backtop


Print 16 comment(s) - last by ipay.. on Jan 25 at 8:03 AM


  (Source: Jason Mick/DailyTech LLC)
DDR4 should land for enthusiasts as early as Q3 2014

When people think memory, or flash storage, most think of Samsung Electronics Comp., Ltd. (KRX:005935) (KRX:005930), the world's largest DRAM and NAND flash memory producer.  By contrast, unless you are a tech investor or an enthusiast you may not have heard of or may know little about Micron Technologies Inc. (MU).  That's understandable given that the American chipmaker has no flashy branded tablets or smartphones to raise its brand visibility in the eyes of the consumer.
 
Micron, for its part, is content with being a quiet leader of the computer industry.  Its position as a dominant power was solidified by the mid-2012 acquisition of Japan's Elpida Memory Inc. (TYO:6665).  Elpida owned several large fabs worldwide, including a fab in the Hiroshima area.  Last year Micron finalized that transaction, just in time to capitalize on a surge in memory spot prices.
 
I. Laser Focus on Memory
 
Today, Micron is a solid second in both the DRAM and NAND markets, outpaced only by Samsung in volume sales.  In its recently announced Q1 2014 earnings, it blew away analyst estimates of 44 cents per share profit, posting a profit of 77 cents per share ($881M USD). 
 
Just a year ago after posting $217M USD loss, Micron had promised it was turning the corner of a turbulent period in the industry, one that sunk many memory producers.  One thing's for certain -- Micron delivered on its promise in 2013.
 
And while the bullish profitability may lull slightly as production at SK Hynix Inc. (KRX:000660) resumes after last year's factory fire and major damage which triggered supply shortages, Micron remains well prepared to continue its strong profitability this year.

DDR stack
Competitors' shortages have driven the world's second largest DRAM producer to impressive profits. [Image Source: Jason Mick/DailyTech LLC]

Key to Micron's success is a laser focus on all things memory -- and only memory.  Selling to every niche from embedded (automotive, etc.) to mobile to console gaming to PCs, Micron has far less distractions than Samsung, even if it must share a piece of the ultimate device markup with various partners who buy its chips.
 
II. DDR4 Reaches Volume Production
 
Another key to Micron's fortunes is technological leadership.  Micron is second to none in terms of memory technology.  Even as Micron moves ahead on work on future successors to capacitor/transistor-based dynamic random access memory (DRAM), it's blazing the trail with DRAM's fourth generation technology, DDR4 (fourth generation double data rate) memory.

Micron DDR4
Since 2002 Micron has pushed DRAM through four generation of technology.
[Image Source: Micron]

Micron has been sampling the next-generation faster, lower-power memory since May 2012.  Last year at the 2013 Consumer Electronics Show (CES 2013) Micron tempted us with the allure of DDR4, "officially" announcing the future product and showing a demo running in the wild on DDR4.  Then again at the 2013 Intel Corp. (INTC) Developer Forum (IDF 2013), it again teased us.  Many asked when real product would actually arrive.
 
According to Jeremy Mortenson, a senior DRAM product manager with Micron's Crucial brand, 2014 will be the year when DDR4 finally arrives for both businesses and possibly for enthusiasts, as well.
 
Volume production is expected to begin this spring, as we approach Q2 2014.  Initially only enterprise buyers will be able to get their hands on the bleeding edge memory modules, which will receive Crucial branding and clock speeds of around 2133 MHz or 2400 MHz.  The first DDR4 volume production will likely be mostly sourced out of Micron's U.S. production facilities, primarily its major Boise fab.

DDR4

In the fall, Micron says it should be set to deliver Crucial Ballistix-branded DDR4 memory to enthusiasts.  This is exciting news, as some previously claimed that wouldn't be the case, such as PC World who in Nov. 2013 wrote, "DDR4 memory won't show up in PCs, tablets until 2015".
 
DDR4 indeed appears poised to land for consumers with cash this year, not next.  While Advanced Micro Devices Inc.'s (AMD) DDR4 support may not land until next year, so that leaves it up to Intel to release the 14 nanometer die shrink of Haswell, Broadwell, a chip which packs a compatible memory controller unit.
 
Intel CEO Brian Krzanich already blew apart the false rumor of Broadwell being bumped to 2015, stating in October's Q3 2013 earnings call that there were issues with the die shrink, but they were less severe than rumored:

We continue to make progress with the industry's first 14nm manufacturing process and our second generation 3D transistors. Broadwell, the first product on 14nm is up and running as we demonstrated at Intel Developer Forum, last month. While we are comfortable with where we are at with yields, from a timing standpoint, we are about a quarter behind our projections. As a result, we are now planning to begin production in the first quarter of next year. It was simply a defect density issue.

IDF 2013
Mass production of Broadwell was bumped from Q4 2013 to this quarter.
[Image Source: Jason Mick/DailyTech LLC]

KitGuru reported this week even more good news, saying that sources at Intel indicated that in Q3 Intel might launch Broadwell chips for enthusiasts in limit quantities. The company will stick to Haswell as its mass-volume seller until Q4 2014, when Broadwell fully takes the reins.
 
III. DDR4 Will Reach Consumer Hands Later This Year
 
Why should you care about DDR4?
 
First, it is expected to drop power consumption on the mobile end.  DDR3's base voltage requirements were 1.5 V.  Low power modules (so called DDR3L or LPDDR3) managed to trim this to 1.35 V.
 
DDR4 starts at 1.2 V, lower than even DDR3L.  Its low power variant is expected to further trim the voltage to 1.05 V, about two-thirds the voltage of desktop DRAM.  
 
With the introduction of DDR (first generation), memory modules gained the power to turn themselves off when not in use.  The cost, to this day, has been the inclusion of a reference voltage, commonly referred to as Vref or VrefDD.  With DDR4 this signal line is eliminated altogether -- a switch from CTT (Center Tab Termination) to POD (Pseudo-Open Drain).  This both reduces noise issues that Vref sometimes suffered and reduces the idle power consumption.
 
[Image Source: Micron]

The cost is that the system must "train" itself to the voltages, adding to the complexity of creating DDR4 modules.
 
Most of DDR4's other benefits come on the performance side.  In addition to the error-correcting code (ECC) that DDR3 added, DDR4 adds cyclic redundancy checks (CRC) (the same CRC as GDDR5, from what I hear) and parity checks.
 
Other improvements that allow faster performance include per DRAM addressability for higher performance and the ability to run in "gear down" mode, which increases the speed of the DQ (data) pins, by decreasing the speed of the CMD/ADD (command and address) signals -- useful for large data transfers at high rates.
 
Where DDR3 topped out at 30 GB/s (officially, at least) at DDR3-1866 (the 1866 parts is the Mbps data transfer rate of a single memory module, DDR4 runs from 34 GB/s (DDR4-2133) to 51.2 GB/s (DDR4-3200).
 
Last, but not least the spec density is also higher, allowing up to 8 Gb modules on a single ranking.  Long story short, you may have seen 8 Gb DDR3 parts, and this used multiple internal chips.  If DDR4 designers like Micron use similar technology they should be able to produce 16 Gb modules.

Single Sided SoDIMM
[Image Source: Micron]

DDR4 (more or less) doubles its predecessors' density, which should prove particular useful for compact form factors such as single-sided SODIMMs (tablets, ultrabooks).
 
IV. No Reason to Cry About DDR4, Adoption Timeline Follows DDR3's
 
With Broadwell soon upon us, Micron's roadmap for DDR4 makes a lot more sense and prophecies of no DDR4 in 2015 -- and some members of the media's tirades regarding Micron's early announcements -- appear unfounded.  Something that Adam Kaufman (a product marketing manager at Micron's Lexar brand) said to me stood out in that regard.
 
Speaking about another technology, he said, "The technology is there, we're just waiting on adoption."
 
When it comes to many technologies, such as CFast (a rival technology to XQd for 4K video storage; the technology Mr. Kaufman was referring to) and DDR4, Micron often does its homework well in advance, and like Samsung, it isn't afraid to lay its cards on the table.
 
Micron has received some flak for taking two years to bring DDR4 to market (see VR-Zone's "Crucial is playing the press with its DDR4 hype").  But in the grand scheme of things, the two-year delay really isn't anything to get upset about.

