Discussion ITT: We predict how low the price floor on 120GB/128GB SSD drops

[VirtualLarry]

Yeah, the ironic thing is how SSDs is going replace HDDs in OEM PCs not because of GB/$ or performance, but purely on a lowest cost of entry basis. The HDD makers can't compete here since the enclosure and assembly of HDDs themselves already cost more than slapping chips on a SSD PCB.

This is true, however, there's a slight confounding factor. Those users who have not yet used an SSD, and are stuck in their ways (likely to be the type to "download everything", and not yet be "hip" to streaming services for content), are likely to look only at the capacity of the primary storage device, when shopping for a PC.

To accommodate those kinds of people, the OEMs would have to include both a small boot SSD (but large enough for OS updates!), AND a "storage drive" (HDD).

Which leads to the obvious question, why not just use a "Hybrid Drive"?

Seagate has been making some progress in this area, their desktop Hybrid Drives are 7200RPM, at least. And now they've got "Gamer" versions, with 32GB of NAND (enough to actually be somewhat useful), instead of only 8GB (useful only for speeding up boot-ups).

Edit: I guess I don't see SSDs replacing HDDs in OEM systems (completely), until we get 500GB SSDs for $50, and 1TB SSDs for $105. Until then, we might see OEMs offering specific models of their PCs with SSDs, but they will still be selling the HDD-equipped models, for those that want more storage, or want something ever cheaper. Though I agree with your premise, that it would likely be cheaper for an OEM to include a 60GB SSD (OEM-branded), than a 500GB HDD.

 

[cbn]

Seagate has been making some progress in this area, their desktop Hybrid Drives are 7200RPM, at least. And now they've got "Gamer" versions, with 32GB of NAND (enough to actually be somewhat useful), instead of only 8GB (useful only for speeding up boot-ups).

I couldn't find any Seagate desktop Hybrid drives with 32GB NAND, but I did find this Seagate 2.5" Gamer SSHD (1TB, 32GB NAND) For $118.95 shipped:

http://www.amazon.com/Seagate-Gamin...0LX001&qid=1465431531&ref_=sr_1_1&s=pc&sr=1-1

Here is a review demonstrating a boot time speed up of 7.5 seconds (compared to 10 seconds for the AMD R7 240GB SSD) after restarting the PC ten times:

http://www.pcworld.idg.com.au/review/seagate/st1000lx001/600632/

Our first test involved simply booting our Windows 10 test rig. This usually runs on a standard 240GB AMD R7 SATA 3 SSD. It takes 20 seconds to boot to desktop. We cloned the disk and switched to the SSHD which then took 30 seconds to boot to desktop - 50% slower. But the story doesn’t end there. We kept rebooting and, over the course of 10 reboots saw the boot time drop gradually down to a consistent 22.5 seconds - just 12.5% slower.

Compared to a 1TB SSD, the drive I linked above does offer a savings of around $100, but at the same time for someone that wants or needs lots of storage a 2TB 2.5" HDD would be cheaper.

Maybe Seagate should also offer a 2TB 2.5" SSHD?

P.S. As far as the mechanism of how SSHDs work, Here is some info from an old Anandtech article (I am assuming the mechanism is the same with new SSHD drives, but I am not certain):

http://www.anandtech.com/show/3734/seagates-momentus-xt-review-finally-a-good-hybrid-hdd

The size of the NAND was a shocker to me when I first heard it. I honestly expected something much larger. In the Momentus XT however, the SLC NAND acts exclusively as a read cache - writes never touch the NAND. The drive looks at access patterns over time (most likely via a history table of LBAs and their frequency of access) and pulls some data into the NAND. If a read request comes in for an LBA that is present in the NAND, it's serviced out of the 4GB chip. If the LBA isn't present in the NAND, the data comes from the platters.

 

[VirtualLarry]

My Reeinno ST120GB S3S3 2.5" 120GB SATA6G SSD that shipped in my Onda M3 mini-PC (complete, not barebones), has some pretty amazing benchmark scores.

I'll try to attach an image later on, but it has:
550MB/sec seq. read
260MB/sec 4K QD32 read, and almost that in write.

