NAND Data Base

[Ao1]

NAND Data Base

I thought it might be useful to build a data base of NAND types as it affects speed, endurance and data integrity. It’s hard to come across manufactures product specs sheets, but if you find one and want to add details to the data base it can be found here:

https://docs.google.com/spreadsheet/...lLTdNQVE#gid=0

Anyone can access/ edit it and changes are auto saved.

[johnw]

Sorry, can't get the image to not resize, but if you zoom with the brower it is readable. Thanks XS for not being able to use attachemnts

That image resolution is only 800 x 220. Contrary to countless bad TV sequences, you cannot zoom in to get greater detail than what is actually present in the bitmap to begin with!

The database is a good idea, but it would be nice if we could read it. Have you considered putting it on google docs?

[Ao1]

Done, thanks for the suggestion

[minpayne]

Got these from a thread somewhere else: http://forum.coolaler.com/blog.php?b=989

ADATA S510 120GB - Micron 29F64G08CBAAA (25nm, 3000 P/E cycles, async)
ADATA S511 60GB - Intel 29F64G08AAME1
ADATA S511 120GB - Intel 29F64G08AAME1
ADATA S511 240GB - Intel 29F16B08CAME1
ADATA S599 64GB -
ADATA S599 128GB - Intel 29F64G08CAMDB (34nm)
ADATA S599 256GB - Intel 29F16B08JAMDB (34nm)

Apacer ProII AS202 - Hynix H27UBG8T2ATR (32nm, 3000 P/E cycles)
Apacer TurboII AS602 120GB - Intel 29F64G08CAMDB (34nm)
Apacer TurboII AS602 120GB - Intel 29F64G08CBAAA

Corsair Performance 3 128GB - Toshiba TH58TVG7D2FBA89 (32nm, 5000 P/E cycles, Toggle)
Corsair Performance 3 256GB - Toshiba TH58TVG8D2FBA89 (32nm, 5000 P/E cycles, Toggle)
Corsair Force Series 3 120GB - Micron 29F64G08CBAAA (25nm, 3000 P/E cycles, async)
Corsair Force Series 3 120GB - Micron 29F64G08CBAAB (25nm, 3000 P/E cycles, sync)
Corsair Force Series 3 240GB - Intel 29F16B08CAME1
Corsair Force Series GT 120GB - Intel 29F64G08ACME2 (25nm, 5000 P/E cycles, sync)
Corsair Force Series GT 120GB - Micron 29F64G08CBAAB (25nm, 3000 P/E cycles, sync)
Corsair Force Series GT 240GB - Micron 29F128G08CFAAB (25nm, 3000 P/E cycles, sync)

Crucial RealSSD C300 64GB - Micron OJB12 NW273 (34nm)
Crucial RealSSD C300 128GB - Micron OHB12 NW274 (34nm)
Crucial RealSSD C300 128GB - Micron OPD12 D9LGQ (34nm)
Crucial RealSSD C300 256MB - Micron OAB12(OAB11) NW172 (34nm)
Crucial RealSSD C300 512GB -
Crucial m4 64GB - Micron 29F64G08CFACB (25nm, 3000 P/E cycles, sync)
Crucial m4 128GB - Micron 29F64G08CFACB (25nm, 3000 P/E cycles, sync)
Crucial m4 256GB - Micron 29F128G08CFAAB (25nm, 3000 P/E cycles, sync)
Crucial m4 512GB - Micron 29F256G08CJAAB (25nm, 3000 P/E cycles, sync)

EZLINK Seraphim MLC - Intel 29F64G08CAMDB (34nm)
EZLINK Seraphim MLC - Intel 29F64G08CAMDA (34nm)
EZLINK Achilles MLC - Intel 29F64G08ACME2 (25nm, 5000 P/E cycles, sync)
EZLINK Emperor MLC - Intel 29F16B16MCME1 (25nm, 10000 P/E cycles, sync)

G.Skill Phoenix Pro 40GB - Intel 29F32G08AAMDB (34nm)
G.Skill Phoenix Pro 60GB - Intel 29F64G08CAMDB (34nm)
G.Skill Phoenix Pro 120GB - Intel 29F64G08CAMDB (34nm)
G.Skill Phoenix Pro 240GB - Intel 29F16B08JAMDB (34nm)
G.Skill Phoenix EVO 115GB - Micron 29F64G08CBAAA (25nm, 3000 P/E cycles, async)

Intel X25-M 160GB - Intel 29F64G08FAMC1 (50nm)
Intel X25-V G2 40GB - Intel 29F64G08CAMD1 (34nm)
Intel X25-M G2 80GB - Intel 29F64G08CAMD1 (34nm)
Intel X25-M G2 120GB - Intel 29F64G08CAMDA (34nm)
Intel X25-M G2 160GB - Intel 29F16B08JAMD1 (34nm)
Intel X25-E 32GB - Intel 29F16B08CAMC1 (50nm)
Intel SSD 310 80GB - Intel 29F16B08JAMDA (34nm)
Intel SSD 320 120GB - Intel 29F16B08CCME1
Intel SSD 320 160GB - Intel 29F16B08CCME1
Intel SSD 320 300GB - Intel 29F16B08CCME1
Intel SSD 320 600GB - Intel 29F32B08JCME1
Intel SSD 510 120GB - Intel 29F64G08CAMDD (34nm)
Intel SSD 510 120GB - Intel 29F16B08JAMDD (34nm)
Intel SSD 510 250GB - Intel 29F16B08JAMDD (34nm)
Intel SSD 710 200GB - Intel 29F16B08CCME1

Kingston SSDNow V100 128GB - Intel 29F16B08CCME1
Kingston SSDNow V100 256GB - Toshiba TH58NVG7D2FTA20 (32nm)
Kingston SSDNow V+100 64GB - Toshiba TH58NVG7D7EBAK2 (32nm)
Kingston SSDNow V+100 128GB - Toshiba TH58NVG7D7EBAK2 (32nm)
Kingston SSDNow V+100 256GB - Toshiba TH58NVG8D7FBAK2 (32nm)
Kingston SSDNow V+180 64GB - Toshiba TH58NVG6D7FBAK2 (32nm)
Kingston HyperX 120GB – Intel 29F16B08CCME2 (25nm, 5000 P/E cycles, sync)
Kingston HyperX 240GB – Intel 29F16B08CCME2 (25nm, 5000 P/E cycles, sync)

Mach Extreme MX DS 100GB - Intel 29F64G08CAMDB (34nm)
Mach Xtreme MX DS Turbo 120GB - Intel 29F64G08ACME2 (25nm, 5000 P/E cycles, sync)

Memoright STM-25 32GB/64GB/128GB - Intel 29F64G08CAMDB (34nm)
MemoRight FTM-25 120GB - Micron 29F64G08CBAAA (25nm, 3000 P/E cycles, async)
MemoRight FTM-25 240GB - Micron 29F128G08CJABA (29nm, 5000 P/E cycles, async)
MemoRight FTM Plus 60GB -
MemoRight FTM Plus 120GB - Micron 29F64G08CBAAB (25nm, 3000 P/E cycles, sync)
Memoright FTM Plus 240GB - Micron IMA12(INA12) NX287 (25nm, 3000 P/E cycles, sync)

Mushkin Callisto Deluxe 40GB - Intel 29F32G08AAMDB (34nm)
Mushkin Callisto Deluxe 60GB - Intel 29F32G08AAMDB (34nm)
Mushkin Callisto Deluxe 120GB - Intel 29F64G08CAMDB (34nm)
Mushkin Chronos Deluxe 120GB - Toshiba TH58TAG6D2FBA49 (32nm, 5000 P/E cycles, Toggle)
Mushkin Chronos Deluxe 240GB - Toshiba TH58TAG8D2FBA89 (32nm, 5000 P/E cycles, Toggle)