DDR4
Standard DDR4 DIMMs (top) branded as Crucial Ballistix sticks, will become available to consumers later this year. [Image Source: Jason Mick/DailyTech LLC]

Back in 2006 Micron became only the second company to begin sampling DDR3 chips at the time, the culmination of a half decade of development work, or so.  (Micron was sampling a 78 nm, 1 Gigabit (Gb) chip; the first DRAM samples came from Elpida and were 512 Mb (Megabit).)  Lo and behold Crucial only delivered DDR3 to market by Oct. 2008, as they had to wait on the Intel X58 chipset (which accompanied the 45 nm Nehalem CPU product).  The crossover point for sales of DDR3 didn't occur until Q2 2010; the crossover point for bits in the wild didn't appear until sometime very late in 2010 or early 2011.
 
A half-decade later, we see DDR4 following a similar timetable, driven by largely the same players.  There's about a half decade that goes into defining the initial implementation of these new memory technologies, another couple years of sampling, and another couple of years before it becomes the dominant memory technology.  You're looking at about a decade's worth of design work to take a memory technology from the drawing border to master of the market.

DDR4
Micron sees 2015 as the year when the full family of DDR4 form factors (pictured) becomes available. [Image Source: Jason Mick/DailyTech LLC]

Micron sees next year -- DDR4's second year on the market -- as the year when commodity DDR4 products, including modules from Micron's third party partners like Taiwan's Transcend Information Inc. (TPE:2451) will become widely available. Further, it expects 2016 to mark the crossover point where the number of bits of DDR4 in use in the wild passes the number of DDR3 bits in use worldwide.
 
But both DDR3 and DDR4 should share some gains such as the coming availability of 8 Gb x 32 DRAM (DDR3L-RS; L="low-voltage", RS="reduced standby") modules.  The '32' part stands for the bit width per chip on your memory sticks -- more bits generally means faster performance.  Again, the DDR3 relies on tricks to achieve 8 Gb modules (specifically, using two linked chip ranks per module), while the DDR4 inherently is more dense.
 
Micron is also working with Intel to finalize the somewhat waylaid XMP 1.3 standard.
 
V. CFast 2.0 -- Show me the Product!
 
While DDR4 is clearly the biggest thing on Micron's roadmap for 2014, its Lexar brand is also pushing another next generation standard, CFast 2.0.
 
CFast isn't exactly new -- it's been kicking around since 2008.  The idea is to use serial ATA (Advanced Technology Attachment) (SATA) to access the card at a much faster rate than current compact flash cards, which are accessed via parallel ATA (PATA).
 
At the show Lexar showed off a 3333X CFast 2.0 card, whose form factor is similar to CompactFlash, its technology precursor.  Capacities ranged from 32 GB up to 256 GB.

CFast 2.0
CFast 2.0 cards in the wild [Image Source: Jason Mick/DailyTech LLC]

For years the new card format was supposedly just around the corner.  However, today there are still no devices that currently support these cards.
 
The main backers of CFast 2.0 is Canon Inc. (TYO:7751) and Denmark-based premium professional photography camera-seller Phase One.  Returning to the previous quote by Mr. Kaufman -- Canon and Phase One have yet to make a camera that supports the format.
 
Lexar is playing both sides of the fence.  It's also selling XQD cards, a format that Sony Corp. (TYO:6758) and Nikon Corp. (TYO:7731) are pushing.
 
From a spec perspective XQD -- which uses the PCI Express (PCI-E) bus -- is inferior to CFast 2.0.  XQD can only push a maximum of 500 MBps (megabytes per second), and Lexar's just-released product is achieving rates of 125 (write)/197 (read) MBps in the wild.  The fastest XQD will allow is 500 MBps.  By contrast Lexar is proclaiming that its CFast 2.0 cards -- while more expensive -- already are achieving 500 MBps.  The spec is rated all the way up to 600 MBps.

CFast 2.0
CFast 2.0 uses the SATA bus and can reach speeds of 600 MBps.
[Image Source: Jason Mick/DailyTech LLC]

But where XQD has somewhat of an advantage is in that Sony and Nikon have announced/released actual camera hardware that supports the high-speed new format.  They certainly aren't cheap, though -- a 32 GB XQD card will cost you $220 USD, while the 64 GB version is a "bargain" at $380 USD (on Amazon.com, Inc.'s (AMZN)) storefront.

