HDD mini-reviews

[Behemot]

My MB2000EAZNL has 56717 hours and counting, 158 start/stops. Seen few more such drives around with same milage. According to S.M.A.R.T. it has 23 % average life remaining so we could extrapolate the predicted factory lifetime to be (at least) 8.40 years.

That said, WD never mastered reliable 10k not to mentions 15k drives so they did not even make them really. That's why they had to buy Hitachi, trolol

[LongRunner]

I'm afraid my very humid shed (this is Australia, you know ) hasn't done any favors to the HDDs stored in it (nor any other metal objects, for that matter):
Two of my WD800JD-00LSA0s (dug out of there just recently) now sound sick (they were healthy when I got them). No bad sectors yet, but they are developing several read errors (and even a few seek and write errors), and their head landings (after spin-down) also sound a bit scratchy. Poor little guys.
There is some corrosion on their aluminium base casts, too (although my ST3120026A and ST340014A have worse corrosion, yet still seem to work and even sound OK; I bet they would still run with no bad sectors, too bad I don't have a working PATA connection to hand).

Fortunately, my ST380817AS and especially ST380013AS remain in pretty good condition (even their PCB solder hasn't fared too badly, with few if any whiskers; my ST3120026A was still running flawlessly the last I checked, and its PCB does show numerous whiskers much like my poor tortured ST380011A did). So I'll bet they can keep going for 20+ years after manufacture (i.e. into 2024, as both were made during 2004) with good care. Enterprise (S)ATA drives, indeed.
(Mark Furneaux did bury a well-used ST380013AS deep underground, in a plastic bag, in Canadian weather; yet although horrifyingly corroded, it still sort-of worked for a short time – albeit with severely degraded performance. I bet any lesser HDD would have been completely dead then.)

So from now on, any HDDs I care about will be kept safely in the house (with proper heating, cooling, and dehumidification). I even have the idea of building a special home for them (and other precious precision hardware), complete with a miniature air-conditioner (presumably Peltier-based) and dehumidifer…

Overall, I'd probably rate the HDDs we've experienced in about the following order, from most reliable to least:

1. Seagate Barracuda ATA IV and V (with the 7200.7 just slightly behind, given its lead-free PCBs and occasional media coating problem )
2. Mid-2000s Western Digitals (after they finally switched to FDB spindles)
3. The better IBM/Hitachi models (although the 75GXP, 40GV and 60GXP were quite horrible, the engineers did get their act together and the 180GXP was reportedly very reliable, with the 7K250 not far behind)
4. Samsung (before their infamous F1 series)
5. Seagate U Series? (They reportedly weren't nearly as good as the Barracudas of the same era; U Series drives often reported quite a few seek errors, whereas Barracudas usually have none. And yes, I am examining the attribute properly. Plus, the claimed seek time for U5 and U6 was blatantly false.)
6. Maxtor (they were a bit cheaper and ‘quirkier’ than most of the rest; it's hilariously ironic how their MaXLine Plus II, the first officially “enterprise” (S)ATA HDD, was – just like its “consumer” version, the DiamondMax Plus 9 – so much more failure-prone than the Barracuda 7200.7 )
7. Barracuda 7200.11/ES.2 (and Momentus 7200.3, for that matter)
8. Toshiba mobile (2.5″, probably also 1.8″) HDDs
9. IBM DeathStar 75GXP, 40GV and 60GXP (YMMV on whether the 120GXP was OK, or just as terrible; I did find a 75GXP with no bad sectors, as noted here)
10. 7200.14 (it's frankly an insult to even call it a Barracuda; I can see why Seagate dropped the name, for a time )

Do tell me if you would agree (or disagree, for that matter).

[Behemot]

Don't really keep track aymore which series are more reliable, only remember many of the IBM/HItachi (DeskStars I think?) ones to be utterly unreliable.

The Seagate U series drives I remember to be shaky, loud and slow, lol. (They've been 5400RPM drives after all.) And their rubber jacket, that one you remember. That was to keep them warm I guess?