Pretty decent specs for a 120GB drive. Better in 4K QD32 than my SM951 128GB PCI-E AHCI M.2 SSD.

Of course, no-one on ebay seems to be selling this model, in 120GB. Sigh.

I did find that xtrememicro, the seller of the Hectron X1 drives, does sell a Reeinno 240GB drive for $68 or so. But for not much more, you could get a PNY CS2211 on sale, which is better still, I think, and from a US company.

Also, there appears to be 120GB and 240GB capacities of the Hectron X1 drives now. Whether those are "fake capacities" or real, I guess, remains to be seen. (If anyone purchases one of these larger-capacity versions, please post benchmarks.)

 

[VirtualLarry]

http://www.ebay.com/itm/KingDian-12...II-3-0-2-5-Digital-w-Cache-E8A5-/131774726346

KingDian 120GB TLC SSD

In the listed specs, they list "BCH" ECC. I thought all of the newer TLC drives used LDPCC. Maybe this is a cheap hodgepodge of a drive, using TLC with an older controller not designed for it? Probably will fail in six months.

 

[hojnikb]

Probably uses Phison S10, which is a TLC capable controller, but doesn't feature LDPC.

 

[Hellhammer]

Probably uses Phison S10, which is a TLC capable controller, but doesn't feature LDPC.

Or JMicron JMF680 aka Maxiotek MK8213/MK2815

 

[hojnikb]

Guys, its SMI 2256 according to internetz. The description is wrong.

http://www.king-dian.com/kingdian-s280-sata3-ssd-480gb

 

[Glaring_Mistake]

Guys, its SMI 2256 according to internetz. The description is wrong.

http://www.king-dian.com/kingdian-s280-sata3-ssd-480gb

Seen a review in which one was opened and there was an SM2256 inside so I also think it most likely that it uses LDPC ECC even if their own website says it uses BCH ECC.
Because why exactly would you use BCH ECC when the controller supports LDPC?

 

[cbn]

I've seen that KingDian S280 listed as SM2256 in other listings too.

 

[cbn]

My KingFast F6 32GB in SATA6G on H110:

As a comparison of low end 32GB SSD vs. 32GB eMMC, below is the CrystalDiskMark score for the 32GB eMMC used in CX-W8 Pro Mini-PC. Review here.

Reads are massively in favor of the KingFast F6 SSD, the only exception being the 4K read which is roughly tied. The eMMC does win Sequential write with 59.56 MB/s vs. 40.69 MB/s but loses both 4K writes by a huge margin. ( 1.88 MB/s vs. 34.13 MB/s and 2.033 MB/s vs. 37.99 MB/s)

For an overall computing experience I think even this low end SSD (KingFast F6) would better than soldered on storage (BGA SSD would be a possible exception).

 

[cbn]

Also, there appears to be 120GB and 240GB capacities of the Hectron X1 drives now. Whether those are "fake capacities" or real, I guess, remains to be seen. (If anyone purchases one of these larger-capacity versions, please post benchmarks.)

Here is a listing I found for the 120GB and 240GB Hectron X1:

http://www.aliexpress.com/item/100-..._3&btsid=99e25fdb-8a6f-4a7e-a3db-bcc0b4af6461

120GB = $33.67 shipped
240GB = $48.96 shipped

 

[cbn]

This is interesting though. Just gave SSD Plus (which apparently uses SM2246XT as well) a bench and this is what i got (mind though, this is a 480GB version).

I ran the benchmark twice. Interesting results. Looking at this, if you have enough NAND, you can easily pull lots of 4K-32 read performance out of this controller.
Although it still baffles my mind, as to why read performance is limited by capacity at all, if sequentially you can pull off 500MB/s+ with only 64GB of flash.

Anyone has any ideas ?

Any ideas of what happened with the following KingSpec SM2246XT 512GB MLC SSD :

http://rageworx.ddns.net/index.php/kingspec-ssd-512gb-acsc2m512s25-review/

Firmware issue causing an extremely low 4K Q32T1 read score?