Lexar's Kaufman hints that the ultimate winner of divisive CFast 2.0 vs. XQD war may be good old SD.  The high speed variant of SD -- SDXC (Secure Digital eXtended Capacity) is expected to reach speeds of 312 MBps via the second generation Ultra High Speed (UHS) bus (UHS-II), as defined in the 2011 issued "version 4.0" specification [PDF].  That should allow SDXC cards to be used for 4K video.

Given the availability of cheap SDXC readers that use USB 3.0 or Thunderbolt ports, SDXC could trump both its PCI-E (XQD) and SATA (CFast 2.0) competitors if it is more broadly adopted.

Lexar currently offers a wide range of SDXC cards ranging from 16 GB to 256 GB, with speeds for Class 6 cards rated at 90 MBps (reads), 45 MBps (writes), or roughly half the speed of its current XQD stock.  But the prices are much lower; you can get a 256 GB SDXC 600x card for $450 USD.

So currently SDXC is about $1.75 USD/GB of capacity, while the twice-as-fast XQD is about $5.93 USD/GB (nearly four times as much).

Micron and its Lexar brand seemed supportive of CFast and SDXC over XQD in the long run, given spec advantages, but ultimately they're format agnostic.  Whatever memory card format wins, they'll be selling NAND flash storage products for it.

VI. Workflow Reader and Long-term Memory Roadmap

To that end, one last little interesting addition from Lexar is the Professional Workflow Reader Solution.  Continuing on swappability -- a hot trend at CES 2014 -- the Workflow Reader allows hot-swappable USB 3.0 reader modules.

Originally announced in September, the bay retails for around $72 USD (at present), while basic SD/SDXC and CompactFlash reader modules will cost you ~$24 USD a piece. 

Professional workflow reader
 The Professional Workflow Reader [Image Source: Jason Mick/DailyTech LLC]

An XQD reader is slightly pricier at $44 USD.  Given that the price of a standalone XQD reader is $42 USD (from Sony), hopefully that price will drop a bit.  But in the long term, the Workflow Reader seems like a smart idea in terms of future-proofing users from memory card format wars.
 
Stay tuned in months ahead, 2014 should be a big year for both memory (DRAM) and flash storage (NAND).  We'll continue to keep you up to date on the rollout of DDR4 and CFast 2.0, as well as other bleeding edge memory technology.
 
Looking ahead, based on the DDR2, DDR3, and (now) DDR4 timetable, expect Micron to start sampling a next-generation memory technology -- possibly MRAM -- around 2018, with mass-production sometime around 2020.  MRAM -- which can double as nonvolatile storage -- may replace NAND flash storage as well, offering unified device memories.
 

MRAM alliance
Micron and Hitachi are working together on a next-generation MRAM-based product.
[Image Source: DailyTech]

Until then, expect DDR4 to slowly become the dominant technology, inheriting the performance crown from DRAM, and expect faster and faster NAND chips to make their way to market in XQD, CF2.0, and SDXC UHS-II cards.


Comments     Threshold


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

DDR4
By techxx on 1/20/2014 1:28:46 PM , Rating: 3
Amazing how much DDR advances every few years. Next build will definitely be sporting DDR4. 8)




RE: DDR4
By ClownPuncher on 1/20/2014 3:23:09 PM , Rating: 2
Yea, I'll be doing mine in 2014/2015 too. I don't see much reason to build a brand new rig today.


RE: DDR4
By StevoLincolnite on 1/20/2014 6:08:53 PM , Rating: 2
I'll probably maketh the jump when my 3930K begins to struggle and do a system overhaul.

Faster Ram though... Would be a boon for AMD's APU's.


RE: DDR4
By tayb on 1/20/2014 6:53:38 PM , Rating: 2
It depends on when these sticks will actually be available and how much they will cost. They are talking about enterprise roll out in Q3 and enthusiast roll out in Q4. To me that means prices are going to be sky high until maybe 2016.

I'm sold on the performance potential but at the outset the biggest benefits are going to be density and power efficiency. I wouldn't be surprised if the first sticks don't perform any better than the highest end DDR3.

I'm looking to upgrade in February and so far the allure of DDR4 hasn't changed my mind. By the time I'm ready to upgrade again in 2016 perhaps DDR4 will be mainstream and price competitive with DDR3 of today.