[LongRunner]

ST3250620AS (firmware 3.AAJ) (83.3GB/surface, 3 heads, 7200RPM, FDB, CSS, SATA 300MB/s NCQ, 16MiB cache)
Well, this one's had it: It came up with 722 seek errors (most healthy Barracudas have none) and a lot of very slow sectors (each one accompanied by a ticking noise) when I surface-tested it yesterday, so I judged it not worth risking any data on (in the presence of so many healthier HDDs to choose from).
By the way, the casing on this drive doesn't even look corroded at all. So while the Barracuda 7200.7 and earlier (with the possible exception of Barracuda ATA II, according to some reports) were extremely reliable, and 7200.8 was probably quite decent too (I don't have personal experience with it); the 7200.10 wasn't much better than average. (I definitely don't think it was mere coincidence that Seagate then ‘jumped the shark’ shortly after buying Maxtor. )

I've decided to put my ST31000528AS out of its misery, too; the 7200.11 and 7200.12 drives have let me down enough already (mind you, mine had the “good” firmwares), and its PCB is even suspiciously brown suggesting that it's made of cheap phenolic resin (the earlier Seagate PCBs were of very high quality).
(And I don't even want to think about the 7200.14 anymore; I'll bet you could make a stripe-set of 19 × ST3160023AS/ST3160827AS – or 25 × ST3120026AS/ST3120827AS or 38 × ST380013AS/ST380817AS, for that matter – and your data would still be far safer than with a single ST3000DM001.)

(It's a shame about the poor QA, as the basic design still looked half-decent; and my ST31000528AS weighed 615g. My ST3250620AS weighed 585g.
Although, I will concede that the ST3250620AS's disks are 1.75mm thick, whereas the ST31000528AS's are only 1.27mm thick.)

WD5000AAKS-08A7B2 (167GB/surface, 3 heads, 7200RPM, FDB, L/U, SATA 300MB/s NCQ, 16MiB cache)
Reallocated a group of 4 sectors during an erase; otherwise seems OK.

If I borrow Red Hill's (vintage) HDD reliability rating system for a moment, here are my (tentative) ratings under it:

• AAA (outstanding): Barracuda ATA IV and V, Barracuda 7200.7 (especially those with Agere microcontrollers and the TI SH6950 motor driver, rather than their less robust STMicroelectronics counterparts)
• AA1 (good—excellent): Barracuda ATA III, probably Barracuda 7200.8; Maxtor/Quantum D740X-6L; WD Caviar (Blue) from late 2004 to ≈2010
• AA2 (fair): U Series 6, Barracuda 7200.9 and 7200.10; typical Hitachi
• AA3 (poor): DiamondMax +9/MaXLine +II; Barracuda 7200.11 and 7200.12; WD Caviar Green
• BA (terrible): Slimline Maxtors; “Barracuda” 7200.14
• AAX (uncertain, but promising): WD Caviar Black
• AAY (uncertain, but worrying): DiamondMax 16/MaXLine II (it seems that the 4A300J0, especially, could be very short-lived indeed )

Looking back, then, Seagate's inclusion of FDB motors as standard equipment in the Barracuda ATA IV was very much bang-on-time.

[Behemot]

With the increasing data density it only gets worse, that's what I see anyways.

[LongRunner]

With the increasing data density it only gets worse, that's what I see anyways.

You're not wrong there (and I would like to see upgrades to the error-rate specifications), but I don't think that alone justifies how hard Seagate (in particular) plummeted. While the Barracuda ATA IV/V and 7200.7 were of true enterprise quality (and no doubt, much stronger even than many “proper” enterprise HDDs from the less competent manufacturers ), the 7200.14 turned out even worse than the DeathStars (and yes, Seagate have been sued accordingly).

(When the 75GXPs got warm, they were indeed good as Dead; but if you kept them cool, they could still survive. Red Hill Technology should know.
And I could have forgiven that problem if IBM had admitted to it, and dealt with it properly. Too bad they weren't so nice.
The 7200.14, on the other hand…I'm not sure if we can even count all of its faults. )

EDIT: It seems rather that the Hungarian-made DeathStars were bad, while those made in the Philippines (by NEC under contract) seem OK.
While compounded by some design deficiencies, it appears that poor quality control was the biggest problem.

Clearly, buying out Maxtor did Seagate no favors at all (firmware bugs don't discriminate between “consumer” and “enterprise” products, with the 7200.11 and Barracuda ES.2 affected alike; I don't think the timing here was a mere coincidence, observing that Seagate's QA never fully recovered since then).

