Taking an Accurate Look at SSD Write Endurance
Last year, I posted a rebuttal to a paper describing the future of flash memory as ‘bleak’. The paper went through great (and convoluted) lengths to paint a tragic picture of flash memory endurance moving forward. Yesterday a newer paper hit Slashdot – this one doing just the opposite, and going as far as to assume production flash memory handling up to 1 Million erase cycles. You’d think that since I’m constantly pushing flash memory as a viable, reliable, and super-fast successor to Hard Disks (aka 'Spinning Rust'), that I’d just sit back on this one and let it fly. After all, it helps make my argument! Well, I can’t, because if there are errors published on a topic so important to me, it’s in the interest of journalistic integrity that I must now post an equal and opposite rebuttal to this one – even if it works against my case.
First I’m going to invite you to read through the paper in question. After doing so, I’m now going to pick it apart. Unfortunately I’m crunched for time today, so I’m going to reduce my dissertation into the form of some simple bulleted points:
- Max data write speed did not take into account 8/10 encoding, meaning 6Gb/sec = 600MB/sec, not 750MB/sec.
- The flash *page* size (8KB) and block sizes (2MB) chosen more closely resemble that of MLC parts (not SLC – see below for why this is important).
- The paper makes no reference to Write Amplification.
Perhaps the most glaring and significant is that all of the formulas, while correct, fail to consider the most important factor when dealing with flash memory writes – Write Amplification.
Before geting into it, I'll reference the excellent graphic that Anand put in his SSD Relapse piece:
SSD controllers combine smaller writes into larger ones in an attempt to speed up the effective write speed. This falls flat once all flash blocks have been written to at least once. From that point forward, the SSD must play musical chairs with the data on each and every small write. In a bad case, a single 4KB write turns into a 2MB write. For that example, Write Amplification would be a factor of 500, meaning the flash memory is cycled at 500x the rate calculated in the paper. Sure that’s an extreme example, but the point is that without referencing amplification at all, it is assumed to be a factor of 1, which would only be the case if you were only writing 2MB blocks of data to the SSD. This is almost never the case, regardless of Operating System.
After posters on Slashdot called out the author on his assumptions of rated P/E cycles, he went back and added two links to justify his figures. The problem is that the first links to a 2005 data sheet for 90nm SLC flash. Samsung’s 90nm flash was 1Gb per die (128MB). The packages were available with up to 4 dies each, and scaling up to a typical 16-chip SSD, that only gives you an 8GB SSD. Not very practical. That’s not to say 100k is an inaccurate figure for SLC endurance. It’s just a really bad reference to use is all. Here's a better one from the Flash Memory Summit a couple of years back:
The second link was a 2008 PR blast from Micron, based on their proposed pushing of the 34nm process to its limits. “One Million Write Cycles” was nothing more than a tag line for an achievement accomplished in a lab under ideal conditions. That figure was never reached in anything you could actually buy in a SATA SSD. A better reference would be from that same presentation at the Summit:
This shows larger process nodes hitting even beyond 1 million cycles (given sufficient additional error bits used for error correction), but remember it has to be something that is available and in a usable capacity to be practical for real world use, and that’s just not the case for the flash in the above chart.
At the end of the day, manufacturers must balance cost, capacity, and longevity. This forces a push towards smaller processes (for more capacity per cost), with the limit being how much endurance they are willing to give up in the process. In the end they choose based on what the customer needs. Enterprise use leans towards SLC or eMLC, as they are willing to spend more for the gain in endurance. Typical PC users get standard MLC and now even TLC, which are *good enough* for that application. It's worth noting that most SSD failures are not due to burning out all of the available flash P/E cycles. The vast majority are due to infant mortality failures of the controller or even due to buggy firmware. I've never written enough to any single consumer SSD (in normal operation) to wear out all of the flash. The closest I've come to a flash-related failure was when I had an ioDrive fail during testing by excessive heat causing a solder pad to lift on one of the flash chips.
All of this said, I’d love to see a revisit to the author’s well-structured paper – only based on the corrected assumptions I’ve outlined above. *That* is the type of paper I would reference when attempting to make *accurate* arguments for SSD endurance.
Subject: Storage | January 3, 2013 - 12:22 PM | Jeremy Hellstrom
Tagged: tlc, Samsung 840, 500gb, ssd
As the lifespan of flash memory in SSDs has become a topic of concern for many users, it is nice to see that the Samsung 840 500GB has a lifespan of some 14 years assuming a daily write load of 10GB. Since most users do not write 10GB to a drive day in and day out, that estimate is probably on the low end. If that doesn't have you excited then consider the cost of the drive, at $350 it is much lower than the $1/GB mark most other SSDs are at. There are some trade offs however, [H]ard|OCP saw comparatively slow extended write speeds though the read speeds were higher than the 256GB model. When you consider this drive do keep in mind that it is still going to be faster than a platter drive even when it is working on a task that other SSDs might do slightly faster.
"Samsung has released the first TLC NAND equipped SSDs into the market, creating the lowest price points we have witnessed for SSDs bringing large capacity SSDs within reach for average users. Today we test the 500GB TLC Samsung 840 Series SSD to test the performance in steady state of a large capacity TLC Solid State Drive."
Here are some more Storage reviews from around the web:
- OCZ Vector 256 GB @ techPowerUp
- OCZ Vertex 3 240GB SSD VTX3-25SAT3-240G Review @ PCSTATS
- OCZ Vector 512GB SSD @ Tweaktown
- OCZ Vector 256GB @ eTeknix
- Kingston SSDNow V300 120GB SSD Review @ Legit Reviews
- Silicon Power Slim S70 120GB 7mm SSD Review @ ModSynergy
- Western Digital Red 4TB HDD @ TechwareLabs
- Western Digital Red (WD30EFRX) 3 TB @ TechARP
- WD Black 4TB Hard Drive Review @ Techgage
- Western Digital Black 4TB @ Legion Hardware
- Western Digital Sentinel DX4000 NAS @ X-bit Labs
- Promise Pegasus J4: 4-bay, 2.5-inch Thunderbolt storage @ Hardware.info
- NETGEAR ReadyNAS Duo v2 NAS Server Review @ Techgage
Subject: Storage | November 20, 2012 - 12:48 PM | Jeremy Hellstrom
Tagged: Samsung, 840, tlc
As part of their review of the Samsung 840 250GB SSD, The Tech Report covers the specifics of the TLC flash memory which is used in the 840 series as opposed to the MLC we saw in the 830 series. As well they show off some of the capabilities of the control software, which Samsung has dubbed the SSD Magician utility. Of course from there the benchmarking begins which showed performance continually below the similarly priced 830 series which hurts the new SSDs on the price to performance chart. Overall they are hard pressed to recommend the drive over the previous models, not only because of the performance but also the shortened lifespan of TLC flash. As that flash technology matures we may see those concerns fade, as Allyn pointed out in his review.
"Samsung's 840 Series SSD combines a next-gen fabrication process with an extra bit per cell to lower the cost per gigabyte. We take a closer look at the implications and see how the drive stacks up against the competition."
Here are some more Storage reviews from around the web:
- OCZ Agility 4 256GB SSD Review @ Neoseeker
- Intel 330 Series 120 GB Solid State Drive Review @ Hardware Secrets
- SanDisk Extreme 480GB SSD @ Tweaktown
- OCZ Vertex 3 240GB SSD VTX3-25SAT3-240G Review @ PCSTATS
- Silicon Power Slim S70 240GB SSD @ Tweaktown
- Lexar JumpDrive S73 32 GB USB 3.0 @ techPowerUp
- Lexar JumpDrive 64GB USB 3.0 Flash Drive Review @ PCSTATS
- Silicon Power Firma F80 32GB USB 2.0 Flash Drive Review @ NikKTech
- Vantec NST-400MX-S3R NexStar MX Enclosure Review @ Pro-Clockers
- Super Talent USB3 Express RC8 100GB Flash Drive @ SSD Review
- ADATA DashDrive Elite 500GB USB 3.0 External Hard Drive @ Kitguru
- Patriot Memory Gauntlet 320 Wireless 2.5 Hard disk Enclosure @ Funky Kit
- WD My Passport Edge 500GB Portable Hard Drive Review @ Legit Reviews
Introduction, Specifications and Packaging
Last week, Samsung flew myself and a few of my fellow peers in the storage review community out to Seoul, Korea. The event was the 2012 Samsung SSD Global Summit:
At this event, Samsung officially announced their new 840 Pro, which we were able to obtain early under NDA and therefore publish in concert with the announcement. The 840 Pro was largely an incremental inprovement over their 830 Series. Newer, faster flash coupled with a higher clocked controller did well to improve on an already excellent product.
As the event closed, we were presented with the second model of the lineup - the 840. This model, sans the 'Pro' moniker, is meant more for general consumer usage. The first mass marketed SSD to use Triple Level Cell (TLC) flash, it sacrifices some write speed and long-term reliability in favor of what should become considerably lower cost/GB as production ramps up to full capacity. TLC flash is the next step beyond MLC, which is in turn a step after SLC. Here's a graphic to demonstrate:
Subject: General Tech | September 27, 2012 - 10:46 AM | Ken Addison
Tagged: tlc, ssd, Sea Islands, Samsung, PSU, podcast, nvidia, IOPS, Intel, evga, amd, 840 pro, 840, 1500W
PC Perspective Podcast #220 - 09/27/2012
Join us this week as we talk about the Samsung 840 Pro SSD, a 1500W PSU from EVGA, AMD GPU leaks, and more!
The URL for the podcast is: http://pcper.com/podcast - Share with your friends!
- iTunes - Subscribe to the podcast directly through the Store
- RSS - Subscribe through your regular RSS reader
- MP3 - Direct download link to the MP3 file
Hosts: Ryan Shrout, Josh Walrath, Jeremy Hellstrom, and Allyn Malvantano
Program length: 1:07:28
Podcast topics of discussion:
- Week in Reviews:
- 0:23:20 This Podcast is brought to you by MSI!
- News items of interest:
- 1-888-38-PCPER or email@example.com
- http://twitter.com/ryanshrout and http://twitter.com/pcper
Subject: General Tech | July 26, 2011 - 09:03 AM | Jeremy Hellstrom
Tagged: ssd, ocz, arm, tlc, sata 6Gps, Indilinx Everest
OCZ is never satisfied with the performance of their SSDs in general and their controllers specifically. After purchasing Indilinx to ensure that their controllers would be of high quality and designed to OCZ's specific needs, they've now been pushing Indilinx to improve on their controllers. That has lead to Everest, which has a dual core ARM processor and 400MHz DDR3 cache that can support up to 512MB. The controller is optimized for 8K writes which is perfect for the current flash utilized in SSDs. OCZ has also optimized the flash memory, developing Triple Level Cell (TLC) which has three layers as opposed to MLC which sports two. The controller will be backwards compatible, which is a good idea if OCZ wants to license the controller to other manufacturers, which makes sense as Everest should hit 200MT/s as compared to SandForce's current 166MT/s. There is more that this controller can do, click on over to The Register to read about it.
"OCZ is sampling a new flash controller that gives a picture of future solid state drives.
The company bought Indilinx for its solid state drive (SSD) controller technology in March this year and has now unveiled the Indilinx Everest controller platform.
It has a 6Gbit/s SATA III interface, a dual-core ARM processor and a number of enticing features, such as 3-bit multi-level cell (MLC) support. This is going to be called TLC, for triple-level cell, to distinguish it from today's MLC, which is 2-bit MLC."
Here is some more Tech News from around the web:
- Intel says it competes with Qualcomm not ARM @ The Inquirer
- Mozilla is developing a mobile operating system @ The Inquirer
- Running high-performance neural networks on a "gamer" GPU @ Ars Technica
- The Isostick @ Hack a Day
- Lawn warfare: Light Strike brings laser tag back home @ Ars Technica
- JMicron develops SATA 6Gbps controller IC for SSDs @ DigiTimes
- The TR Podcast 92: Fusion, the cloud, and dongles galore
- Sony Alpha NEX-C3 Review @ TechReviewSource
- Real World Labs And A.C.Ryan Joint Contest
Subject: Storage | April 27, 2011 - 07:06 PM | Allyn Malventano
Tagged: tlc, ssd, slc, ocz, mlc, flash
A while back, Intel and Micron jointly announced the beginnings of 20nm flash memory production, promising a 50% increase in die count per wafer (or a 50% reduction in per die production cost, depending on how you slice it). This shrink only did just that - shrink the die. Capacity remained at 64Gbit (8GB).
A few days ago IMFT also announced another way to shrink that die, but this time keeping with the now 'old' 25nm process. It turns out they have refined 25nm to the point where consumer-grade TLC flash can be produced. TLC is Triple-Level-Cell. While SLC (Single) holds 1 bit per cell, and MLC (Multi) holds two, TLC holds 3 bits per cell. Compared to the MLC 25nm dies, this gives a capacity increase without changing much else. IMFT, however, is happy with the 8GB 'sweet spot', so instead of jumping to a 12GB die of the same physical size, they are opting to instead shrink the current 25nm die to 131mm^2.
25nm TLC die, same 8GB capacity, but less area than the 25nm MLC die.
This gives Intel and Micron two options for ultimately reducing the price of flash - either by shrinking the process and getting more 8GB MLC dies out of a 20nm wafer, or by squeezing more bits into each cell of existing 25nm flash.
This is good stuff. Let's hope it gets even more SSD's into even more machines this holiday season.