RE: DDR4
By ritualm on 1/21/2014 5:50:44 PM , Rating: 2
quote:
To me that means prices are going to be sky high until maybe 2016.

Give it 2-3 quarters after enthusiasts get theirs. Enterprise sales are a major profit center, while the enthusiast community isn't satisfied with anything but the latest and greatest; combined, these two groups literally subsidize the higher initial production costs for everyone else.

If I'm building a gaming desktop in Q3/Q4 this year, I'm going to future proof it as much as possible. Cost goes out of the window, as I'm not interested in a cheaper part if there is no more substantial improvement in the years ahead.


RE: DDR4
By purerice on 1/24/2014 2:21:53 PM , Rating: 2
You're probably right but here's the catch:

Reduced demand = future product delays to cover development costs. Future product delays = reduced demand.
Vicious circle for Intel's Broadwell/Skylake/Cannonlake.
DDR4 won't be in Intel desktops until at least H2 2015.

AMD's Steamroller is like Intel's Netburst. You won't want AMD in your rig (if speed is your goal) until at least Q1 2016.

DDR4 will be available Q3 2014 but there will be no enthusiast/consumer machine to put it in.


Whoa
By Tequilasunriser on 1/20/2014 6:56:22 PM , Rating: 2
I very well might end up skipping the entire generation of DDR3 by the time I'm ready for my next build.

Current build has an Phenom II 940BE processor, 6GB of DDR2, and an HD6850 GPU.




RE: Whoa
By Silver2k7 on 1/23/2014 12:52:30 PM , Rating: 2
lol also have DDR2 8GB, would be nice to upgrade but feels silly to get on the DDR3 train just before DDR4 hits the market. Would be nice to see some DDR4 compatible CPU's and motherboards soon.


Not so laser focused
By ipay on 1/25/2014 8:03:57 AM , Rating: 3
"Key to Micron's success is a laser focus on all things memory"

your having a laugh surely, as you well know that Wide IO was ratified long before JEDEC slipped in this so called DDR4 (to milk the markets)that's in fact slower and has higher latency than the far better low power WIO2 and its sibling HMC at 7 to 15 times the data throughput of DDR4 right now.

WIO is a 512bit 4xDDR3 at 266Mhz single module/IO channel.

"Wide I/O 2 has a four-channel architecture, but the channels have a higher bandwidth of 6.4GBps and have 64 I/Os per channel, yielding an overall 25.6GBps data rate.

Later Wide I/O 2 devices will have up to 128 I/Os per channel and a data rate of 51.2GBps . Wide I/O 2 also opens the possibility of a 1,024-bit interface to memory."

"The first Wide I/O 2 devices will probably be 2.5D interposer-based architectures. These take more space than 3D die stacks, but 2.5D chips also have fewer thermal problems, more flexibility to rework connections, and lower costs. Wide I/O 2 (3D)is expected to go into mass production in 2015."

"Hybrid Memory Cube - Low Power, Blazing Fast

Want to go really, really fast - say, up to 320GBps ? And while we're at it, how about with 70% less energy per bit than DDR3 and 90% less board space than today's RDIMMs ? Then take a look at the Hybrid Memory Cube (HMC), which is expected to go into mass production in 2014 ."

please explain exactly why the so called industrial server markets would pay for DDR4 at best x2 the speed of the very best DDR3 today, over the HMC at 7x DDR4 data throughput in 1 or 2 years.

or for that matter why the masses of low power mobile markets or even desktop at that time would pay for DDR4 over the far better lower latency, lower power and far higher data throughput of WIO is beyond me...

as for your MRAM coverage , its also interesting you didn't mention that its already commercially here today from Everspin/Freescale

https://www.youtube.com/watch?v=kTKivwuzvNs#t=125 in production for some 5 years already on 90nm with the same speeds and data throughput as DDR3...
http://www.everspin.com/PDF/press/Buffalo_Memory_R...

so i ask you where exactly is Micron's laser focus in providing these 3 form's of ram today

http://www.cadence.com/Community/blogs/ii/archive/...




Innovation over Iteration
By EricMartello on 1/20/14, Rating: -1
RE: Innovation over Iteration
By JasonMick (blog) on 1/20/2014 2:41:30 PM , Rating: 3
You can't have the innovation without the iteration. I've covered technologies such as MRAM extensively too... e.g.

http://www.dailytech.com/Coalition+of+20+Tech+Firm...

... the reason I cover both is that you never know which innovations will turn to actual product and when. I remember back in 2004 interning at Lear Corp. and being tasked with preparing a presentation for the group manager on this "new display technology" -- OLED. It sounded great at the time, I even went above and beyond and got some rep from an early producer in to talk about it.

It would take probably another five years before the tech would start showing up in large quantities in the small display market (e.g. cameras, etc.), even longer before it was picked up in full size screen.

By the same token even iteration can cause big changes, if the iteration is big enough. For example the advent of mobile GPUs and CPUs is absolutely iterative in many regards, as is the path of display and memory tech in the mobile space. But it's enabled devices that have changed our lives -- both for the better and worse -- such as tablets and smartphones.

Iteration, isn't always bad.

DDR4 is arguably a large and helpful leap forward, although I'm still looking forward to unified memory technologies like MRAM most of all.


RE: Innovation over Iteration
By Sivar on 1/20/2014 6:17:12 PM , Rating: 2
There are two kinds of scientific progress: the methodical experimentation and categorization which gradually extend the boundaries of knowledge, and the revolutionary leap of genius which redefines and transcends those boundaries. Acknowledging our debt to the former, we yearn, nonetheless, for the latter.

--Academician Prokhor Zakharov, "Address to the Faculty" (Alpha Centauri)


RE: Innovation over Iteration
By EricMartello on 1/21/2014 4:43:20 PM , Rating: 1
quote:
You can't have the innovation without the iteration.

Iteration, isn't always bad.

DDR4 is arguably a large and helpful leap forward, although I'm still looking forward to unified memory technologies like MRAM most of all.


True enough, but that doesn't make DDR4 more interesting than DDR3 in my view. Not trying to bash your article since you write a lot of good ones, but I don't see DDR4 as doing anything much different (or unexpected) than DDR3 - it's not like the transition from EDO to DDR, where DDR was actually an innovation that effectively doubled bandwidth and it was interesting to know the inner workings that allowed it to be an improvement.

I think it's time that they stopped with the whole local bus paradigm and went to a unified system interconnect. Get rid of all these silly standards and interfaces like USB, thunderbolt, SATA, PCI and the "bridges" that allow them to talk to each other just make one interconnect to rule them all - which all components and peripherals connect to and share.

Brought this up a while ago and apparently there is some work going on to this end, but we're still clinging to some archaic designs that really just need to be retired...but that is something I find interesting.


RE: Innovation over Iteration
By haukionkannel on 1/20/2014 2:55:13 PM , Rating: 2
Allso you have to remember that mobile devices is the most important part of computer technology (for normal consumers) so any product that reduces less power is very important!
In desktop... not so much, but desktops are not important any more to most peoples. It is definitely for me, but I am in very small minority. Even in desktop usage some more speed can be really important with APU product that may become the main products to all CPU makers! Even Intel pack more and more GPU cores in each new CPU that they make and develop, so even a little bit faster ram can be grusial! At least more important that is used to be.


RE: Innovation over Iteration
By gppinky on 1/20/2014 3:19:51 PM , Rating: 4
The DDR5 you are referring to is GDDR5, which is a graphics branch memory which is based on DDR3.

DDR4 is system memory which is the next generation of DDR3.

They are completely different branches.


RE: Innovation over Iteration
By extide on 1/20/2014 5:34:32 PM , Rating: 2
His first mention of DDR5 is a mistake, it was supposed to be something like, "DDR3 is now DDR4 and next year we will have DDR5."

Obviously DDR5 will come eventually, (not next year) but you can see his point.

He isn't talking about GDDR of any sort.


“So far we have not seen a single Android device that does not infringe on our patents." -- Microsoft General Counsel Brad Smith




Latest Headlines
4/21/2014 Hardware Reviews
April 21, 2014, 12:46 PM
4/16/2014 Hardware Reviews
April 16, 2014, 9:01 AM
4/15/2014 Hardware Reviews
April 15, 2014, 11:30 AM
4/11/2014 Hardware Reviews
April 11, 2014, 11:03 AM










botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki