What SSDs besides Samsung 840 use planar TLC NAND?

[cbn]

Some more information to fill out the list with.

Samsung 750 EVO Samsung MGX 16nm TLC NAND

Samsung PM851 Samsung MDX 19nm TLC NAND

Silicon Power S55 32/60GB MLC 120/240/480/960GB TLC

SanDisk Ultra II 120/240GB Marvell 88SS9190 480/960GB Marvell 88SS9189 19nm TLC NAND

ADATA SP550 SM2256 SK Hynix 16nm TLC NAND (heard previously that they planned to switch over to using Micron 16nm 128Gbit TLC NAND instead but don't know for certain that they will or perhaps already have)

And you haven't filled out the NAND for Trion 100 (19nm) or Trion 150 (15nm).

Thanks. I updated the OP.

P.S. The June 2015 press release for the ADATA SP550 mentioned Micron TLC, but none of the reviews listed on Johnny Lucky's website reporting it being used.

 

[Glaring_Mistake]

Initially they were going to use Micron TLC but for some reason (don't remember why but likely Micron's TLC wasn't ready) and therefore went with NAND from SK Hynix.
Heard about two months ago that they planned to switch over to Micron TLC, don't know if that has been long enough for them to do it or not or if they perhaps decided not to do so.

 

[cbn]

Initially they were going to use Micron TLC but for some reason (don't remember why but likely Micron's TLC wasn't ready) and therefore went with NAND from SK Hynix.
Heard about two months ago that they planned to switch over to Micron TLC, don't know if that has been long enough for them to do it or not or if they perhaps decided not to do so.

That is interesting and if they did change I wonder how it would compare to the BX200?

In the Anandtech ADATA SP550 review the reviewer mentioned the SM2256 having updated firmware (compared to BX200):

http://www.anandtech.com/show/10131/the-adata-premier-sp550-ssd-review

In November we looked at the performance of the Crucial BX200 based on the SM2256 controller and Micron 16nm TLC NAND, and we found that the situation had gotten much worse: performance was bad and power consumption was high. In the meantime, Phison's S10 controller (and Toshiba's custom branded variant of it) became the most popular solution for TLC SSDs and the market was flooded with nearly-identical drives from new and old players in the market.

That brings us to today's review of ADATA's Premier SP550 SSD. ADATA manufactures drives with almost every controller available on the open market, and the Premier SP550 is their SM2256 offering. ADATA has in turn positioned the drive as the entry-level model of their mainstream consumer Premier product line. Meanwhile with a different choice of flash (SK Hynix 16nm TLC NAND) and updated firmware, the SP550 has the chance to partially redeem the SM2256 controller.

And being a better performer than the BX200:

http://www.anandtech.com/show/10131/the-adata-premier-sp550-ssd-review/10

To that end the ADATA Premier SP550 takes shape as a properly working implementation of an SM2256-based SSD, without the ugly surprises of the Crucial BX200. Write speeds - both sequential and random - are an acute weakness, though this admittedly is nothing new for a low-end TLC drives and at this point is becoming a known trade-off to reach lower prices. Unfortunately this also means that it still lags behind the prototype that used Samsung NAND and that it rarely pulls ahead of the pack of retail competitors. Meanwhile latency is consistently worse than drives based on other controller architectures that have multi-core processors.

Or perhaps another way of looking at the situation would be how the BX200 performs with updated firmware?

NAND type (SK Hynix 16nm TLC vs Micron 16nm TLC) vs. firmware, I wonder how much each is factoring in?

 

[Glaring_Mistake]

Going by the tests on reduced read speeds that I have performed on for example BX200 and SP550 I would lean towards the NAND being the differentiator here.

When I look at my results for them I can see that the read speed for the BX200 has for some small areas dropped by around 200-300MB/s while for the SP550 read speeds have only dropped by about 50MB/s.

It may not be proof that the NAND itself is the reason the BX200 is slower than the SP550 but I do think it is an indication that it's some really poor NAND they're using in the BX200.

