Who makes the best consumer disk drives?

by Robin Harris on Wednesday, 11 July, 2007

I just did a post over at the more consumer-oriented StorageMojo doppelgänger Storage Bits titled “Who makes the best hard drives?”

In it I attempted to determine, based on simple search requests, if any drive vendor had a real advantage or disadvantage.

Maxtor *had* a lot of online negative comment
Maxtor was 3-4x ahead in raw negative comment – defined as either “Maxtor sucks” or “Maxtor drives suck” – over the next worse in the industry. On a market-share weighted basis they were even worse: 5x or more.

Here’s the data. The total number of Google search results returned in descending order of negative comment. The market share numbers are from an iSuppli report quoted in Digitimes. Market share fluctuated quite a bit for Maxtor due to the acquisition, so they are a very rough gauge. Presumably we’re mostly seeing complaints about the entire installed base, not just the first n months. There could be a bias for complaint when a newish drive fails. Hm-m-m.

Nonetheless the results are startling. Maxtor garnered a weighted 74% of internet negativity. Yikes!

table 1

Adding “IBM drives” to Hitachi’s numbers hurt Hitachi some, but very little in the last year.

Looking at only the last year
I went back and did the same searches while limiting results to just the last year. Everyone in the industry improved – except Maxtor! Double yikes!

table 2

Calling Dr. Seagate, calling Dr. Seagate
Perhaps readers with more insight than I can comment on Maxtor’s problems. There are a lot of ways to anger consumers including flaky drives, poor out-of-box experience and unresponsive customer service. I suspect drive quality had to be an issue.

For as Seagate noted after their acquisition of Maxtor in one of their SEC reports

We are engaged in integration and restructuring processes whereby we are driving to replace Maxtor-designed disc drive products with Seagate-designed disc drive products. We expect this process to be substantially complete by the end of calendar year 2006. We foresee there to be a transitional period through the first six months of fiscal year 2007 during which we expect the inefficient use of the Maxtor manufacturing infrastructure as we wind down the volume of Maxtor-designed disc drive products and incur up-front investment needed for capacity additions to support the ramp-up of Seagate-designed disc drive products.

My translation: Seagate dumped all the Maxtor designs AND manufacturing as quickly as they could. Endemic problems everywhere? The designs I can understand. The manufacturing? Hm-m-m.

The StorageMojo take
I know this is a crude technique, but in the absence of good information we have to go where we can. Probably the reason that Excelstor, the disk drive company you’ve never heard of, came out on top is the lack of English-language complaints. They’re a fast-growing Chinese vendor, presumably focused on the internal Chinese market. Their stuff could be junk and we’d never know.

It certainly looks like Seagate performed a service in taking Maxtor’s stuff – not the brand – off the market. In a consumer-driven technology market it hurts everyone when a single player is a point off the curve. Everyone gets tarred with the same brush.

Comments welcome, of course. If anyone is using the 10k Raptor in their home machine please tell me how much of a difference it made over a 7200 rpm drive. I’m thinking I “need” one for the new StorageMojo quad-core Xeon mainframe with 5 GB of RAM and 4 SATA drive bays.

John had a great comment. Let me quote:

I turned the question around a little, and asked, “If somebody is talking about a drive that sucks, what is the probability that it was manufactured by [vendor name]?” Algebraically: P([name] drive | drive sucks)

My results:

Vendor Suckage Probability
Maxtor .386
Seagate .104
WD .167
Samsung .103
Fujitsu .013
Hitachi .047
Toshiba 0.180
Excelstore .000

End of John’s comment
Is that a great comment or what?

{ 14 comments… read them below or add one }

Matt July 12, 2007 at 3:23 am

It’s good that you note the questionable reliability of figures like this. I think you can fairly claim strong anecdotal evidence but not really anything more than that. On a personal level, I would tend to avoid Maxtor products but for a reason that I think directly impacts on the reliability of these figures.

If you go into any consumer electronics store in the UK, you’ll find plenty of Maxtor product in the external hard-drive section. Bearing in mind that the majority of Joe Consumer’s have absolutely of the HD brand in their computer. They know very well what brand that external hard-drive full of MP3s is though.

