SSD "which controller?" list

[oupimiquo]

In light of the number of questions regarding which SSD to buy, and which controller each SSD has, I thought a bit of a list could be made up. This is an almost certainly incomplete list of SSDs and which controller they use. This is just the "manufacturers" that I could think of off the top of my head. Let me know if I've missed any (or got any of the controllers wrong) and I'll update it. Note that all drives with the same controller will perform very similarly. There's only a handful of companies that actually design and manufacture the drives, and OCZ, Patriot, etc just stick the manufactured PCB into a case (or in some cases, just stick a label on).

On a related note, I've been building up a collection of tools to benchmark SSDs. If there's people in Canberra who'd be happy to lend me their drive for a few days, I can start building up a list of detailed performance specs on the drives (waaaaayyy more detailed than any review I've seen). I say Canberra-only simply for ease of transaction - I'm happy to pay the shipping costs if you're outside Canberra, but you'd obviously have to just trust me to return it ...

Controllers

First, for an explanation of page mapping, see my post here, or the prettier but not quite correct Anandtech article here. Block mapping uses the same basic idea as page mapping, but at a much coarser level - one logical block per physical block, or even multiple physical blocks per logical block. The logical block size is a critical factor in SSD performance. Smaller block sizes lead to better small-write performance, larger block sizes lead to better sequential performance.

JMicron JMF601: Presumed predecessor to the JMF602. Appears to have the same overall structure as the JMF602, but only 4 channels instead of 8.
JMicron JMF602: Everyone's favourite Uses block mapping, which with MLC, not recommended for a drive that will be seeing a large number of writes, especially as a system drive. MLC drives will drop down to ~4 writes/sec under random write conditions (compared to at least 50 for most recent HDDs). SLC-based drives using the JMicron controllers are much better and quite usable as system drives due to the advantages of SLC over MLC flash (primarily the smaller block size and faster erase). Has eight 8-bit flash channels, 3.0 Gbps SATA II, and USB II support.
JMicron JMF602B: Very little information exists on this controller. From the name, it looks like a revision of the JMF602. Apparently marginally higher random write performance compared to the JMF602. Could be simply a higher-clocked version, who knows.
JMicron JMF612 (not yet released): JMicron said that they'd be showing off their new controller at Computex 2009 - I wasn't there, but none of the sites reporting on Computex said anything about JMicron, and there's no mention of the controller on their site yet. Not an especially good sign. Rumoured to have an ARM9 embedded processor (specific specs not given), supports DDR and/or DDR2 (probably 16-bit width, unknown bus frequency, up to 256 MB supported) and sports eight 8-bit flash channels (unknown data rate). Like the 600-series chips it supports USB II and 3.0 Gbps SATA II. Logical block size etc and performance are unknown at this stage.

Dual JMicron JMF602B: Not as bad as a single JMicron, but still not recommended as a system drive if it's MLC based (will still stutter under random write load).

Indilinx Barefoot: Works great for a system drive. Slower under random writes than the Intel drive, but faster for sequential writes. Looks like it uses logical pages somewhere around 128 KB, so somewhere between block and page mapping. Has four 16-bit flash channels, uses an embedded ARM7TDMI processor (estimated ~233 MHz), and supports 16 to 64 MB 166 MHz SDRAM (32 bits wide). There's multiple known part numbers: IDX22A4BI0-A0ER (pre-production ES silicon), IDX110M00-FC (production silicon), IDX110M00-LC (same as the -FC, but with more chip select pins so larger capacities possible), IDX110M01-LC (minor tweaks to the M00 to allow the use of 34 nm Intel flash).
Indilinx Jet Stream (not released yet): The successor to the Barefoot. Not much is known about this controller, besides that it should support 6.0 Gbps SATA. Pencilled in for a 3Q09 release back in Feb 09, so details should hopefully emerge soon.

Intel PC29AS21AA0: Dominates for random writes, but a bit pokey for sequential writes unless you go for the SLC drives. No problems whatsoever as a system drive. Appears to use pure page mapping (4 KB for MLC, 2 KB for SLC). Has ten 8-bit flash channels. Intel designed and manufactured? Information is sketchy on the issue.
Intel PC29AS21BA0: The "G2" controller appears to largely identical to the "G1" controller, and has only a minor change in part number. Perhaps as little as a small die revision to make it compatible with the 34nm flash, nothing too much though. The sequential write issue with MLC flash (likely caused by poor hardware support for the ECC algorithms used with MLC, which are significantly more complex than those used with SLC) does not appear to have been resolved or improved.

Samsung: Samsung have several different controllers, but don't say which controllers are used in which drives. Reviewers seem reluctant to pop the casings of the Samsung drives, so it's hard to narrow things down. Mostly pretty good though.
Samsung S3C49RBX01: It's ARM based. Needs external SDRAM. That's about all I can find out about it ... Performance numbers suggest four 8-bit flash channels, and block mapping. Used only in SLC drives?
Samsung S3C29RBB01: A newer controller from Samsung, used in MLC drives. It's got eight flash channels according to SuperTalent (benchmarks indicate 8-bit channels). However, it's not clear whether it uses block mapping or something more fine-grained. Proper used-drive numbers appear to show around 200 random 4K writes/sec, which could be done with pure block mapping.

Mtron: According to Mtron, drives prior to 4Q07 used FPGA-based controllers. After this point, they used an ASIC. It's likely that both offer identical features and performance, with the ASICs simply high-volume versions of the FPGA design. It appears that the current design has four 8-bit channels to the flash devices, and a PATA host interface. External SDRAM is required. Random write IOPs (60 for the Mobi 3000, 130 for the Mobi 3500, 25 for the MLC-based Mobi 1000) suggest a block mapping (JMF602 style) implementation with an unknown logical block size. Apparently an 8-channel version is in the works.

Drives

A-Data

S391S - MLC, JMicron JMF602
S591S-<nnn>GS - SLC, unknown controller, probably JMicron JMF602.
S591S-<nnn>GM - MLC, unknown controller, probably JMicron JMF602.
S592S - MLC, Indilinx. Interestingly, the PCB layout is different from that used by either OCZ/G.Skill/etc or SuperTalent.
SINTS - X-25M rebadge.

SX81 - MLC, JMicron JMF602.
SX93 - MLC, JMicron or Dual JMicron, not clear.
SX94 - MLC, Dual JMicron.
SX95 - MLC, Indilinx.

Corsair

S128 - MLC, Samsung S3C49RBX01
P256 - MLC, Samsung S3C29RBB01

Crucial
CT128M225 - MLC, Indilinx. Interestingly, they use SuperTalent's PCB rather than whoever it is that makes them for G.Skill, OCZ, etc.

G.Skill

Falcon - Indilinx
Titan - dual JMicron JMF602B
(no series name) - JMicron JMF602B

Falcon II - Indilinx M01 with 34 nm Intel flash.

G.Skill part numbers indicate whether it's SLC or MLC. They are of the form: FM-<xxx>-<nnn>GB<zzz>
<xxx> = "25S2" for SLC, "25S2S" for MLC
<nnn> = size in HDD-manufacturer-GB
<zzz> = empty for single JMicron, "T1" for dual JMicron (Titan), "F1" for Indilinx (Falcon).

Intel

All Intel drives use the same basic controller (Intel designed and manufactured? Information is sketchy on the issue). Their -M drives are MLC, the -E drives are SLC.

X25-M: MLC, Intel PC29AS21AA0.
X25-M G2: MLC, Intel PC29AS21BA0.
X25-E: SLC, Intel PC29AS21AA0.
X25-X: MLC, Intel PC29AS21BA0, but only half the flash channels used.

Various flash part numbers are 29F64G08FAMC1 (80 GB X18-M) and 29F32G08CAMC1 (80 GB X25-M). Presumably both are of the same generation, just different capacities or number of dies on the package.

The G2 Intel drives use a new revision of the controller that is used on the G1 drives, and 34 nm flash (part number 29F16B08JAMD1).

Kingston

SSDNow E Series : Rebadged Intel X25-E (MLC, Intel PC29AS21AA0)
SSDNow M Series : Rebadged Intel X25-M (MLC, Intel PC29AS21AA0)
SSDNow V Series : JMicron JMF602B, sort of. It's a Toshiba-branded chip, and claims to be using tweaked firmware plus has 64 KB RAM instead of 16 KB. However, it's still the same problematical controller, so until contrary evidence turns up in the form of 4 KB write tests on a "full" drive not being in the 4-8 IOPS range, it's best to assume it's no different than any other JMF602B drive.
SSDNow V Series 40 GB Boot Drive: Rebadged X25-X (MLC, Intel PC29AS21BA0, only half the flash channels used).
SSDNow V+ Series: MLC, Samsung S3C29RBB01

Mtron

I don't really know much about Mtron drives. AFAIK, they all use the same Mtron-designed FPGA/ASIC controller.

Mobi 1000 - MLC
Mobi 3000 - SLC
Mobi 3500 - SLC

OCZ

Agility - MLC, Indilinx, supposed to be slower than the Vertex through the use of lower cost flash chips. They appear to be using Intel 29F64G08FAMC1 flash, which is the same family as that used in the first-generation Intel SSDs.
Agility EX - SLC, Indilinx, again supposed to be slower than the Vertex EX through the use of lower cost (but still SLC) flash chips. Possibly Intel inside again rather than Samsung? Who knows ...
Apex - MLC, dual JMicron JMF602B
Core - MLC, JMicron JMF602
Core V2 - MLC, JMicron JMF602B
Solid - MLC, JMicron JMF602B
Solid 2 - MLC, crippled Indilinx controller. Until someone pops the lid, it's not clear how it's been crippled. The spec'd speeds seem to indicate only half the channels being used or something.
Summit - MLC, Samsung S3C29RBB01
Vertex - MLC, Indilinx
Vertex Turbo - MLC, Indilinx, overclocked controller + RAM, but the flash is apparently still running at the same speed so it's unclear whether there's any real-world performance advantage.
Vertex EX - SLC, Indilinx
"OCZ SATA II SSD" (ie: no series name, part numbers OCZSSD2-1S32G, OCZSSD2-1S32G) - SLC, unknown Samsung controller?
Z-Drive - Currently unknown. Specs indicate either 1) a cacheless Highpoint RocketRAID card with four Summit or Vertex drives, or 2) a Highpoint RocketRAID card with 256 MB cache, and 4 Apex/Core V2/Solid drives.

Patriot

Warp - MLC, JMicron JMF602?
Warp V2 - MLC, JMicron JMF602B
Warp V3 - MLC, dual JMicron JMF602B
Torqx - MLC, Indilinx
Torqx M28 - MLC, unknown, though 128 MB RAM implies Samsung S3C29RBB01 rather than Indilinx.

PQI

X25-M - The same as the Intel drive of the same name ...
S525/S518 - JMicron, appears to be both SLC and MLC versions floating around.

RunCore

RunCore do a lot of mini-PCIe and ZIF form factor drives as well. The following is just their 2.5" SATA lineup.

Pro III - MLC, JMicron JMF602B
Pro IV - MLC, Indilinx

Samsung

Note that Samsung tends to supply drives to computer assemblers (Dell, Apple, etc) as opposed to selling direct to the end user. So there's probably lots more out there that aren't included below.

PB22-J - MLC, Samsung S3C29RBB01

SiliconSystems (acquired by Western Digital)

SiliconSystems presumably uses their own controller in their drives, but there's no details available about what this controller is.

SiliconDrive III - No information known.

Solidata

K1 - SLC, JMicron
K2 - MLC, JMicron
K5 - SLC, Indilinx
K6 - MLC, Indilinx
X1 - SLC, dual JMicron
X2 - MLC, dual JMicron

SuperTalent

UltraDrive LE - SLC, Indilinx
UltraDrive ME - MLC, Indilinx. SuperTalent use their own PCB layout for this drive (and presumably their SLC version as well). There don't appear to be any functional differences though.
UltraDrive GX (MLC) - New name for the UltraDrive ME.
UltraDrive GX (SLC) - New name for the UltraDrive LE.
MasterDrive BX - SLC, unknown (only SATA I, so whatever it is, it's old )
MasterDrive OX - MLC, JMicron
MasterDrive PX - SLC, JMicron
MasterDrive RX - SLC (parts starting with FTD) or MLC (parts starting with FTM), dual JMicron JMF602B
MasterDrive SX - MLC, Samsung S3C29RBB01

Transcend

For the models below, <nnn> is 32, 64, etc depending on the SSD capacity.
TS<nnn>SSD25S-S - SLC, JMicron JMF602B
TS<nnn>SSD25S-M - MLC, JMicron JMF602B

Western Digital

See SiliconSystems.

Flash devices

To be filled out a lot more ... At the moment it's just a list of flash components and example drives. Remember that most drives using the same controller are actually identical at the component level, so any 120 GB Indilinx drive will be using K9HCG08U1M chips for example.

Samsung MLC:
K9HCG08U1M - Used in the 120 GB OCZ Vertex and similar drives.
K9HCGZ8U5M - Used in the 256 GB Samsung PB22-J and similar drives.
Both of these are from the same family of flash devices - the die in the K9HCGZ8U5M's are simply twice the capacity (4 planes instead of two) of those in the K9HCG08U1M (both have 4 dies/package).

Samsung SLC:
Samsung K9NCG08U5M - Used in the 120 GB OCZ Vertex EX.
4 planes/die, 2 dies/package. Not really much else of interest here.

