Seagate is the world's largest purveyor of mechanical hard drives. As the company prepares for mortal combat in the consumer SSD space, are its wits, Toshiba's Toggle-mode NAND, and SK hynix memory solutions' 87800 controller enough to get by?
Seagate raked in almost $15 billion in revenues last year. Not much of that came from its enterprise-oriented Pulsar line of solid state drives, which first surfaced in 2009. But Seagate knows that, even though it'll always be a hard drive company first, branching out into the SSD space is increasingly important as time passes.
There are lots of little reasons why SSDs make sense. Disk drives are terribly complex, requiring tons of components and robotically-precise assembly. By contrast, SSDs are pretty easy to put together. Just slap components on a PCB and you're done. That's a big departure for a company familiar with sourcing or building some two hundred different parts for every hard drive it sells.

So, Seagate's strategy for the consumer space is pretty simple, then. Offer a trifecta of storage options, from cheap and cheerful mechanical disks to solid-state storage, and a strange union of the two in between: solid-state hybrid drives. Over time, revenue from flash will increase as SSD adoption picks up from lower prices, and having a better product mix can only help matters. Factor in a remix of enterprise-oriented solid-state storage options, and Seagate appears to have a comprehensive assault on the space all planned out.
And that's where Seagate's new 600 SSD family comes into play. Designed for the ubiquitous consumer and client storage applications, Seagate leaves the heavier lifting to its 600 Pro (Seagate 600 Pro-Series 200 GB SSD Review: For The Enterprise), 1200, X8 Accelerator, and (somewhat outdated) Pulsar lines. That frees up the 600s to compete on value and lower cost per GB. The company might be facing an uphill battle though, since Seagate is probably the last name that comes to mind when you think about solid-state drives. Is there anything the 600 can do to change that?
Technically, there's a lot of overlap between the 600 and 600 Pro. For client and light enterprise applications, the 600 Pro adds more over-provisioning and power loss protection. The standard 600 getting reviewed today does away with those two features, presumably for more competitive pricing. The 600 ships in 120, 240, and 480 GB capacities, and will eventually be available in a 5 mm form factor for environments that require super-slim drives. Those skinny models may or may not utilize a traditional SATA connector, so most folks will probably find themselves looking at the more common 7 mm version when they start moving in greater volume.
| Seagate 600 2.5" 7 mm z-Height | 480 GB | 240 GB | 120 GB |
|---|---|---|---|
| 128 KB Sequential Read/Write MB/s | 500+/400+ | 500+/400+ | 500+/300+ |
| Max 4 KB Read/Write IOPS | 80,000/70,000 | 80,000K/70,000 | 80,000/60,000 |
| Maximum Drive Writes Per Day | 40 GB | 40 GB | 20 GB |
| Max Warranty TBW | 72 TB | 72 TB | 36.5 TB |
| Price | ? | ? | ? |
The 600's specifications are fairly competitive, though slightly vague. Claims of 400 MB/s sequential writes and 500 MB/s sequential reads are pretty common, really. Random performance is just on the cusp of more elite offerings. So, all's well on that front.
Actually, the most interesting line items aren't performance-related, or even the projected 0.58% Annualized Failure Rate. The 600's three-year warranty coverage has the equivalent of a mileage restriction on it. You get three years or 72 TB for the larger two drives, whichever comes first, and half as many writes for the 120 GB model. We're not fans of this. While it's true that most users won't ever come close to those numbers, we don't want to see vendors adding caveats to their coverage. Most of the big SSD companies are pushing five-year guarantees on most of their products as Seagate lays the groundwork for dragging warranty protection into post-flood hard drive territory. Let's hope this doesn't catch on. Ever.
Warranty shenanigans aside, opening up the 600's rugged metal chassis to get at the good stuff is fairly easy. Seagate warns reviewers that popping the enclosure will destroy the drive (as if that'd stop us). But the process is fairly easy. Just unceremoniously yank the bottom chassis cover's metal tabs out of the top half. There are no tiny screws to lose, no fuss, no mess, no sad or broken SSDs. Naturally, the bent metal and twisted tabs make it obvious that someone tampered with the drive, and that's probably the point.