 

[cbn]

Updated the opening post with Samsung CM871a (which uses the new Maia controller rather than the MGX controller found in the 750 EVO and 850 EVO):

http://www.anandtech.com/show/10181/samsung-readies-new-maia-ssd-controller

The new Samsung Maia controller supports multiple types of NAND flash memory, including TLC and 3D NAND/V-NAND. At the Samsung SSD Forum Japan, the company demonstrated two new SSDs based on the Maia chip. Both drives come in 2.5” form-factor and with a SATA III (6Gbps) interface. Initially Samsung intends to use the controller to build drives for OEMs, but as is often the case with Samsung, it is quite possible that it will eventually show up in retail SSDs as well.

The new Samsung CM871a OEM SSD based on the Maia controller will feature Samsung’s 16nm TLC NAND, which indicates that the company has no plans to drop production of planar NAND flash memory for consumer SSDs any time soon. The drive is rated for maximum sequential read speed of up to 535 MB/s as well as sequential write speed of up to 215 MB/s in default mode or up to 515 MB/s with TurboWrite technology active. Like many modern TLC NAND-based SSDs, the CM871a uses part of the non-volatile memory in SLC mode for rapid caching. The SSDs can perform up to 97K random read IOPS as well as up to 59K random write IOPS (or 88K IOPS with TurboWrite technology). Samsung will offer its CM871a in 128 GB and 256 GB configurations, which indicates that the family was designed for the low end of the market.

Meanwhile in the retail channel Samsung sells the inexpensive 850 EVO and 750 EVO SSDs. These SSDs are based on the company’s MGX controller, coupled with the company's second-gen 32-layer TLC V-NAND or 16 nm planar TLC, respectively. The aforementioned drives offer performance levels very similar to those of the forthcoming CM871a and PM871a, which may indicate that Samsung’s new Maia platform focuses mostly on cost efficiency, endurance and predictable performance, but not on maximum sequential read or random IOPS numbers (especially given limitations of SATA). Nonetheless, the development of Maia shows that SATA is still an important part of the market for Samsung as lower-end PCs continue to rely on it.

 

[cbn]

That's... horrid! Especially when you consider the data-retention issues with at-rest data decaying, and using up precious P/E cycles re-writing the data.

Using the example from post #15, 2Xnm planar TLC has 150% the capacity of 2Xnm planar MLC (re: 3 bits vs. 2 bits), but only 25% the endurance.

Or another way of looking at things would be that MLC has 67% of the capacity of TLC, but 4X the endurance

So yes, a drastic change in endurance for a bit more capacity when going from 2Xnm planar MLC to 2Xnm planar TLC.

To combat this massive change in endurance the SSD controller maker adds additional error checking (that according to the link below discussing the SM2256 controller kicks in when the drive begins to age).

http://www.anandtech.com/show/9165/...-reference-design-ssd-review-tlc-for-everyone

Architecturally the SM2256 shares the same core design as its predecessor SM2246EN. The design is modular, which allows Silicon Motion to change parts of the controller without redoing the rest. It features the same single 32-bit Argonaut RISC processor core as the SM2246EN, which is quite unique because we have seen many SSD controller vendors moving towards multi-core ARM architectures. A single custom core obviously brings efficiency gains and we've witnessed those in the SM2246EN, but the downside of such limited CPU power is sustained performance when the controller has to perform garbage collection at the same time as processing host IOs.

The only dramatic change is in the error correction circuitry as the SM2256 supports Low Density Parity Check (LDPC) error correction codes instead of more common and less powerful BCH ECC. Silicon Motion calls its ECC technology as NANDXtend, and it's a combination of LDPC hard and soft decode along with RAID5-like data recovery. The benefit of having three levels of ECC is performance because LDPC soft decode and recovery from parity both have a relatively noticeable impact on performance and are typically only needed when the drive approaches its end of life (i.e. when the NAND has been cycled a lot). Uncycled NAND has much higher reliability because the tunnel oxide hasn't worn out due to P/E cycles, so only very little ECC is needed and LDPC hard decode is sufficient and also doesn't have a dramatic impact on performance.

The reason why hard decode is faster than soft decode lies in how the voltage of a cell is sensed. Hard sensing is binary based, so for an SLC cell like in the graph above the cell can be either 1 or 0. However, as you can see, the voltage threshold distributions overlap slightly and that's actually far worse with MLC and TLC since there are more voltage states. In soft sensing the voltage distributions are divided into several segments, which requires more precision and iterations. For example in segment 4 the bit value can be either 1 or 0 as the distributions overlap, so probability algorithms are used to figure out the correct value. To be honest, ECC codes and the way they work are way over my head, but in case you are familiar with ECC and want to learn more, I suggest you simply google LDPC as there are numerous publicly available academic papers that go into more depth about this topic.

Silicon Motion claims that its NANDXtend technology can extend the endurance of TLC NAND by up to three times, making TLC more robust for heavier workloads and also allowing the use of lower quality NAND that some OEMs may use anyway due to the lack of in-house binning equipment. Unfortunately I didn't have any time to do extended endurance testing with the SM2256 yet to validate Silicon Motion's claims, but I will be sure to test that once we have a retail drive on our hands,

With that above article noted, I do wonder how the speed of the drive will be affected as it wears over time (re: the above mentions LPDC soft decode is slower than the hard decode that is used when the drive is newer). However, I would think with a larger drive (used in a consumer PC) the use of LPDC soft decode shouldn't happen as soon as it would with a smaller drive due to the effect of wear leveling. This assuming the large capacity planar TLC SSD and small capacity planar TLC SSD are being used in the same fashion.

P.S. Here is the marketing graph from post #3 (taken from the Tom's SM2256 article):

So if this is actually true, then a SM2256 drive with its planar TLC NAND would be fastest the first ~1/3 of its life (ie, the first 500 P/E cycles) and then slow down the last ~2/3 (500 to 1500 P/E cycles) of its life as the LPDC slow decode and RAID5-like data recovery kick in.

Interestingly 1500 P/E cycles is very close to the 2000 P/E cycles quoted in post #15 for the Samsung TLC V-NAND.

 

[Glaring_Mistake]

The ADATA SP550 using SK Hynix NAND has 1000 write cycles.
Now if they are already taking into account that NANDXtend can increase write cycles by up to three times that would mean it would otherwise be rated at about 300 write cycles.


Don't know what the NAND in BX200 is rated at, though not for lack of trying.


For the Toshiba/SanDisk TLC NAND I have seen several different claims of it's write endurance.

The first SSD using their 19nm TLC NAND was the SanDisk Ultra II and it was rated at about 500 write cycles.
The controller used in it was not made for TLC NAND however so that might have had an effect.

After that I have heard claims of 700, 1200, 1500 and 2000.

The highest claim(s) of 2000 was for the Lite-On CV2 (19nm) and Plextor M7V (15nm) which was said to be an increase by 33% by using LDPC meaning that with BCH ECC they would be rated at 1500.

 

[Glaring_Mistake]

Minor update: SanDisk X400 is supposed to use Marvell 88SS1074-BSW2 as controller and 15nm TLC NAND.

 

[Maxima1]

Why the heck did SATA Express go away? It just seems like an excuse to segment the market with low performance SSDs to differentiate them from the high performance ones. When is NVMe going to go down to a reasonable level?

 

[cbn]

Minor update: SanDisk X400 is supposed to use Marvell 88SS1074-BSW2 as controller and 15nm TLC NAND.

Thanks, I updated the OP.

 

[cbn]

Opening Post updated with AMD R3 SSD which uses the SM2256 and SK Hynix TLC NAND (same controller and NAND combination as the Intel 540s and ADATA Premier SP 550)

 

[cbn]

Updated Opening post with Kingston UV400 (Thanks to info gained from Johnny Lucky's database --> http://www.johnnylucky.org/data-storage/ssd-database.html )

 

[BonzaiDuck]

I hope so, it would be awesome for all kinds of lower end computers out there including those with 2GB RAM (Eg, Core 2, with only two DIMM slots and not using the more expensive higher capacity 2GB DDR2 DIMMs).

P.S. With regard to frequent paging out and disk swapping (in a situation of limited RAM), I did find this marketing graph (from the Tom's Silicon Motion SM2256 article ) interesting.