Those Maxtor external drives are universally, the cheapest (by some margin) on the shelves. Cheaper even that the Packard Bell boxes. If there’s something I’ve learned working in the storage industry it’s that cheapest doesn’t necessarily mean best. In the world of individual HDs, it’s definitely not the case (although I suffered hard from the Deskstar problems a few years ago).

It’s likely that these cheapy external HDs are failing at a significantly higher rate that average – likely not just because of the drive in the box. Bearing in mind that they’ll be taking people’s MP3 collections with them, they’re probably not very popular.

I’ve no more real evidence for the above than you(mine is straight up personal opinion) but it’s fair to say that for consumers (certainly in the UK), Maxtor is a much more visible brand than Seagate and definitely the almost invisible Hitachi\IBM.

Harold July 12, 2007 at 11:48 am

On Maxtor’s problems, design vs. manufacturing:

If it was the case the Maxtor’s manufacturing was so bad the best solution was to close it down and expand Seagate manufacturing to cover it (note lower sales of Maxtor design+manufactured drives could have made that “easier” if their rep has indeed gotten bad enough), that could drive the move from Maxtor designs, since it will cost Seagate a lot less to make more of the same thing than to start producing Maxtor designs in their facilities.

Also, figuring out exactly what parts of Maxtor were bad, how bad, and if they could have been fixed would have consumed precious resources, the sort that e.g. (hopefully) proactively keeps Seagate quality up. Plus there’s market perception: Seagate’s story WRT Maxtor is simple, a pox on their house which we burned to the ground, now it’s just our 3 year warranty brand.

The fact that their 7210.10s and upcoming .11s are 5 year warranty no questions asked (see below) will keep me with Seagate. To quote a friend who’s an 18 year veteran of the industry with “A Fist Full of Patents” including 2 or 3 in every drive you own, “One should start planning the death of the drive, and all one’s data, beginning the day one buys the drive.”

Re: WD Raptor: due to a problem I just haven’t had time to spare to debug doing an SCSI boot of OpenSolaris (for ZFS, natch, although I like Solaris anyway), I looked at the WD Raptors, and one thing sticks in my craw: while they have a five year warranty, you have to prove to them at the time of warranty service that you bought the disk from an authorized dealer. Seagate doesn’t mess with that, you have a bad drive in warranty, they replace it. Frankly, I don’t want to take a chance that WD will find an excuse to not replace a drive, or have to keep the sales papers in the disk enclosure to make sure I have them when I need them (and it would be bad if a disk caught on fire…).

What I’ve done this decade for ALL my system disks is to buy Cheetah 73GB 10K.6/7s (now the ST373207LW), which I can assure you SCREAM compared to 7200.4 or 10 Barracudas (not that they’re very loud). $2.40/GB vs $1.96/GB for the same sized Raptor (from ZipZoomfly, which knows how to repack disks!), and you have to buy a HA and cable, but for one or two disks you aren’t going to notice much if any speed loss from only doing Ultra-2 or 3, and NCR/Symbios/LSI based cards are standard, fast enough and reasonably cheap.

(And if you remember WD’s ill fated Enterprise SCSI attempt, one doubts that the write firmware of the Raptors is anywhere near that of Seagate’s Cheetahs, they just haven’t had a fraction of the time necessary, nor would it be likely they could buy this sort of thing (nor would it be worth it for a boutique drive, whereas Cheetahs are business workhorses)).

(Note, on the (called into question by the CMU paper) presumption that SCSI enterprise disks are more reliable that consumer ones, I’ve been using them as primary system disks since the early 90s, Seagate->HP->IBM->Seagate again. I did just loose one 10K.6 out of 2 in a system after 3.5 years of service, but that’s probably pretty good, that’s the only failure in a decade and a half and they’re lasting more than 3 years.)

For straight 33MHz/33bit PCI (which again is “fast enough”) HAs I’ve been using the Tekram DC-390U2B (Ultra 2) which comes with a cable, and to drive a LTO-3 fast enough and to get the extra data integrity I just got an Ultra-3 LSI LSI20160B-F (plus a Granite Digital cable ($$$)), which works fine except for the above mentioned OpenSolaris only boot bug (XP no problem, Solaris install and safe mode boot no problem).

Of course, as I’m sure you know, 10K.7s are now 2 generation old drives and Seagate isn’t going to make any more 10K generations this small (or so they say), but 15K.Xs will be cheaper in 5 years, and maybe there will be more and better Raptor class SATA drives—who can say, except that now is a GREAT time to buy LOTS of disk space (remembers getting on the net when the Datacomputer with its Ampex Terabit Memory Systems was the cat’s meow :-).

– Harold

John July 12, 2007 at 11:50 am

Robin:

You’ve got a methodological problem here, albeit one that occurs quite commonly.

The number of hits that Google reports on the first page of a search is only an estimate, based on the conditional probabilities of the word pairs. That estimate is refined as you drill deeper into the search, until it finally gives you the correct number of unique hits.

So, for example, if I google “maxtor sucks”, it reports 1650 hits on the first page. However, as I proceed along the pages, the number of hits slowly decreases, until finally, on page 25, it reports only 245 hits. Even if I click on the option to include omitted duplicates, I only wind up with 665 hits. (Note that with duplicates, Google continues to report “about 1610” hits right up until I hit the last page.)

Tony in SV July 12, 2007 at 1:28 pm

Well, Seagate basically had to choose to do things either its way or Maxtor’s way – because Maxtor does just about everything reversed from the rest of the industry (e.g. drives spin the opposite direction, heads numbering reversed).

Interesting, Excelstor is doing reverse outsourcing – their R&D facility is in Longmont, CO. That’s probably a good location, because I suspect there are a lot of laid off (from Maxtor, Seagate, etc) HDD people in Longmont.

I don’t think Excelstor will be having any big impact in the US anytime soon – their biggest drive is 160G. Also, with the big players (Seagate, Hitachi, Fujitsu) already vertically integrated or going integrated (WD), it’s going to be hard to get heads (only one independent supplier left, TDK) and media (only three independents left, all Japanese – and likely to be fewer soon).

I don’t have any personal experience with the Raptors. They have a reputation for being the best (or second best) workstation drive, regardless of price, but not so great as a server drive (SCSI/SAS is better). If I ever get the money to build a good workstation instead of looking at Fry’s bargains and eBay (I’m a married man now), I’d probably use one or two. See storagereview.com, for example http://storagereview.com/WD1500ADFD.sr

Tony in SV July 12, 2007 at 1:39 pm

Harold,
WD can certainly make excellent desktop drives – the Caviar series has had a reputation as a fast drive for a long time. But the tuning is different for a server drive.

Look at storagereview.com – in the Office Drivemark 2006, the Raptor is #1 and the first Seagate drive is way down the list. But if you look at IOMeter File Server -128 I/O (a very server oriented benchmark), the top is dominated by Seagate with the Raptor a ways down.

So how the drive will be used has a major impact on which drive is fastest.

BTW, I’ve seen a lot of HDD parts over the years, and sometimes there is a significant quality difference between the server (SCSI/FC) drives and consumer drives.

John July 12, 2007 at 3:11 pm

OK, I am an unrepentant geek, married to a computational linguist, so I just *had* to find another way to analyze the data.

Weighting your results by market share is one approach, but probably a better one is to weight it by the amount of discussion surrounding the company. So, when somebody is talking about Maxtor, how often do they say that their hard drives suck? Or, expressed algebraically, what’s P(suck|Maxtor)?

Unfortunately, Google truncates the search results for large results like ‘Maxtor drives’. So I’ll use the same methodology I faulted you for above, and pretend it’s intentional by calling it “Google-based smoothing”.

I did the same searches you did, excepting that I added the word “drive” outside of the quotation marks to samsung, hitachi, toshiba, and fujitsu, to avoid bias against companies with other product lines. I also did a base collocation search of “company name” and drives, to get the base frequency. Then I divided the suck-sum by the base frequency to get what I’ll call the suckitude coefficient.

My results:

Maxtor 0.00063
Seagate 0.00017
WD 0.00029
Samsung 0.00014
Fujitsu 0.000023
Hitachi 0.00008
Toshiba 0.000258
Excelstor 0

My summary: Maxtor does indeed suck, but only twice as much as WD, and five times as much Seagate. What’s a surprise is that Toshiba also sucks, more than Seagate but less than Western Digital. Samsung and Seagate run almost neck-and-neck, while Hitachi and Fujitsu lead the pack at not sucking.

John July 12, 2007 at 6:27 pm

Because I just can’t help myself, and I love crunching numbers:

I turned the question around a little, and asked, “If somebody is talking about a drive that sucks, what is the probability that it was manufactured by [vendor name]?” Algebraically: P([name] drive | drive sucks)

My results:

Maxtor .386
Seagate .104
WD .167
Samsung .103
Fujitsu .013
Hitachi .047
Toshiba .180
Excelstor .000

An interesting thing to note is that Western Digital and Toshiba flipped around in the rankings, so that even though it’s more probable that a WD drive sucks, it’s more probable that a given sucky drive is a Toshiba. That’s because there are more discussions around Toshiba drives.

And that’s all the fun I can squeeze out of these numbers. Last word from me.

John July 12, 2007 at 9:28 pm

Robin:

Sorry to be obsessing about this but I’ve gone from being just amused by this approach to being fascinated. After reading the comments on your zdnet blog, I’m interested in the fact that at least some of the anecdotal evidence seems to be backing up these numbers.

A little while back, I was trying to convince a couple of colleagues to drop out of the storage game and start pursuing text mining. One of the things I proposed to them was doing some kind of simple frequency analysis stuff on brand identity. You’ve given a really engaging example of how that might work… (I *do* think that my numbers are better, though.) 🙂

Robin Harris July 12, 2007 at 10:54 pm

Damn, StorageMojo has a great audience!

John, I didn’t know about that Google bug, so I will redo the numbers with that in mind. Text mining is SigInt with static data. I wish Google enabled arbitrary time-period searches. Imagine the time-series information one could construct.

Now think about the internet in 100 years.

And thanks for the advice on Raptors. Since I have a SATA-based system, and I want a fast drive, I’ll probably buy a Raptor. And tell you all about how it goes.

But today my 8 month old 22 inch widescreen LCD monitor went dark. This is when I wish I could hustle down to Fry’s and have a new one in 20 minutes. Instead it was an hour drive to Flag – tourists and construction – and another hour back. But it was a beautiful drive up and down Oak Creek Canyon. Maybe I needed that.

Robin

Robin

Chris Weiss July 13, 2007 at 3:50 pm

I’m curious as to how you handle queries for Western Digital? Did you search for “WD” or “Western Digital”? IF both, what effort (if any/if any was needed) was there to eliminate duplicate results?

Robin Harris July 13, 2007 at 5:13 pm

Chris,

I tried “WD” and didn’t get much in the way of hits, IIRC. So I stuck with “Western Digital”.

Robin

Allen Cole July 18, 2007 at 3:29 pm

About Seagate model number ST373207LW. If you search that drive on support.dell.com, you get a recommended drive firmware update to version D704 or D404 depending on the rotational speed.

Harold July 20, 2007 at 7:40 am

Allen: Thanks!

Although, if I read the part numbers and dates correctly, this fix is for fairly old drives (12/04 to 09/05 releases by Dell). The oldest 10K.7 I own was bought early this year, and SMART/Seatools reports the firmware level is
0005, which I assume would correspond to a future “D705” (Dell probably hasn’t qualified/modified Seagate’s base 0005 yet).

But point well noted: drive firmware is getting as complicated as a moderate sized OS at minimum, and firmware upgrades should be a part of one’s PM. Unfortunately Seagate doesn’t officially offer firmware updates for “for our standard distribution ATA/SATA drive models.” I’ve just asked Tech Support for a certificate to get access to the SCSI firmware download area … we’ll see what response I get.

(As a side note, Seagate’s fancy site seems to be often dependant on IE (vs. Firefox 1.5.x).)

– Harold

Smith January 9, 2008 at 7:25 pm

Oh, lord yes! Avoid Maxtor drives like the plague. In 15+ years of doing data recovery I have seen no other brand come close to the failure rate of Maxtor drives. Some folks get lucky, but you’re taking a major gamble with your data putting it on a Maxtor drive. Maxtor has officially been acquired by Seagate now, and Seagate says they will phase out production of Maxtor-technology drives using Seagate technology; and supposedly the Maxtor name will be done away with in reasonably short order. This is *good* news…

Leave a Comment

Previous post:

Next post: