New Seagate Hybrid Drives Hampered By Slow Mechanical Guts

crookedvulture writes "Seagate announced its third-generation hybrid drives last month, revealing a full family of notebook and desktop drives that combine mechanical platters with solid-state storage. These so-called SSHDs are Seagate's first to be capable of caching write requests in addition to reads, and the mobile variants are already selling online. Unfortunately, a closer look at the Laptop Thin SSHD reveals some problems with Seagate's new design. While the integrated flash cache reduces OS and application load times by 30-45%, overall performance appears to be held back by its 5,400-RPM mechanical component. Seagate's last-gen Momentus XT hybrid spins its platters at 7,200-RPM, and it's faster than the new SSHD in a wide range of tests. The upcoming desktop SSHDs will also have 7,200-RPM spindle speeds, so they may prove more appealing than the mobile models."

Comments

Hey, Seagate:

[CanHasDIY]

2000 called, they want their crappy hardware back.

Re:Hey, Seagate:

[CanHasDIY]

Whoosh... or douche?

I'll let the reader be the judge.

Re:Hey, Seagate:

[girlintraining]

2000 called, they want their crappy hardware back.

Actually, this kind of thing has preceded every major storage advance in computers. As the replacement technology matures and becomes mainstream, the producers of the legacy technology cut corners on quality in order to maximize profit ahead of decommissioning of their production facilities for that technology. Zip disks, floppies, consumer tape drives, etc. All of these had major quality control issues near the end of their production runs. You would be hard-pressed to find a technology in this field that as it sunsets doesn't have its quality turn to absolute crap.

What Seagate is doing here is an attempt at prolonging that period to maximize profits on its existing (mechanical drive) production lines by gluing a turbo-charger onto the I/O equivalent of a four banger. They figure the consumers are idiots and will fall for four color marketing glossies saying these are the "fastest mechanical drives ever!" and boldly print the percentages all over the packaging... and then praying they don't look an aisle over and realize that a modest SSD would blow it out of the water for not much more cash. You can bet these drives are not built to the same specs or tolerances of previous models -- they will fail more often, and because of their hybrid nature, will be more difficult to recover data from when they do, if you can recover anything at all.

It's a douche move, but... it's sound business practice. Sell your customers down a river to keep profits up until you can turn up production on the Next Big Thing, and then try to buy them back later with discounts and deals.

Re:Hey, Seagate:

[Anonymous Coward]

2000 called, they want their crappy hardware back.

2000 called? Did you warn them about Katrina? 9/11? No? You dick!

(Oblig. xkcd)

Re:Hey, Seagate:

[plover]

Certainly, technology producers change their designs as they go, making them cheaper and cheaper until they just barely stay within the lower end of their specs. It's called "optimization", and it's the responsible thing for a manufacturer to do for its shareholders. Same volume, same price, higher return on investment. If you don't personally like it, you can pay more for a device with better lower-end specs from someone else.

In the case of hard drives, Seagate knows that 5,400 RPM machines are far more reliable than 7,200 RPM machines, even after optimization of both. I suspect that in order to make 7,200 RPM drives more durable, the manufacturing costs exceeded that of SSDs. And if 7,200 RPM drives can't be made more reliable for an affordable price, I expect that is why Seagate is dropping them completely.

Hybrids are a way to sell a slightly faster version of the mechanical drive for people on a budget who still need reliability. No, it's not going to out-perform a 7,200 RPM drive, but over time it will do better than a 5,400 RPM drive without a cache. If you want performance, spend the extra money for a real SSD. If you want cheap speed without reliability, you'll have to buy a faster drive from someone else.

Re:Hey, Seagate:

[girlintraining]

It's called "optimization", and it's the responsible thing for a manufacturer to do for its shareholders. Same volume, same price, higher return on investment. If you don't personally like it, you can pay more for a device with better lower-end specs from someone else.

Are you some kind of capitalist apologist or something? I never offered an opinion on whether I like it or not, I was simply pointing out that this is what businesses do.

Same volume, same price, higher return on investment. If you don't personally like it, you can pay more for a device with better lower-end specs from someone else.

Perhaps you didn't read carefully enough my previous comment: All the manufacturers of mechanical drives are going to be doing the same thing. There isn't anyone else to buy these magical unicorns you speak of from.

And if 7,200 RPM drives can't be made more reliable for an affordable price, I expect that is why Seagate is dropping them completely.

As has been covered before, reliability is not what is driving these changes.

Hybrids are a way to sell a slightly faster version of the mechanical drive for people on a budget who still need reliability.

Lolwut? You're taking the exact same product, gluing it to another product, and that improves reliability? What planet are you from where increasing the complexity of a device improves reliability? Engineers, software and mechanical, rely on the KISS principle for a reason -- it makes troubleshooting easier, and it improves reliability. Complexity is antithetical to reliability. If you want a classic example of how complex design can cause all kinds of reliability issues, look at the LOX engines on the space shuttle. Some of the most complex machinery ever designed -- and it was built top-down, not bottom-up. And every engineer who worked on it will tell you, they honestly don't know all the possible failure conditions because of that. It's simply too complex.

Re:Hey, Seagate:

[hawguy]

and then praying they don't look an aisle over and realize that a modest SSD would blow it out of the water for not much more cash.

Only if they don't need the storage capacity of the spinning hard drive... many laptops don't have the room for both an SSD and a hard drive.

The smallest Seagate SSHD is 500GB and costrs around $99. The cheapest 500GB SSD I can find costs around $350.

So, for those that need the higher capacity, you can't get an SSD for "not much more cash".

So yes, an SSD would have much better performance, but not equivalent capacity at the same price point. For the kinds of things most people use a laptop for (booting windows, loading apps) the SSHD gives close to SSD performance, while still letting them keep their large media files on the hard drive

It's a douche move, but... it's sound business practice. Sell your customers down a river to keep profits up until you can turn up production on the Next Big Thing, and then try to buy them back later with discounts and deals.

Slower speeds aren't just a cost cutting move - cutting the speed reduces noise, power consumption (so you get better battery life on your laptop) and lowers heat production (so you get better reliability for your hard drive).

Re:Hey, Seagate:

[nabsltd]

Their hybrid nature does not affect data recovery. All the onboard SSD does is cache data that exists on the HDD.

There is no way to know what will happen to the overall usability of the drive if the flash fails (either through normal write exhaustion or catastrophic failure).

Hopefully, Seagate did the right thing in this case and the drive would turn into the equivalent of a pure mechanical drive. But, failure of the flash or its controller might cause the drive to become completely unusable. Unless they specifically deal with this as a "special" failure mode, it wouldn't be that different from some essential part of the controller on a purely mechanical drive failing (like the DRAM cache), and that usually turns the drive into a doorstop.

Re:Hey, Seagate:

[plover]

And why are the magical unicorns going extinct? No demand. People on a budget buy a slow drive, people who can afford slightly more buy an SSD hybrid, people who have the means and require the performance buy an expensive SSD. If Seagate saw there was a market for expensive yet still slow 7,200 RPM drives, they'd keep making them. (Actually they still do have a few, it's their corporate customers who are too large and slow to make a more rational decision.)

Sorry I wasn't clear about the reliability thing. Seagate's 5,200 RPM drives have twice the reliability when compared to their 7,200 RPM drives (the Annual Failure Rate is predicted at 0.48%[1] vs 1.065%[2]), and the difference made by adding the complexity of a cache gives them a predicted AFR of only 0.50%. And reliability is absolutely driving this: if they have to double the reliability of their 7,200 RPM drives, they will cost more than plain old SSDs (also at 0.50% AFR[1]).

[1] http://origin-www.seagate.com/internal-hard-drives/laptop-hard-drives/
[2] http://origin-www.seagate.com/internal-hard-drives/enterprise-hard-drives/hdd/enterprise-value-hdd/

Laptop 7200rpm drives discontinued

[SpaceManFlip]

This is because they recently announced the discontinuation of their entire lineup of 7200rpm drives in the 2.5" form factor.

Desktop drives in 3.5" size still have the 7200rpm drive speeds, so the smaller ones have been gimped by Seagate's mfr'ing decision

Re:Laptop 7200rpm drives discontinued

[spire3661]

You dont trust ANY 'disk drive'. You backup and recover, just like any other storage medium. Trust doesnt even enter the equation.

semi serious question

[scream at the sky]

Why are we not seeing more 10K drives? Other than the WD Raptors, I haven't seen 10K desktop drives in forever.

I would think it would be a better compromise, am I missing something?

Re:semi serious question

[AuMatar]

Noise issues, power issues, and the likelihood that cheapening SSD will make magnetic disks obsolete. People who really care about speed just go solid state. With the price dropping I'm sure we all will in a few years.

Re:semi serious question

[sl4shd0rk]

Why are we not seeing more 10K drives?

Only reason I can think of is direct competition with top-shelf scsi hardware. a 900G 10K SCSI is about $500 bucks. I sure would be tempted to RAID twice as many 10K SATA for half that price if I could get the same RPM.

Re:semi serious question

[dlakelan]

Higher density at constant speed means higher signalling rates now vs before. We're already reading more off the disk per second at 7200 rpm than we were at 7200 rpm back when 200GB was big. Power requirements have taken a bigger position, and also at the higher densities tolerances need to be more exact and even more so at higher speeds. Going to lower spinning speeds allows you to get better results without tightening tolerances as much.

Re:semi serious question

[fuzzyfuzzyfungus]

Why are we not seeing more 10K drives? Other than the WD Raptors, I haven't seen 10K desktop drives in forever.

I would think it would be a better compromise, am I missing something?

10Ks on the desktop(and, at least to some degree, although less of one, 10 and 15Ks in the enterprise) have been curb-stomped by SSDs, actually harder than their slower brethren.

Everybody knows that 5.4s and 7.2s are horribly slow, for everything except very well behaved linear reads or writes; but they are insanely capacious and cheap, so people who don't need speed buy them anyway, and by the truckload. On the consumer end, cheap shit sells by the pallet, and needs something to boot from, and on the enterprise side the (partial) unification of SAS and SATA means that a lot of stuff that you used to have to dump right to tape can now be handed off to crazy-cheap 'nearline' HDD storage(and, in sufficient quantity, a lot of less demanding storage tasks are perfectly fine on prosaic 7.2K SATA, and since SATA drives drop right into SAS slots/connectors, they all play nicely with the RAID backplanes and hot-swap trays and things, which wasn't the case back in the PATA/SCSI days).

Among people who need I/O above all, any mechanical drive is an amusing little smudge clinging to the X axis when graphed against the performance of any halfway decent SSD. When a good SSD can easily be several orders of magnitude faster, the fact that you might(best case) triple performance by going from 5.4k to 15k barely registers; but the price of increasing spindle speeds certainly does.

Velociraptors, and their ilk, had a brief period of popularity back when all the 15Ks were SCSI(and so were either wildly expensive, or dodgy fleabay gear, and usually needed an add-on card that cost more than most consumer hard drives, even used) and SSDs were either nonexistent or more expensive than entire workstations. Now, they just aren't a terribly impressive offering. If you don't care much, you can get a rather larger and quieter HDD for substantially less money. If you do care, a surprisingly small premium will get you an SSD that will blow the Velociraptor out of the water.

The gulf between good solid state storage and mechanical storage, in terms of latency, is just so enormous that we will probably see more retreating from higher spindle speeds than advancing. High precision, high reliability mechanical parts are stubbornly costly, so increasing spindle speeds isn't free; but the performance gap is sufficiently vast that even some terrifying HDD built with ultracentrifuge technology just isn't going to be as fast as an SSD. Flash prices are still high enough that HDDs have plenty of retreating room into high capacity/high latency applications; but any attempt to achieve parity in low-latency work would just be comedic(if probably impressive from an engineering standpoint, and when it tore itself apart and shredded everything nearby)...

Re:semi serious question

[scream at the sky]

I get that, but I also see the flip side where 10K drives have typically been aimed at Enterprise, rather than consumer business.

I have a pair of ancient 74 gig raptors that I use for the boot system (raid 0) on my home NAS, and I love the disks. I'd love to see some consumer grade 10K drives with a standard warranty.

And, yes, I have SSD in my laptop, and I agree that spinning platters have a limited number of days, but for a company like Seagate who is pussying around with these hybrid drives, it would make sense for them to do what they do well, instead of half assing something, and giving us half a solution.

Battery Life?

[TheNinjaroach]

See the current Slashdot poll about laptops. The winning request is better battery life. How much extra juice does it take to spin those platters at 7200RPM? Perhaps Seagate's manufacturing decision to use 5400 is better informed than it appears.

Re:Battery Life?

[BLToday]

"How much extra juice does it take to spin those platters at 7200RPM? "

From my experience with their 1st-gen Momentus XT (500GB), about 15% reduction in overall battery life compare to my previous 5400 rpm drive.

Re:Battery Life?

[TheGratefulNet]

I know YOU see no reason for not having vari speed in mechanics.

then again, its very clear to me that you have no clue how the electronics, clocking and mechanics would DO that.

maybe, just maybe, its not do-able within reason.

varispeed is marketing BS. it makes no sense to vary platter speed. stupid idea, in fact!

Re:Battery Life?

[ackthpt]

I see no reason why a hard drive can't just putt putt around at a slow speed and ramp up that speed if the IO queue starts filling faster than it can be emptied.

Does anyone know why this has not been done more already, especially in laptop drives? There has been some "eco" desktop 3.5" drives that can lower their speed when idle, but not much else.

Seagate Barracuda ST3000DM001 (3TB 7200RPM) spin down to conserve power, unless you use power management tool to tell them to remain spinning at all times.

Re:Battery Life?

[dgatwood]

I see no reason why a hard drive can't just putt putt around at a slow speed and ramp up that speed if the IO queue starts filling faster than it can be emptied.

I thought I had read that WD's green drive did exactly that.

Re:Epic failure

[fuzzyfuzzyfungus]

I would strongly suspect that it's an OEM thing, mostly:

Intel, for one, sets some fairly strict boot-time requirement for an OEM to be able to slap "Ultrabook!!!!" on the laptop and possibly get some Intel 'marketing assistance' cash. Microsoft has also been doing a bit of leaning on OEMs in terms of how fast Win8 machines need to boot in order to earn their little sticker of meaningless approval.

OEMs, of course, still need to shove $400 black-friday specials out the door. What will we do? Well, it just so happens that our good buddies at Seagate have a hard drive that is super cheap, being a very undemanding mechanical model with only a small amount of flash; but just so happens to be able to(if configured and pre-cached and whatnot properly) boot the OS like a bat out of hell... Seagate proceeds to sell a giant pile of the things.

Given that Seagate knows that benchmarks are going to happen, they have no realistic hope of pulling the wool over the eyes of informed enthusiasts. I'd be surprised if they care: less cost-sensitive enthusiasts are going to buy SSDs anyway, more cost-sensitive ones may well buy if the price is right, and making the spindle slow and the cache small will definitely help there.

As a strategy for launching a successful enthusiast storage brand, Seagate's choices would be suicide; but 'enthusiast storage' isn't a terribly big market anyway, and the SSD guys own it now, so Seagate doesn't have a choice about not playing there. The OEMs, on the other hand, are caught between certification demands(which generally specify boot time, resume-time, etc. not 'IOPS Random 4k' scores) and price pressures. This product looks like it is tailor-made to be pitched right at them.

Re:replaced by SSDs and 2.5" drives

[tlhIngan]

10K was typically an enterprise thing. Enterprise has generally moved to either SSDs or to 2.5" drives (currently available in 10K and 15K).

The increased areal density gives decent capacity for the 2.5" drives, and the smaller platter means it's more robust, causes less vibration, and uses less power. It also takes up less space in a server.

Most 10K 3.5" drives used 2.5" platters because at the speeds they spun at, there was no way a 3.5" platter would survive the rotational induced stresses. (Most of the additional bulk was used by cooling fins to help cool the platters - a good chunk of the heat a hard drive produces comes from friction... with air!)

In fact, the latest desktop 10K drives were 2.5" drives bolted to a huge heatsink (it was recommended to NOT remove the heatsink).

Due to their temperature sensitivity (they had to have good cooling) they were a limited market (servers, mostly, there were a few desktops that had it but they were limited by case designs), and relatively low capacity versus the more contemporary 7200RPM drives.

With SSDs, low capacity high speed drives were basically extinct - why buy a 320GB 10K drive when a 512GB SSD would often be cheaper (and faster, to boot).

Re:7200RPM less reliable than 5400?

[TsuruchiBrian]

5400 rpm models use less power than 7200 models. In a notebook everything is about power usage and physical size, and I guess speed too.

I have 4TB 5400 rpm drives in my raid because my network is the bottleneck for speed anyway, and the cooler the drives run, the better (there is 5 of them).