backtop


Print 21 comment(s) - last by johnnyMon.. on Jul 2 at 1:56 AM


A peek at one of Micron/Intel's second generation NAND flash chips shows how tiny it is. The chips, available in SLC and MLC varieties have shrunk substantially and offer faster read/write speeds from the first generation.  (Source: DigiTimes)
New SSDs should be faster, more power efficient, and smaller thanks to die shrink

Die shrinks can be a wonderful thing.  A die shrink to 40 nm is helping AMD/ATI further its competitive edge over competitor NVIDIA in the desktop graphics market.  Die shrinks make consoles like the PS3 and Xbox more affordable and more profitable.  And both AMD and Intel have used die shrinks to deliver faster, more core CPUs over the last decade.

In the field of flash memory, die shrinks have been carried out at a frenzied pace.  In 2008, Micron hit 34 nm and Samsung hit 32 nm.  This year Toshiba/SanDisk caught up, hitting 32 nm.  The shrink dies brought with them performance improvements -- lower power consumption, smaller size, and faster access speeds.

Now, Micron is preparing to further reap the rewards of its own die-shrink campaign.  Micron's flash unit, co-owned by Intel, is reportedly readying new 34 nm NAND flash memory chips for the market.  After seeing relatively strong sales of its first generation, Micron will soon be releasing revised 16 and 32 Gigabit NAND chips. 

Both chips are much smaller, and available in Single Level Cell (SLC) and Multi-Level Cell (MLC) designs.  The SLC designs are the premium models, featuring faster write speeds, lower power consumption and higher cell endurance.  The MLC meanwhile will form the basis of budget offerings. 

The new MLC 32 Gb NAND chip is 17 percent smaller, while the 16 Gb chip is a tiny 84mm².  The tiny footprints should help to increase memory capacity of devices such as iPods, iPhones, as well as solid-state drives.

Brian Shirley, vice president of Micron’s memory group brags, "Our industry-leading NAND products are opening new possibilities for some of the world’s most popular consumer electronic devices.  With our new 16- and 32Gb NAND chips in mass production, we are enabling customers to design cost-effective, high-capacity storage in their small-form factor products, using less space and fewer die. In addition, the high-speed interface is ideal in the industry’s quest to continue to increase throughput performance for SSDs."

The chips will be used in a variety of Lexar memory products to be released in September.  The chips' ONFI 2.1 synchronous interface delivers transfer speeds up to 200 Mbit/s, compared to 40 Mbit/s with traditional NAND SLC.  This allows manufacturers to offer 6 Gbit/s SATA speeds, offering twice the throughput of standard 3 Gbit/s SATA speeds, according to Micron.



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

"Mb/s"...
By void5 on 6/30/2009 10:33:51 AM , Rating: 5
Do you really think that using "Mb/s" for "megaBYTES per second" and "Gb/s" for "gigaBITS per second" in the same article is such a good idea?

If you are too lazy to write "Mbyte/s" / "Mbit/s" - at least use "standard" abbreviations like "MB/s" for bytes and "Mb/s" for bits. While not perfect, this will be MUCH better than now.




RE: "Mb/s"...
By void5 on 6/30/2009 1:39:41 PM , Rating: 4
Further congratulations to Jason Mick or whoever has changed original article text. "Mb/s" has been replaced with "Mbit/s", so now instead of inappropriate abbreviation we have factual error because ONFI 2.1 bandwidth is 200 MegaBYTES per second (http://onfi.org/news-events/onfi-introduces-21-spe...


RE: "Mb/s"...
By icanhascpu on 6/30/2009 3:30:42 PM , Rating: 3
Why is this rated down?

An article on a TECH site that cant even get this shit straight is in SERIOUS lack.


So how do I read this?
By Golgatha on 6/30/2009 11:00:36 AM , Rating: 2
quote:
The chips will be used in a variety of Lexar memory products to be released in September. The chips' ONFI 2.1 synchronous interface delivers transfer speeds up to 200 Mbit/s, compared to 40 Mbit/s with traditional NAND SLC. This allows manufacturers to offer 6 Gbit/s SATA speeds, offering twice the throughput of standard 3 Gbit/s SATA speeds, according to Micron.


Does this mean they will have SATA 3 interfaces on the drive, or that the throughput of the drive will reach 6Gbit/sec? 600MB/sec sustained transfer rates with nearly no seek time would be insane!




RE: So how do I read this?
By VaultDweller on 6/30/2009 1:11:20 PM , Rating: 3
This article isn't about drives, it's just about the flash chips. Manufacturers that buy this flash for SSDs can use whatever interface they want for the drives. They're saying that it will be possible for drives to be created using their chips that can achieve 6Gbit speeds, not that any particular drive is doing so or will have the SATA 6 Gbit interface to enable it.


RE: So how do I read this?
By Souka on 6/30/2009 1:56:52 PM , Rating: 2
such improvements will push a new SATA standard...SATA-3...

This will occur as a single non-raid device will hit the limits of SATA2...

Good news, if I understand correctly, all the SATA interfaces are compatible...with the slowest on a chain being the speed which it operates. (eg, you can plug in a new super-fast SSD into your old SATA1 Dell on-sale/special CoreDuo pc and it'll run....)