[LongRunner]

Maxtor 52049H4 (5.12GB/surface, 4 heads, 7200RPM, BB, CSS, UDMA 5, 2MiB cache)
Seagate Barracudas were by no means alone in dying from chip failures, even while the mechanics are still good. This one doesn't short out the power supply, and the voltage regulators check out OK, but it still doesn't even attempt to spin; my chief suspect is the parallel flash chip, made by STMicroelectronics (of course ). It did work fine before the chip failed (in my shed), but I don't have a particular use for a drive this primitive (if I want 20GB at 7200RPM, I'll use my ST320011A instead); so I've ended up taking it apart. (Anyway, I did see some specks of alumina laying on the disk. )

(I do like that the PCB bears the codename “FISH'N M16+” ; indeed, this drive was a worthy competitor to Seagate's Barracuda ATA II, observing StorageReview. Overall, I think it was one of Maxtor's better efforts.)

Maxtor 31536H2 (7.68GB/surface, 2 heads, 5400RPM, BB, CSS, UDMA 5, 512KiB cache)
The cheaper and less-impressive “Value Line” counterpart to the above drive (it even has the same basic PCB, albeit with some different parts); while still reasonably fast (for its age), this has only an aluminium (instead of stainless-steel) top cover (at least the upper damper layer is still stainless-steel), and the bearings are of noticeably inferior quality too. At least the electronics have survived (so far), perhaps because this one isn't unlucky enough to have any STMicroelectronics chips. AAM is actually supported, and quite effective (not that it's terribly helpful, with the noisy spindle bearings). The PCB was able to spin the 52049H4's mechanism to 5400RPM (with considerably less bearing noise), but obviously can't operate with such a gross parameter mismatch (just clicked); when put back on the 31536H2, it seems to work again.

The 50249H4 weighed 555g overall (almost as heavy as a Barracuda 7200.7, or Barracuda ATA IV/V with the SeaShield removed), and its top cover (including gasket) weighs 135g; the 31536H2 weighs 475g. Each disk from the 50249H4 weighs about 18g and measures 0.99mm (near enough to 1mm) thick; so doing the sums puts the 31536H2's cover at 75g or so.

So, 7200RPM HDDs of 1999 and the early 2000s were often better-built than their 5400RPM counterparts; and this difference was only amplified when several manufacturers adopted FDB spindles in 7200RPM drives, while leaving 5400s in the past with BBs. (U Series 6 weren't as durable as the Barracuda ATA IV, for instance; but at least Seagate had the sense to stop using BBs after the U6, and even the dinky little U Series X then had FDB motors.
Maxtor and Western Digital, on the other hand…)

[Behemot]

Yeah, just had a 160GB Barracuda, not sure which one, horrifyingly failed with about one third of the surface failed and the rest failing too. Just like that, it appeared OK, did write test, the next read test was a disaster. So not always only the multi-terabyte drives fail with gazillion bad sectors

[LongRunner]

Yeah, just has a 160GB Barracuda, not sure which one, horrifyingly failed with about one third of the surface failed and the rest failing too.

Well, ‘a 160GB Barracuda’ could be any of these:

PATA
ST3160021A (7200.7, 4 heads, 2MiB cache)
ST3160023A (7200.7, 4 heads, 8MiB cache)
ST3160212A (7200.9, 2 heads, 2MiB cache)
ST3160812A (7200.9, 2 heads, 8MiB cache)
ST3160215A (7200.10 slimline, 2 heads, 2MiB cache)
ST3160815A (7200.10 slimline, 2 heads, 8MiB cache)

SATA
ST3160021AS (7200.7, 4 heads, 2MiB cache) (OEM only)
ST3160023AS (7200.7, 4 heads, 8MiB cache)
ST3160827AS (7200.7, 4 heads, 8MiB cache, NCQ capable)
ST3160228AS (7200.7, 4 heads, 2MiB cache, SATA 300MB/s) (OEM only)
ST3160828AS (7200.7, 4 heads, 8MiB cache, SATA 300MB/s) (OEM only)
ST3160212AS (7200.9, 2 heads, 2MiB cache)
ST3160812AS (7200.9, 2 heads, 8MiB cache)
ST3160211AS (7200.9 slimline, 2 heads, 2MiB cache)
ST3160811AS (7200.9 slimline, 2 heads, 8MiB cache)
ST3160215AS (7200.10 slimline, 2 heads, 2MiB cache)
ST3160815AS (7200.10 slimline, 2 heads, 8MiB cache)
ST3160310AS (short-stroked variant of ST3250310AS?)
ST3160813AS (7200.11 slimline, 1 head, 8MiB cache)
ST3160318AS (7200.12 slimline, 1 head [short-stroked], 8MiB cache)
(I have the least trust in these last two, of course.)

So, that hasn't really narrowed it down very much …
(To be fair, even the better models could well be knackered after all these years, if treated harshly. At least, all are new enough to have FDB spindles.)

[LongRunner]

Samsung HD250HJ (125GB/surface, 2 heads, 7200RPM, FDB, CSS, SATA 300MB/s NCQ, 8MiB cache)
I have 3 of these; they're very quiet drives, both at idle and seeking (even with AAM off). Performance (at least in low-level tests) looks decent for the era.

Now, the thing that bothers me: The MCU on this HDD gets quite hot during sequential reading, even while the HDA is fairly cool. (At least it's in a QFP.) So if you want to keep using them, I'd recommend adding a thermal pad under the PCB, behind the MCU (Seagate did include them on many 7200.10s, though not on the 7200.11/12 ). Vcore on this measures at +1.43V, which is quite high for a relatively-modern HDD like this (the ST380817AS already had a +1.25V core!).

So, I think I still trust a Caviar Black or an older Barracuda more…

Samsung HD082GJ (80GB/surface, 1 head, 7200RPM, FDB, CSS, SATA 300MB/s NCQ, 8MiB cache)
One of the last 80GB desktop HDDs made (160GB 2-head versions also exist); a precursor to the HD250HJ, and almost identical-looking (and sounding).

The MCU still gets warm (as many do), but not as scorching as the HD250HJ's. (The data rates on this are only modestly higher than on the ST380817AS; this drive might actually use longitudinal recording, whereas the HD250HJ would have to use perpendicular recording.) Vcore measures at +1.4V.

And I'm pleased to report that Samsung gave up on their ‘half-minutes’ silliness; this drive reports 14,300 POH. So far it looks healthy.

Now, some bad news: I've just killed a WD800JD-00LSA0 (one of those exposed to the shed), by turning it around its spindle axis (its upper head suspension got bent out of shape by the stiction+rotation). Overall, they were still decent budget drives, but not as rugged as the Barracudas a few years before them.
(Although frankly, observing the amount of corrosion in that shed, it's a wonder that any HDD can survive there. It may become a PCB donor…)
My ST32140A recently died, too (somewhat suspiciously, when I plugged it into a JM20337-based bridge; I know the JM20337 has a problem with periodic data corruption, and it also operates from +3.3V but PATA originally used +5V signaling ). But my Barracuda ATA IVs and 7200.7s have survived!

For what it's worth, here are the weights of (some of) my HDDs:

• Maxtor 6L020J1 (×2): 549~550g
• Maxtor 6L040L2 (×1): 540g
• Samsung SP0411N (ultra-low profile) (×1): 447g
• Samsung HD082GJ (×1): 459g
• Samsung HD250HJ (×3): 457~459g
• ST320011A (w/o SeaShield) (×1): 555g
• ST340016A (w/o SeaShield) (×1): 546g
• ST360021A (w/o SeaShield) (×1): 573g
• SeaShield from a Barracuda ATA IV: 44g
• ST340014A (×3): 530~540g
• ST380013AS (×1): 545g
• ST380817AS (×1): 537g
• ST3120026A (×1): 566g
• WD800JD-00LSA0 (×6 including the dead one): 423~440g
• WD1600BB-55RDA0 (late version with 2-layer stainless-steel top cover) (×1): 559g
• WD5000AAKS-08A7B2 (I have quite a few, but I've weighed just the one mentioned): 596g

A curious observation: The earlier Singapore-made Barracuda 7200.7s had the neatest labels (61mm wide, same as the Barracuda ATA IV which AFAIK were all made in Singapore); then most of the China-made units have somewhat wider labels (70mm), and the Thailand-made units have ridiculously wide labels (79mm, making up a lot of padding; Seagate didn't bother putting the extra width to good use).