The chassis' top housing is sturdy and feels great in-hand. That's almost completely irrelevant in a review, but when you spend as much time handling solid-state storage as us, quality construction is easy to appreciate. The textured black top housing also has an extension to mate up with the controller, drawing heat away from the logic and into the metal chassis.

Speaking of controllers, it's not surprising to see Link_A_Media Device's (now SK hynix memory solutions') LM87800AA processor at the heart of this desktop-oriented drive. Seagate was once an equity partner in LAMD prior to SK hynix's purchase. LAMD controllers power Seagate's Pulsar.2 enterprise offerings, and now the eight-channel design finds a home in the 600.
Next to the controller are two 128 MB Micron DDR2-667 DRAM packages. This 240 GB model uses just 256 MB of total cache, but the 480 GB model comes armed with twice that amount for page mapping and caching, according to Seagate.

Pairing up Toshiba's Toggle-mode NAND with the LM87800 is a potent combination, which is why Corsair also picked those components for its Neutron GTX. Comparisons between the Seagate and Corsair drives are inescapable, and you'll see the two SSDs go up against each other in our benchmarks.
The 600's 19 nm Type C Toggle-mode MLC flash is very fast, and each of the packages on this 240 GB SSD contains four die, meaning every IC adds 32 GB of capacity. As we saw in yesterday's review of the 600 Pro, there aren't any components on the PCB's naked back; everything fits on the front.

- Seagate's 600 SSD: LAMD And Toshiba, Together Again
- Test Setup And Benchmarks
- Results: 128 KB Sequential Performance
- Results: 4 KB Random Performance
- Results: Tom's Storage Bench v1.0
- Results: PCMark 7 And PCMark Vantage
- Results: Power Consumption
- Seagate 600: Mediocre Power Numbers, But Solid All-Around
2. I dont get why you use QD greater than 4 in the synthetics. All of thses drives are for PC users, who will rarely get QD even equal to 4.
3.I would have liked more real world tests like : Copying to and from drive, restoring backups, decompressing large ISO files , doing all of the above and then noting the time it takes to open Photoshop,
4. Can you do a pre and post defragment test, just for lolz ?
5. Can you do a test where the windows system is paging on the SSD ? basically a measure of the read/write disc speed when the OS is low on RAM and is using the SSD for pagefile.
6. IMHO, if you use completely incompressible data to check the perf of SSD, you are deliberately biasing against the Sandforce based SSD's. Could you use a better mix of compressible and incompressible data ? The dynamic compression will definitely improve the perf of Sandforce SSD's in real world desktop usage.
2. I dont get why you use QD greater than 4 in the synthetics. All of thses drives are for PC users, who will rarely get QD even equal to 4.
3.I would have liked more real world tests like : Copying to and from drive, restoring backups, decompressing large ISO files , doing all of the above and then noting the time it takes to open Photoshop,
4. Can you do a pre and post defragment test, just for lolz ?
5. Can you do a test where the windows system is paging on the SSD ? basically a measure of the read/write disc speed when the OS is low on RAM and is using the SSD for pagefile.
6. IMHO, if you use completely incompressible data to check the perf of SSD, you are deliberately biasing against the Sandforce based SSD's. Could you use a better mix of compressible and incompressible data ? The dynamic compression will definitely improve the perf of Sandforce SSD's in real world desktop usage.
1. The time it takes to do a full drive complete error checking (check file errors+recovery of bad sectors).
2. The time it takes for a deleted file to be recovered ,using a third party data recovery freeware.
I dont have a problem with the dual condition warranty either, its a lot like a car; (10 year, 100,000 miles) I think the problem is that they are only giving a 3 year warranty, where almost everyone else in the ssd market has 5 year warrantys, and unconditioned at that. Even if the drive is faster than some of the other models tested here, id feel much safer buying a drive with a longer warranty, knowing its going to be protected for an extra 2 years.
False. Neither Ocz or samsung have limits on how much data is written on the drive. And their the only two ssd brands worth buying.
Such charts are the domain of dodgy advertisers, not tech sites that seek to convey useful
information, etc.
Ian.
Oh, fully agree with the OP about the Samsung 840/840-Pro, these should have been
in the mix, not the 830. At least in the UK, one cannot buy the 830 anymore, it may
aswell never have existed - even the old product pages for it have gone from most
seller sites (rather unusual IMO). If possible, please replace the 830 data with 840 and
840 Pro, then the Samsung info would be much more relevant.
Ian.
True enough. We've been in the middle of a transition, retesting every drive and trying some different tests. That means that to review this one drive, I had to retest some fourteen other drives. Tedious, and time consuming, I couldn't have wrangled every one in under the gun.
Regards,
C. Ryan
Are you upset that I didn't test at QD 0, or are you concerned because I changed the scaling for the 128 KB Sequential Line graphs? No data is lost, but it does certainly make them easier to read. It's not intended to trick the reader, but rather to impart the information with more clarity and less confusion. This way, the differences are easier to detect. I can always drop in the charts without the scaling as well.
Regards,
C. Ryan
Are you sure ? Samsung drives (my 830 for example) have lifetime counter (counting down from 100%), I believe it's there to protect them from having to honor warranty on drives that have exceeded projected number of writes in their lifespan.
(note that this doesn't mean the drive will fail immediately when counter reaches 0%)
> ...That means that to review this one drive, I had to retest some fourteen
> other drives. Tedious, and time consuming, ...
I can definitely empathise with you there.
I have though commented several times in the past three months
that the 830 is effectively null & void now. I really wanted to
get another for my 3930K build because I was amazed at the way
it maintains steady state performance, but they'd just vanished
from all the usual sellers ("It's *gone* McCready!"), so I bought
a standard 840 250GB which I have to say is quite good overall.
Irony is, not long after, I won two new Vector 256s on eBay...
Anyway, definitely a more relevant product than the 830 now is
the Vertex4 256GB which is still 'current' (continues to be for
sale in the normal way from the usual dealers). It's a little
more expensive than the 840 Pro, but it's a lot cheaper than the
Vector (why is the Vector now so costly? Scan lists it for 236 UKP).
Btw, have you heard anything about an updated Vertex3, called the
Vertex3.20? It's on Scan's site (code LN50566, have a look) with
a slightly higher IOPS rating than the normal Vertex3, though
bizarrely it's priced higher than the 840 Pro. Can't imagine why
OCZ would bother doing an update.
> ... or are you concerned because I changed the scaling for the 128 KB
> Sequential Line graphs? ...
Correct, the 128KB sequential graphs should have a Y-axis origin
of 0. I assume the X-axes are queue depth - as Sakkura says, both
just say 'Title' atm.
> No data is lost, but it does certainly make them easier to read. ...
I think it makes them harder to read, because one cannot use the
instant overall visual look of the graph to gain some idea of
relative performance. It makes the performance differences seem
wider than they really are. I know that zooming in on a narrow
range allows one to move the lines apart to make them clearer, but
the result allows one to infer incorrect relative performances (eg.
it makes the Intel units look terrible).
> ... I can always drop in the charts without the scaling as well.
Better idea: how about having both? eg. just click on the image
and it switches back and forth between the whole graph with 1,0
origin vs. the zoomed in graphs as they are atm? ie. I would
suggest the default should be the whole graph, then click to zoom
in to the way it looks just now, click again to zoom out. Is
that possible?
Ian.
How does conditioned warranty works exactly?
How do they know how much many times the data has been written?
is there any counter/register or anything like that for each block?
How does conditioned warranty works exactly?
How do they know how much many times the data has been written?
is there any counter/register or anything like that for each block?
Yeah, the SMART attributes register information like that.