 

[hojnikb]

could be poor alignment..

 

[cbn]

Some info on Micron Gen 2 3D NAND :

http://www.tomshardware.com/news/micron-3d-nand-intel-xpoint,32116.html

Capacity doubles to 64GB MLC or 96GB TLC per die.

Here is what the article mentioned about NAND parallelism:

Such radical density increases can hamper performance, as SSDs require plenty of parallelism to offer speedy performance. Micron and Intel designed their 3D NAND die with four planes to increase performance in comparison to competing dual-plane designs.

However, there was no specific mention how many planes these new high capacity Gen 2 3D NAND dies would have.

 

[VirtualLarry]

Some info on Micron Gen 2 3D NAND :

http://www.tomshardware.com/news/micron-3d-nand-intel-xpoint,32116.html

Capacity doubles to 64GB MLC or 96GB TLC per die.

What happened to 1st-gen Micron 3D NAND? Did that even ever hit the market? Will it? Still waiting for uber-cheap, high-capacity, SSDs with 3D NAND. (Without the Samsung price premium.)

(Any truth to the rumor that the Team Group "L7 EVO" SSD is 3D NAND?)

Edit: According to documentation from Team, it uses a "new Marvell controller", and comes in 120GB and 240GB capacities.

Wouldn't 3D NAND come in "odd" capacities?

 

[cbn]

What happened to 1st-gen Micron 3D NAND? Did that even ever hit the market? Will it? Still waiting for uber-cheap, high-capacity, SSDs with 3D NAND. (Without the Samsung price premium.)

It just got released.

http://www.anandtech.com/show/10274/the-crucial-mx300-750gb-ssd-review-microns-3d-nand-arrives

I was surprised at the 4K read being so slow compared to other drives:

(Any truth to the rumor that the Team Group "L7 EVO" SSD is 3D NAND?)

Edit: According to documentation from Team, it uses a "new Marvell controller", and comes in 120GB and 240GB capacities.

Wouldn't 3D NAND come in "odd" capacities?

If it used Micron 3D TLC NAND it would be 90GB, 180GB, 360GB, etc.

P.S. I wonder what Marvell controller it uses?--> http://www.marvell.com/storage/ssd/

(The only one on that list that fits this kind of drive would be 88NV1120, which is the dram-less 28nm controller (with or without embedded SRAM?) also used in the planar TLC ADATA SP580)

 

[VirtualLarry]

Well, I ordered two Team Group "L7 EVO" 240GB SSDs. We'll see, I might even take one apart. But you're right, probably planar TLC, with a DRAM-less Marvell controller.

 

[hojnikb]

Wouldn't 3D NAND come in "odd" capacities?

Not necessary. With 48GB die you can have 240, 480 and 960 capacity points just fine (but with less OP than its usually with 240gb+ drives). Although parallelism might suffer since there are odd number of dies in case of 240gb model.

 

[VirtualLarry]

Some info on Micron Gen 2 3D NAND :

http://www.tomshardware.com/news/micron-3d-nand-intel-xpoint,32116.html

Capacity doubles to 64GB MLC or 96GB TLC per die.

IMFT Gen 2 3D NAND could help to reduce the cost of SSDs into the sub-$0.15-per-GB region when it comes to market, which makes insanely capacious SSDs even more attractive.

$0.15/GB? Please! $150 1TB SSDs...

 

[VirtualLarry]

The ebay seller (xtrememicro) also has KingFast K6 F6 32GB 2.5" for $15.54 shipped and KingFast K6 64GB 2.5" for $25.69 shipped.

http://www.ebay.com/itm/SSD-32GB-Ki...182795?hash=item3f618716cb:g:JokAAOSw8aNXGJ-J

Now $23.39 for the 32GB.

Why the steep price rise? Are we starting to see the precipice of inflation? Or is that the MO of the China sellers; introduce with a low price, and then keep raising it?

 

[cbn]

http://www.ebay.com/itm/SSD-32GB-Ki...182795?hash=item3f618716cb:g:JokAAOSw8aNXGJ-J

Now $23.39 for the 32GB.

Why the steep price rise? Are we starting to see the precipice of inflation? Or is that the MO of the China sellers; introduce with a low price, and then keep raising it?

There is another seller on ebay offering that drive for $21.19 free shipping (from Hong Kong).

But yes, that and the price you found are both good increases from the previous $15.54 shipped sale we saw earlier.

Not sure what was going on? Maybe the earlier price was an inventory reduction sale?

 

[VirtualLarry]

Now they're both $20.99 and $20.89. Weird.

 

[cbn]

Another cost reduction in NAND to consider is Samsung and SK Hynix 14nm planar NAND :

http://www.custompcreview.com/news/samsung-mass-produces-14nm-nand/27476/

So for the price floor SATA 6 Gbps drives, I am thinking we might see the DRAM-less SM2246XT coupled to SK Hynix 14nm MLC as a possible combination.

Some other possible DRAM-less combinations:

SM2258XT with Micron 3D TLC (with that mentioned, I am surprised that SM2258XT does not have SLC caching mentioned in the specs--> http://www.siliconmotion.com/A3.2_Partnumber_Detail.php?sn=7. If this holds, then this controller/NAND combination would not have a write advantage against SM2246XT/SK Hynix 14nm MLC )
Marvell 88NV1120 with SK Hynix 14nm MLC
Marvell 88NV1120 with Micron 3D TLC
Marvell 88NV1120 with either Toshiba or Sandisk 15nm TLC (ADATA SP580 already has this combination using Sandisk NAND)
Phison S11 with 15nm Toshiba TLC
Phison S11 with Micron 3D TLC (less likely than the Toshiba 15nm TLC)
Phison S11 with SK Hynix 14nm MLC ((less likely than the Toshiba 15nm TLC)
Maxiotek MK8115 (believed to be a DRAM-less Jmicron JMF680) with Micron 3D TLC (ADATA SU700 SSD has this combination of parts)
Maxiotek MK8115 with SK Hynix 14nm MLC

P.S. I do think the possibility also exists that the SK Hynix will allow their 14nm NAND to be configured as TLC (This assuming, of course, it works well enough with three bits per cell).

 

[VirtualLarry]

I wonder which is more reliable, Micron 3D TLC, or the SK Hynix 14nm planar (TLC)?

Previously it was expected that Samsung would abandon planar NAND development following the launch of their 3D V-NAND product; however, Samsung launched their 16nm NAND product earlier in the year and has further completed development on 14nm NAND as well. Planar NAND based on the 14nm manufacturing process is very tough to develop. With every manufacturing process shrink, the number of electrons that can be stored in a floating gate decreases. At the 14nm level, that could be just a few electronics per die, which seriously reduces the margin of error when reading or writing. This is the primary reason why most other memory makers have decided to cut development of planar NAND beyond 15nm.

Samsung has not yet stated whether or not it plans on development 3-bit MLC (TLC) NAND on the 14nm manufacturing process. Experts believe that they will face hurdles due to the same reason mentioned above. If successful however, production cost could be further reduced.

 

[cbn]

Not necessary. With 48GB die you can have 240, 480 and 960 capacity points just fine (but with less OP than its usually with 240gb+ drives). Although parallelism might suffer since there are odd number of dies in case of 240gb model.

Do you know what capacity NAND die Intel 540s, ADATA SP550 and AMD R3 use?

I know they use SK Hynix 16nm planar NAND. I also know looking back at the SMI production tool you posted in #37 for Hectron X1 that there is a 128 Gbit MLC die.

So If Intel, ADATA and/or AMD use that 128 Gbit (16GB) MLC as 192Gbit (24GB) TLC that means the 120GB SSDs use an odd number of dies (eg, 120GB Intel 540s would use five 192Gbit (24GB) TLC dies).

Maybe this also explains why the Intel 540s offers better performance when configured as 180GB (eight 192Gbit (24GB) TLC dies with 12GB overprovisioning) rather than 120GB (five 192Gbit (24GB) TLC dies with no overprovisioning) :

http://www.anandtech.com/show/10432/the-intel-ssd-540s-480gb-review