RE: So how do I read this?
By PandaBear on 6/30/2009 3:13:57 PM , Rating: 2
That is only the transfer rate between the NAND and the controller. Most of the bottleneck is the processing time for read/write/erase in the NAND itself.

Running it at a very high frequency is bad for power consumption as well. There is a reason Toshiba/Sandisk/Samsung are not using ONFI serial interface: power consumption.


the whole article is screwed up.
By Randomblame on 6/30/2009 11:40:07 AM , Rating: 1
"In 2008, Micron hit 34 nm and Samsung hit 32 nm.
Now, Micron is preparing to further reap the rewards of its own die-shrink campaign. Micron's flash unit, co-owned by Intel, is reportedly readying new 34 nm NAND flash memory chips for the market."

so in 2008 micron made 32nm flash chips and now at the end of 2009 they are excited to announce that they have regressed to 34nm!




RE: the whole article is screwed up.
By poundsmack on 6/30/2009 12:40:39 PM , Rating: 2
"... Samsung hit 32 nm. This year Toshiba/SanDisk caught up, hitting 32 nm. "

Basic reading comprehension skills would indicate that Samsung, Toshiba, and Sandisk are using the 32mn process, while Micron is using the 34nm process. Mircon choose to go 34 around the same time Samsung went 32mn, so it isn't a regression as this is the smallest fab Micron has done to date (that has hit the market).


By PandaBear on 6/30/2009 3:16:47 PM , Rating: 2
Samsung, Toshiba and Sandisk are shipping 32nm with high yield. Micron started 34nm last year but didn't get the yield up until this year.


By PandaBear on 6/30/2009 3:15:18 PM , Rating: 3
Micron was having a very huge yield issue until recently, and the reason why Intel's SSD was so expensive is that the 34nm failed in the SSD usage, and Intel has to use the very expensive (by today standard) 50nm generation memory.


Bytes vs. Bits
By johnnyMon on 6/30/2009 1:17:20 PM , Rating: 2
Question: The article describes a 32Gb chip and shows what one looks like. Gb means gigabits, right? If so, then are 8 chips necessary to make a 32GB memory module? Or are there 9 chips so there is a parity bit used for error checking? If there are 8 or 9 chips, how do all of these fit onto a tiny little SD card? Thanks!




RE: Bytes vs. Bits
By PandaBear on 6/30/2009 3:20:02 PM , Rating: 2
The chip support 32Gb + overhead space for the page. The data encoded in the NAND is usually: header + data + ECC/checksum that is about 4 sectors to 16 sectors each page. Then you have overhead info for wear leveling and look up table, so a 32Gb chip will probably only provide 4000 MB (not 4096MB) of useful space to the file system. You can think of it as the hard drive's type of 4GB, not the DRAM's type of 4GB.


RE: Bytes vs. Bits
By johnnyMon on 7/2/2009 1:56:01 AM , Rating: 2
Thank you, PandaBear. This was helpful - I'll think about it this way from now on.


Still slow
By Belard on 7/1/2009 5:31:35 AM , Rating: 2
Compared to the PCIe SSD cards... but those cost $1500~4000.

But the data rates are 600MB/s to 1500MB/s

The DREAM drive:

500GB unit, $100: 600MB/s in both read & Write. Boots into Windows7 in 5 seconds.




RE: Still slow
By Jeffk464 on 7/1/2009 3:40:52 PM , Rating: 2
Well yeah, don't forget to mention longer laptop battery life.


RE: Still slow
By Jeffk464 on 7/1/2009 3:43:14 PM , Rating: 2
I'm thinking until they get cheap the solution might be a two drive setup. SSD for the OS and all your programs, and a spinning hard drive for data.


6Gb/s SATA | SSD
By Shig on 6/30/2009 10:28:04 AM , Rating: 1
So overkill for the average desktop PC user, yet I'm still going to get one. :)

I simply need my games to load fastar!




RE: 6Gb/s SATA | SSD
By icanhascpu on 6/30/2009 4:00:04 PM , Rating: 3
Most computers made in the last 10 years are "overkill" to the "average" PC user.

To people that come to this site, however, typically are not average, and opening a huge bottleneck is not an overkill.


More info here
By eilersr on 6/30/2009 12:15:03 PM , Rating: 2
EE Times has more background on the NAND scaling wars between Intel/Micron and Toshiba/Sandisk:

http://www.eetimes.com/news/latest/showArticle.jht...

More geared at an engineer's technology (process) focus, so not as much info on interfaces, etc., but good supplemental reading.




By riku0116 on 6/30/2009 12:34:31 PM , Rating: 2
quote:
New SSDs should be faster, more power efficient, smaller, and most importantly, cheaper thanks to die shrink


There, fixed it for you.




"My sex life is pretty good" -- Steve Jobs' random musings during the 2010 D8 conference




Latest Headlines
2/10/2012 Daily Hardware Reviews
February 10, 2012, 5:50 PM
2/9/2012 Daily Hardware Reviews
February 9, 2012, 11:54 AM
2/8/2012 Daily Hardware Reviews
February 8, 2012, 1:11 PM
2/7/2012 Daily Hardware Reviews
February 7, 2012, 12:23 PM










botimage
Copyright 2012 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki