The Seagate 600 & 600 Pro SSD Review
by Anand Lal Shimpi on May 7, 2013 8:00 AM ESTIf you had asked me back in 2008 who I thought would be leading the SSD industry in 2013 I would’ve said Intel, Western Digital and Seagate. Intel because of its commanding early lead in the market, and WD/Seagate because as the leaders in hard drives they couldn’t afford to be absent from the long term transition to SSDs. The days of having to explain why SSDs are better than mechanical drives are thankfully well behind us, now it’s just a question of predicting the inevitable. I figured that the hard drive vendors would see the same future and quickly try to establish a foothold in the SSD market. It turns out I’m really bad at predicting things.
Like most converging markets (in this case, storage + NAND), the SSD industry hasn’t been dominated by players in the market that came before it. Instead, SSDs attracted newcomers to the client/enterprise storage business. Not unlike DRAM, owning a NAND foundry has its benefits when building a profitable SSD business. It’s no surprise that Intel, Micron and Samsung are some of the more frequently discussed SSD vendors - all of them own (either partially or fully) NAND foundries.
Whether or not ownership in a foundry will be a requirement for building a sustainable SSD business is still unclear, but until that question gets answered there’s room for everyone to play in the quickly growing SSD market. This year, Seagate re-enters the SSD market with a serious portfolio. Today it not only announces two 2.5” SATA drives, including its first client-focused SSD, but also a 2.5” SAS product and a PCIe SSD solution.
The products that we’re focusing on today are the two 2.5” SATA drives: Seagate’s 600 and 600 Pro.
Architecture
The 600 and 600 Pro are both based on Link A Media Device’s LM87800 controller. The LAMD controller is the same as the one used in Corsair’s Neutron and Neutron GTX. Previous Seagate SSDs actually used a two-chip solution, with Seagate’s custom silicon controlling the host interface while Link A Media provided a NAND interface chip. The LM87800 is apparently a single chip integration of the earlier Seagate designs. The controller uses the drive chassis for cooling, with a thermal pad acting as an interface layer.
The firmware on the 600/600 Pro is unique to Seagate. It’s unclear whether or not Seagate has access to firmware source, but the solution is definitely custom (as you’ll see from the performance/consistency results). The LM87800 doesn't use any data de-duplication/compression and allegedly uses a DSP-like architecture.
The controller is paired with two DDR2-800 devices, with roughly 1MB of DRAM per GB of NAND storage. The high ratio of DRAM to NAND is common in drives with flat indirection tablets as we’ve come to notice. It’s a more costly (and potentially more power hungry) design decision, but one that can have tangible benefits as far as performance consistency is concerned.
Seagate 600 NAND/DRAM Configuration | |||||||||||
# of NAND Packages | # of Die per Package | Total NAND on-board | DRAM | ||||||||
480GB | 8 | 8 | 512GB | 512MB | |||||||
240GB | 8 | 4 | 256GB | 256MB | |||||||
120GB | 8 | 2 | 128GB | 128MB |
The LM87800 controller is a bit dated by modern standards, especially if we look at what is possible with Crucial’s M500. There’s no hardware encryption support and obviously no eDrive certification. Despite launching in 2013, the 600/600 Pro feature a controller that is distinctly 2012. Admittedly, that seems to be the case with most SSD makers these days. Everyone seems to be waiting for the transition to SATA Express before launching truly new controller designs.
Seagate has deals in place to secure NAND supply from both Samsung and Toshiba, although all of the 600 series will show up with 19nm Toshiba 2-bit-per-cell MLC NAND. The LM87800 controller features 8 NAND channels, and can access even more NAND die in parallel through interleaving.
59 Comments
View All Comments
Jestre - Tuesday, May 7, 2013 - link
2 requests:Can you incorporate the average power usage measured during the testing benchmarks, both "the Destroyer" and the 2011 benchmarks?
It would be nice to see the actual realistic measured power consumption averages for worst case tests like this. You could also list an average power with a reasonable downtime calculated based on a typical usage pattern and the actual time to complete.
Can you also list the time is take to complete these test for each drive,
Thanks for all you good work.
Jestre
Kristian Vättö - Tuesday, May 7, 2013 - link
At least I don't have the hardware for measuring power consumption over time (all I have is a simple multimeter, unless it has features I'm not aware of). Anand might but I'm guessing not as I'm sure he would have taken advantage of it. The problem is that such hardware is not exactly cheap, so it's a rather big investment for only one test (I'd love to have one though).lotharamious - Tuesday, May 7, 2013 - link
Another wonderful SSD review, Anand. Well done!Since I started reading your site over 11 years ago your reviews have been really outstanding and particularly so with SSDs.