[DiGiTaL MoNkEY]

Fantastic work oupimiquo, this will be very handy for a lot of people.

Many thanks!

[oohms]

this site posts internal shots:
http://benchmarkreviews.com/index.ph...id=3&Itemid=60

this guy opens some of the ssd's he reviews (japanese and a bit hard to navigate though)
http://blog.nabe.jp/archives/000169.html#more
http://blog.nabe.jp/archives/000157.html#more

http://www.linustechtips.com/how-to/...rnals-pictured

no ssd internals but interesting reading for those who want to:
http://www.anandtech.com/storage/sho...spx?i=3531&p=1

recent review.. controllers listed
http://techreport.com/articles.x/16848

[oupimiquo]

Cheers oohms. Managed to do the good old information-multiplier-via-Google trick from there

[supernerd]

It's a little annoying how few reviewers do a 4KB random write test. They base their conclusions on a sequential read and write times which are meaningless if the drive is going to be your boot device. Other than Anand - you have to wonder about these 'reviewers.' Anyways, thanks for the list of controllers.

[DirkDiggler76]

Originally Posted by supernerd

It's a little annoying how few reviewers do a 4KB random write test. They base their conclusions on a sequential read and write times which are meaningless if the drive is going to be your boot device. Other than Anand - you have to wonder about these 'reviewers.' Anyways, thanks for the list of controllers.

well then for a boot/os drive what do i want the most randoms read/write, sequentiual? more info please about that.


BTW top read . good work

[pher0x]

Great work mate, I've been hoping someone would do this.

Just a quick note though, it's Indilinx not Indilix :-)

Also the torqx drive does use the Indilinx Barefoot controller.

[The Mafia]

excellent thread, well done mate.


At least everyone knows now which devices use the right controllers.

How do we go about Firmwares?


As far as I know, Intel and OCZ (vertex series) are the only ones that have decent firmwares that allow the drive to not get slowdowns after use, and have a trim tool etc.

Do all of these other brands that use indilix controllers have as much firmware development?

[oupimiquo]

Originally Posted by pher0x

Just a quick note though, it's Indilinx not Indilix :-)

Err, oops, fixed

Originally Posted by pher0x

Also the torqx drive does use the Indilinx Barefoot controller.

Cheers. I've put a "?" after it as I haven't seen any photos with the top off, but the specs do line up with other Vertex drives.

Originally Posted by The Mafia

As far as I know, Intel and OCZ (vertex series) are the only ones that have decent firmwares that allow the drive to not get slowdowns after use, and have a trim tool etc.

Do all of these other brands that use indilix controllers have as much firmware development?

All Indilinx drives use the same firmware, support the same features, etc. The only difference is the ID string. While I wouldn't be game to try, I'd expect that you could run a SuperTalent drive with an OCZ firmware (for example).

[pher0x]

Originally Posted by oupimiquo

Cheers. I've put a "?" after it as I haven't seen any photos with the top off, but the specs do line up with other Vertex drives.

Yup, I emailed Patriot last week and they confirmed Indilinx Barefoot controller with 64 DRAM of cache.

Originally Posted by oupimiquo

All Indilinx drives use the same firmware, support the same features, etc. The only difference is the ID string. While I wouldn't be game to try, I'd expect that you could run a SuperTalent drive with an OCZ firmware (for example).

This I wouldn't be too sure - Tony from OCZ stated that they had to modify the firmware to stop competitors using their firmware on their drives - whether this is a simple check of the ID string (being a valid OCZ one), I'm not sure.

May be something to look into, I've shot off another email to Patriot to see if they can provide any more info.

[DirkDiggler76]

so i guess for a systems o/s drive.. then the vertex ? correct?

[jlien]

So long story short, avoid anything with a JMicron controller if I want to use it as a OS drive?

that being said, isn't that what people use ssd's for? i mean it's not like anyone will have them for storage lol.

[pher0x]

Originally Posted by jlien

So long story short, avoid anything with a JMicron controller if I want to use it as a OS drive?

I certainly would.

Originally Posted by jlien

that being said, isn't that what people use ssd's for? i mean it's not like anyone will have them for storage lol.

I use them for system disks, but if you do a lot of analytical/graphical work where you are working with large images/datasets you could also use it as a type of scratch disk.

Having it as a system disk is great though, windows feels so snappy and responsive.

[frieten]

subscribing, useful for when i get my SSD.

[DirkDiggler76]

want to buy a sdd over weekend.. intel bit expesvie or the vertex, as i want 120 gig 600 buck ok .. but for 32 gig intel 600 bucks

will i notice a difference between them as a o/s drive?