OCZ NOCTI 30GB - 打磨
OCZ NOCTI 60GB - 打磨
OCZ NOCTI 120GB - 打磨
OCZ Vertex 2 60GB - Hynix H27UBG8T2ATR (32nm, 3000 P/E cycles)
OCZ Vertex 2 60GB - Micron ONB17 NW169
OCZ Vertex 2 60GB - Intel 29F32G08AAMDB (34nm)
OCZ Vertex 2 120GB - Micron 29F32G08CBABA (29nm, 5000 P/E cycles, async)
OCZ Vertex 2 Extended 120GB - Intel 29F64G08CAMDB (34nm)
OCZ Solid 3 60GB - Micron 29F64G08CBAAA (25nm, 3000 P/E cycles, async)
OCZ Solid 3 120GB - Micron 29F64G08CBAAA (25nm, 3000 P/E cycles, async)
OCZ Agility 3 120GB - Micron 29F64G08CBAAA (25nm, 3000 P/E cycles, async)
OCZ Agility 3 120GB - Micron 29F128G08CFAAA (25nm, 3000 P/E cycles, async)
OCZ Agility 3 240GB - 打磨
OCZ Agility 3 240GB - Micron 29F128G08CFAAA (25nm, 3000 P/E cycles, async)
OCZ Vertex 3 120GB - 打磨
OCZ Vertex 3 60GB - Intel 29F64G08AAME1
OCZ Vertex 3 120GB - Intel 29F64G08AAME1
OCZ Vertex 3 120GB - Intel 29F64G08ACME2 (25nm, 5000 P/E cycles, sync)
OCZ Vertex 3 240GB - 打磨
OCZ Vertex 3 240GB - Intel 29F16B08CAME1
OCZ Vertex 3 240GB - Intel 29F16B08CCME2 (25nm, 5000 P/E cycles, sync)
OCZ Vertex 3 240GB - Intel 29F128G08CFAAB (25nm, 3000 P/E cycles, sync)
OCZ Vertex 3 Max IOPS 120GB - Toshiba TH58TAG7D2FBAS9 (32nm, 5000 P/E cycles, Toggle)
OCZ Vertex 3 Max IOPS 240GB - Toshiba TH58TAG7D2FBA89 (32nm, 5000 P/E cycles, Toggle)
OCZ Vertex 3 Max IOPS 240GB - Toshiba TH58TAG7D2FBAS9 (32nm, 5000 P/E cycles, Toggle)
OCZ Synapse Cache 64GB - Micron 29F64G08CBAAA (25nm, 3000 P/E cycles, async)
OCZ Octane 512GB - – Intel 29F32B08JCME2 (25nm, 5000 P/E cycles, sync)
OCZ RevoDrive 50GB - Intel 29F32G08AAMDB (34nm)
OCZ RevoDrive 120GB - Intel 29F64G08CAMDB (34nm)
OCZ RevoDrive X2 - Intel 29F32G08AAMDB (34nm)
OCZ RevoDrive Hybrid - 打磨
OCZ RevoDrive 3 - Intel/Micron 29F64G08CBAAA (25nm, 3000 P/E cycles, async)
OCZ RevoDrive 3 X2 - Intel/Micron 29F64G08CBAAA (25nm, 3000 P/E cycles, async)
OCZ Z-Drive R4 1.6TB - Intel 29F16B08CCME2 (25nm, 5000 P/E cycles, sync)

OWC Mercury Electra 6G 120GB -
OWC Mercury Electra 6G 240GB - Intel 29F16B08CAME1
OWC Mercury Extreme Pro 6G 120GB - Micron 29F64G08CBAAB (25nm, 3000 P/E cycles, sync)
OWC Mercury Extreme Pro 6G 240GB - Micron 29F128G08CFAAB (25nm, 3000 P/E cycles, sync)
OWC Mercury Extreme Pro 6G 120GB - Toshiba TH58TAG7D2FBA89 (32nm, 5000 P/E cycles, Toggle)
OWC Mercury Extreme Pro 6G 240GB - Toshiba TH58TAG7D2FBA89 (32nm, 5000 P/E cycles, Toggle)

Patriot Torqx 2 128GB - Toshiba TH58NVG7D2FLA89
Patriot Torqx 2 256GB - Toshiba TH58NVG7D2FTA20
Patriot Pyro 120GB - Micron 29F64G08CBAAA (25nm, 3000 P/E cycles, async)
Patriot Pyro 240GB - Intel 29F16B08CAME1
Patriot Pyro SE 120GB - Micron 29F64G08CBAAB (25nm, 3000 P/E cycles, sync)
Patriot Pyro SE 240GB - Micron 29F128G08CFAAB (25nm, 3000 P/E cycles, sync)
Patriot Wildfire 120GB - Toshiba TH58TAG6D2FBA49 (32nm, 5000 P/E cycles, Toggle)

Plextor PX-128M2S - Toshiba TH58TVG7D2FBA89 (32nm, 5000 P/E cycles, Toggle)
Plextor PX-256M2S - Toshiba TH58TVG8D2FBA89 (32nm, 5000 P/E cycles, Toggle)
Plextor PX-128M2P - Toshiba TH58TVG7D2FBA89 (32nm, 5000 P/E cycles, Toggle)
Plextor PX-256M2P - Toshiba TH58TVG8D2FDA88 (32nm, 5000 P/E cycles, Toggle)
Plextor PX-128M3 - Toshiba TH58TEG7D2HBA4C (24nm, 5000 P/E cycles, Toggle)
Plextor PX-512M3 - Toshiba TH58TEG9D2HBA89 (24nm, 5000 P/E cycles, Toggle)

SanDisk Ultra 120GB - 打磨
SanDisk Ultra 240GB - 打磨

Silicon Power Velox V20 60GB - Hynix H27UBG8T2ATR (32nm, 3000 P/E cycles)
Silicon Power Velox V20 120GB - Micron 29F64G08CFABA (29nm, 5000 P/E cycles, async)
Silicon Power Velox V30 60GB - Intel 29F64G08AAME1
Silicon Power Velox V30 60GB - Intel 29F64G08ACME2 (25nm, 5000 P/E cycles, sync)
Silicon Power Velox V30 120GB - Intel 29F64G08AAME1
Silicon Power Velox V30 120GB - Intel 29F64G08ACME2 (25nm, 5000 P/E cycles, sync)

Team Xtreem S3 - Intel 29F64G08ACME2 (25nm, 5000 P/E cycles, sync)

[nuttcase21]

great idea, now why didn't i think of this...

[Ao1]

[Ao1]

Here’s a quick summary of NAND performance metrics that I have been able to collect so far. It seems that as the page file size increases the page program and random read speed increase disproportionately; whilst the impact on sequential read and block erase times have little impact.
It would be nice to how an SSD using 50nm SLC NAND performed with the latest controller technology.



Same graph without page file size

[johnw]

Here’s a quick summary of NAND performance metrics that I have been able to collect so far. It seems that as the page file size increases the page program and random read speed increase disproportionately; whilst the impact on sequential read and block erase times have little impact.

There are some anomalies in the data. For example, the Intel JS29F64G08AAME1 flash appears to have a random read time of better than 1/4 the other 8KiB page flash. It is hard to believe that both specifications are talking about the same thing with that big of a difference.

Also, a good SSD can achieve about 6000 IOPS for QD=1 4KiB random read (4000-5000 would be on the slow side, 7000 would be very good). 6000 IOPS is equivalent to 167us. But that includes latency from the flash as well as the controller and SATA. So the flash itself would have to be even faster than 167us. But most of the 8KiB flash seems to have 200us or greater for random read (except for the Intel, again).

So it seems something is not right with some of the data.

[Ao1]

Here are a couple of screen shots showing the variation in random read times. I don't think the data sheets are showing anything that can not be observed in performance differences between different SSD's.





Cache/ read ahead plays a role.

Micron state:

The READ PAGE CACHE RANDOM (00h-31h) command reads the specified block and page into the data register while the previous page is output from the cache register. This command is accepted by the die (LUN) when it is ready (RDY = 1, ARDY = 1). It is also accepted by the die (LUN) during READ PAGE CACHE (31h, 00h-31h) operations (RDY = 1 and ARDY = 0).

To issue this command, write 00h to the command register, then write five address cycles to the address register, and conclude by writing 31h to the command register. The column address in the address specified is ignored. The die (LUN) address must match the same die (LUN) address as the previous READ PAGE (00h-30h) command or, if applicable, the previous READ PAGE CACHE RANDOM (00h-31h) command. There is no restriction on the plane address.

After this command is issued, R/B# goes LOW and the die (LUN) is busy (RDY = 0, ARDY = 0) for tRCBSY. After tRCBSY, R/B# goes HIGH and the die (LUN) is busy with a cache operation (RDY = 1, ARDY = 0), indicating that the cache register is available and that the specified page is copying from the NAND Flash array to the data register. At this point, data can be output from the cache register beginning at column address 0. The CHANGE READ COLUMN (05h-E0h) command can be used to change the column address of the data being output from the cache register.

In devices that have more than one die (LUN) per target, during and following interleaved die (multi-LUN) operations the READ STATUS ENHANCED (78h) command followed by the READ MODE (00h) command must be used to select only one die (LUN) and prevent bus contention.

If a MULTI-PLANE CACHE RANDOM (00h-32h, 00h-31h) command is issued after a MULTI-PLANE READ PAGE operation (00h-32h, 00h-30h), then the addressed pages are read into the data registers while the previous pages can be output from the cache registers. After the die (LUN) is ready (RDY = 1, ARDY = 0), the CHANGE READ COLUMN ENHANCED (06h-E0h) command is used to select which cache register outputs data.

Hynix state:

Upon initial device power up, the device defaults to Read1(00h/01h) mode. This operation is also initiated by writing 00h to the command register along with followed by the four address input cycles. Once the command is latched, it does not need to be written for the following page read operation.

Three types of operations are available: random read, serial page read and sequential row read. The random read mode is enabled when the page address is changed. The 528 bytes (x8 device) of data within the selected page are transferred to the data registers in less than access random read time tR. The system controller can detect the completion of this data transfer tR by analyzing the output of R/B pin. Once the data in a page is loaded into the registers, they may be read out in 30 ns cycle time by sequentially pulsing RE. High to low transitions of the RE clock output the data stating from the selected column address up to the last column address.

After the data of last column address is clocked out, the next page is automatically selected for sequential row read. Waiting tR again allows reading the selected page. The sequential row read operation is terminated by bringing CE high.

The way the Read1 and Read2 commands work is like a pointer set to either the main area or the spare area (Refer to Figure 19) . Writing the Read2 command user may selectively access the spare area of bytes 512 to 527. Addresses A0 to A3 set the starting address of the spare area while addresses A4 to A7 are ignored.

Unless the operation is aborted, the page address is automatically incremented for sequential row Read as in Read1 operation and spare sixteen bytes of each page may be sequentially read. The Read1 command (00h/ 01h) is needed to move the pointer back to the main area. Figure 9 to 11 show typical sequence and timings for each read operation.

[Christopher]

Couldn't some of the discrepancy be explained by min/max/typical/average numbers?

[Ao1]

In the examples above the specs are directly comparable.

Faster flash interface design enables latency to be reduced or remain static with larger page file sizes, whilst bandwidth can be increased. Here is a link to a Micron presentation that talks about ONFI 3.0. http://extmedia.micron.com/webmedia/onfi3/onfi30.html

Hynix, Intel, Micron, SanDisk and several less well-known corporations formed the ONFI consortium (Open NAND Flash Interface)

Samsung and Toshiba have their own specification: Toggle Mode DDR

I believe the Plextor PX-256M3 is the first to use 166MT/s Toggle Mode DDR

ONFI specs as below.

ONFI 2.0 – 133MT/s
ONFI 2.1 – 166MT/s and 200MT/s
ONFI 3 – 400MT/s

[Christopher]

I thought the 830 was using 166MT Toggle mode. (BUT I just looked and supposedly it uses 133MTs)

But Samsung started shipping Toggle 2.0 400MTs NAND last year, in 64gbit packages.

[Christopher]

打磨

I tried translating this:
Polished, in English. What the hell, Google?

Is there an actual Chinese person who can tell me what the fudge this means?

[neeyuese]

打磨 means make the original brand & model number to another special mark.

For example:

IMFT Flash but OCZ MARK

[Christopher]

Excellent. That makes a lot more sense.

[Ao1]

Out of curiosity I checked to see what NAND my Lumia 800 uses. It is listed in the product schematics as eMMC Hynix H26M52002CKR, although a photo in the same document appears to show a Toshiba product.

It’s easy to see that smaller geometries are being driven by mobile devices rather than SSD’s. I’d guess the market for NAND in the SSD market is quite small compared to the plethora of mobile device applications.

[Christopher]

It’s easy to see that smaller geometries are being driven by mobile devices rather than SSD’s. I’d guess the market for NAND in the SSD market is quite small compared to the plethora of mobile device applications.

I would guess that Samsung is putting all most all of their NAND in cell phones and tables, and selling the other stuff for other's peoples gear. Notably, I'm not aware of any SSD using Samsung NAND at the moment, aside from older drives that may or may not still be manufactured (like the second brand new Vertex Turbo I have in the mail).

Samsung must be making huge volumes of NAND -- it's just going into everything but SSDs (save their own). I wonder how many drives Samsung ships, as they are OEM for many manufacturers and their consumer and enterprise drives are probably a only tiny fraction of the whole.

Actually, I'd be surprised if more than 10% of Samsung's NAND goes into SSDs, so perhaps the SSDs do get the best of the best from that total.

[Dis07]

I thought the 830 was using 166MT Toggle mode. (BUT I just looked and supposedly it uses 133MTs)

But Samsung started shipping Toggle 2.0 400MTs NAND last year, in 64gbit packages.

http://www.toshiba.com/taec/Catalog/...lineid=1735086

Toshiba is working with JEDEC on a new standard for the most advanced high-performance NAND flash memory, a DDR NAND flash with a 400Mbps3 interface. This next generation Toggle-Mode DDR NAND 2.0 is targeted to provide a three-fold increase in interface speed over Toggle DDR 1.0, and a ten-fold increase over the 40Mbps single data rate NAND that is widely used today.

[Christopher]

Samsung and Toshiba both are making it, Samsung had a press release about it some time ago. But that doesn't mean it's ready for prime time yet.

[kaktus1907]

is there any solid info that's on Toshiba server that says so whether it's 5000 or 3000 P/E is for Toshiba's 24nm Toggle NANDs ? how can we be sure from a random blog post??

BTW i think latest Corsair Performance Pros are shipping with 24nm nands not like the ones on reviewers say 32nm..

for example Corsair P128 Pro with TH58TEG7D2HBA4C 24nm NANDs, production date is 2012 8th week..
http://www.coolaler.com/showthread.p...98%E9%80%9FSSD

and pretty much all reviewers Corsair P256 Pro with TH58TVG8D2FBA89 32nm NANDs, production date is 2011 41th week..
http://www.xbitlabs.com/articles/sto...d_2.html#sect1


also i found Toshiba has 2 types of 24nm NANDs.. one is H and other is G coded i couldnt find any more info about.. maybe one is 3000 P/E and other 5000 P/E ?

http://www.mediafire.com/?l5wjai3l22vuvsp (page 5)

[kaktus1907]

as AnandTech confirms Corsair Performance Pro uses 24 nm Toggle Nands(TH58TEG8D2HBA8C) at least newest ones..

http://www.anandtech.com/show/5785/c...256gb-review/2