I really love delving into the data presented, but you guys really need to get with the times on data plotting. Your charts can be somewhat confusing and downright frustrating to read.
When showing scatter plots please include an option for plotting more statistics of the dataset, e.g. trendline about the mean, standard deviation, etc. It's really difficult to see just how well the 800 is gaining performance towards the end of the test run without a noticable trendline for your eyes to follow along the graph.
In a chart, when displaying a line that corresponds to the unit that is currently being reviewed please draw that line ON TOP of the other references units' data lines. It's frustrating to try to compare the performance of different units when the one you really care about is blocked from view because the other 50 lines on the chart make a large enough noisy mess to not be able to tell right from left.
And just for fun, I really wish you guys used a javascript plotting library for your reviews. It would be awesome if you could click on different graphics cards in a graph to see the percentage performance gain you would get between the card you clicked on and all the other cards in the graph. I've seen this kind of stuff on other sites, and it would be amazing to have here.
Sorry for being so negative, because I still love reading this site every day. It's been incredible to see how you, the industry, and the drives themselves have morphed since 2008.
Here's to many more great reviews!
mike55 - Tuesday, May 7, 2013 - link
What causes the sharp drop in IOPS in the random write tests after so many minutes for SSDs? Is it because the drive has run out of empty blocks and is then doing read-modify-writes?mike55 - Tuesday, May 7, 2013 - link
Never mind, just realized my question was answered in the article.HisDivineOrder - Tuesday, May 7, 2013 - link
I imagine Seagate will tap its longterm partner, Samsung, to help it out with SSD controllers once LAMD is gone. Who did Samsung sell its hard drive business to? Seagate.Who did Seagate contract out for NAND from? Samsung.
Seagate'll go back to Samsung once it's time for a new controller that can handle SATAe.
kyuu - Wednesday, May 8, 2013 - link
While this is a decent drive, another source revealed that Seagate is trying to pull warranty shenanigans, limiting the warranty to 3 years OR 72 TB of writes (36 TB for the smaller drives), whichever comes first. Putting a "mileage" limit on the warranty is a first in the SSD space as far as I'm aware, and not something that should be supported. Definite pass on Seagate's SSDs unless and until they change their warranty terms.Kristian Vättö - Wednesday, May 8, 2013 - link
If you look at the warranty terms of other manufacturers, there is always a phrase that says warranty is invalidated if the SSD is operated outside of public specifications (and endurance is one one of them):Intel: "any Product which has been modified or operated outside of Intel’s publicly available
specifications"
http://download.intel.com/support/ssdc/hpssd/sb/1s...
Crucial: "The above warranties cover only defects arising under normal use and do not include malfunctions or failures resulting from misuse, abuse, neglect, alteration, problems with electrical power, usage not in accordance with product instructions"
http://www.crucial.com/company/termsofsale.aspx#li...
OCZ: "Improper use of product, Normal wear and tear"
http://ocz.com/consumer/support/warranty
Or at least that's how I interpret their warranties. In a nutshell, warranty only covers failures which are results of defects in materials or assembly - it's clear that endurance is limited and hence exceeding the specification means that a failure may not have been due to a defect.
daniel_mayes - Wednesday, May 8, 2013 - link
I would like to see consumer vs enterprise in a SMB server. Under a certain price point for example at $1,300 and lower you can get 3 x Intel 710 Series 100GB ($1,200), 5 x Intel DC S3700 100GB ($1,250), 4 x Seagate 600 Pro ($1,300), 5 x OCZ Vector 256 GB ($1,225), 5 x Corsair Neutron GTX 240 GB ($1,100), 5 x Samsung 840 Pro 256 GB ($1,200). Size doesn't matter as much as low latency and highest consistent performance. Since the consumer drives have more space over provision them at 25% and %50. That way we can see if more consumer ssd's at a higher provision rate are faster in a SMB environment than the enterprise drives. Yes I know the consumer ssd's will probably die faster but I'm sure most SMB's would rather pay half the cost of an enterprise drive and chance having to replace it twice as often.Jestre - Wednesday, May 8, 2013 - link
Kristian,is $59.95 with free shipping in your budget?
See http://www.rc-electronics-usa.com/ammeters/dc-amp-...
Not a sweet and an Agilent EPM power meter but it should do the trick at 3% of the price.
Jestre .