test on the 905P that may not have had all of the same tests as this XG6. However, it's frustrating to me when I see it one benchmark (I do like that it's included), but then when I want to compare against it on another one, it's not there.

    So I would say if you don't have it for all of the tests, just exclude it from all of them.


    Thanks for the feedback, it is appreciated. Over the last few months we've changed our test regimen entirely to more accurately reflect real-world use cases, much of this due to reader feedback. We are still in a state of flux somewhat with the current line up of tests as we optimize the suite to focus on the things that matter most, and avoid things that don't. Reader feedback has a big impact on our decisions. In either case, the adjustments have led to a few devices not being included until we can complete backfill testing, but we'll work to rectify that.
  • jpe1701
    Where is Chris?
  • PaulAlcorn
    Anonymous said:
    Where is Chris?


    Chris decided to pursue another opportunity, which is unfortunate but part of the business at times. Sean, who has years of experience testing and reviewing storage, has stepped into the role now.
  • uglyduckling81
    "Optane's Earth Shattering Performance" - does it though?
    It has low latency, sure. But it's transfer rates are that great are they?
    Basically everything I've read about it is it's pretty good but not as good as the Sammy Pro.
  • 2Be_or_Not2Be
    Anonymous said:
    "Optane's Earth Shattering Performance" - does it though?
    It has low latency, sure. But it's transfer rates are that great are they?
    Basically everything I've read about it is it's pretty good but not as good as the Sammy Pro.


    The part where Optane drives shine is in QD1 (queue depth = 1; that is, only one transaction to complete). It hits top performance transfer-wise and remains consistent. If you think about it, all data transactions start off with the first one; so the faster you fulfill the first one, the faster you can fulfill the next one. Thus, you generally are fulfilling data requests as fast as they come in & keeping the queue low. The performance level remains high, and usually doesn't hit high queue depths (not including servers, which can have significant & varied loads; of course, and perhaps even more so, Optane drives will shine with servers).

    SSDs generally rely on pseudo-SLC caching (small portion of overall storage capacity) to hit peak performance; however, that benefit is gone if the cache is filled up. Also, the max transfer rates of those SSDs are often only seen when the QD is in the QD32 range. So Optane drives should generally feel "snappier" than a regular SSD, and if you need more total transfer bandwith than one individual drive delivers, then you just add more Optane drives (like you would with almost any other storage solution).

    Now, for most consumers, I would agree that they could be served just as well with a Samsung SSD as an Optane drive as generally consumers aren't generating much of a storage I/O load. But if you had to have the best, then Optane is the best right now. Personally, I am just waiting for these drives to come down in price (especially price!) and add more capacity. I can justify the price for servers, but it's harder to do so for a personal system.
  • djay.alpha
    Hi Sean,
    Are you planning to move your test platform to Windows 10 anytime soon? Would be interesting to see the difference between Redstone 4(meltdown patch included) vs Redstone 3 vs Windows 8.1.
  • seanwebster
    Anonymous said:
    Hi Sean,
    Are you planning to move your test platform to Windows 10 anytime soon? Would be interesting to see the difference between Redstone 4(meltdown patch included) vs Redstone 3 vs Windows 8.1.

    Hi. All benchmarks reflect performance with Win 10 Pro RS4 and spectre and meltdown patches enabled. The motherboard firmware is updated for it as well. From my testing, I've seen a reduction of about 10-15% QD1 4k read performance and about 40% reduction in 4K write performance with the updates in place.
  • WyomingKnott
    "Toshiba measures performance when the data lands in the SLC cache...."
    In other words, they publish specs for that small window where it runs like an actually good SSD and hope very few people notice that they are lying through their teeth.

    I dislike ad-speak. Had to teach my daughter that 12 is not "3 times more" than 4. Which was the correct answer in math class, for the sake of the deity of your choice.
  • jpe1701
    @PAULALCORN Well I wish him the best, I've enjoyed his articles. Looking forward to seeing the new guys take on things.