In a nutshell, the idea is the SM2256 controller (with DRAM cache) is supposed to extend the life of the NAND. (Will be interesting to see how this pans out in the Wild, Long life or drives dying sooner than expected?).

I estimate that just with the hibernation file and swapfile alone, I might have about 30GB of writes per day if the system hibernates more than three times in 24 hours. Anything beyond those two causal factors is fairly miniscule.

I could eliminate the hiberfil.sys, but it's important to let my systems go to sleep, and I'd rather have them hibernate after a few hours asleep than let them simply maintain a sleep-state for longer than that. A system that sleeps for long periods of time can wear out the PSU, or so I'd been advised. I was told that certain components of a PSU heat up to maintain the sleep-state, and there's no airflow to compensate.

 

[Glaring_Mistake]

Looks like Mushkin Triactor uses 15nm TLC NAND.

 

[cbn]

Looks like Mushkin Triactor uses 15nm TLC NAND.

Very Interesting as it uses Sandisk NAND with the SM2256 controller.

http://www.nikktech.com/main/articl...drives/6398-mushkin-triactor-480gb-ssd-review

So that is a departure from the usual combination of SM2256 with either SK Hynix TLC or Micron TLC NAND.

 

[Glaring_Mistake]

Lite-On MU-II (controller Phison S10 15nm NAND) and Transcend SSD220 (controller SM2256 unknown NAND) should be added to the list.

 

[Glaring_Mistake]

Very Interesting and it uses Sandisk NAND with the SM2256 controller.

http://www.nikktech.com/main/articl...drives/6398-mushkin-triactor-480gb-ssd-review

So that is a departure from the usual combination of SM2256 with either SK Hynix TLC or Micron TLC NAND.

Yeah, not a combination I've seen before.

 

[cbn]

Updated list with Zotac T400 (Phison S11 and 15nm Toshiba TLC NAND)

P.S. Tweaktown Preview of the Phison S11 (with Toshiba 3D TLC and Micron 3D MLC) found in this post. (Still waiting to see review of Phison S11 with planar TLC NAND)

 

[VirtualLarry]

Interesting about the Phison S11 turning in such high 4K QD32 read scores. I ordered two 120GB Zotac T400 SSDs to benchmark.

 

[cbn]

Plextor S2C, SM2258 and SK Hynix 16nm TLC:

http://www.anandtech.com/show/10735/plextor-launches-s2c-ssds-16-nm-sk-hynix-tlc-nand-and-smi-sm2258

 

[BFG10K]

I'd never buy a TLC drive, not when the likes of Crucial have MLC drives for comparable prices.

 

[Elixer]

I'd never buy a TLC drive, not when the likes of Crucial have MLC drives for comparable prices.

Problem is, they are all going for TLC. The new Crucial mx300 is (3D) TLC.
They don't make the mx100 & mx200 anymore which were MLC.

 

[cbn]

Problem is, they are all going for TLC. The new Crucial mx300 is (3D) TLC.
They don't make the mx100 & mx200 anymore which were MLC.

Something to look out for is Samsung and SK Hynix 14nm MLC.

 

[Glaring_Mistake]

Something to look out for is Samsung and SK Hynix 14nm MLC.

Plextor has actually recently released the S3-series which uses SK Hynix 14nm TLC NAND combined with SM2254 as a controller.
Was not sure that SK Hynix would go down to 14nm since their 3D NAND should be in its fourth generation by now but I guess they may still have seen some profit in shrinking the lithography a bit more.
Going by the TBW though it seems that endurance may have suffered since it is down to 35/70TBW from the 75/150TBW that the S2-series had.

 

[cbn]

Plextor has actually recently released the S3-series which uses SK Hynix 14nm TLC NAND combined with SM2254 as a controller.
Was not sure that SK Hynix would go down to 14nm since their 3D NAND should be in its fourth generation by now but I guess they may still have seen some profit in shrinking the lithography a bit more.
Going by the TBW though it seems that endurance may have suffered since it is down to 35/70TBW from the 75/150TBW that the S2-series had.

Thanks. Will update the Opening post.

P.S. Very surprised to see SM2254 as the controller. I couldn't find that one on the Silicon Motion Website.