HDD mini-reviews

[LongRunner]

This thread is so you can post your HDD opinions and experiences. Set it in this basic format:
HDD model (and firmware version if desired) (key specifications – MB or GB/surface and # of heads (if not documented and not clear from examining the drive, enter “unknown configuration”; for Maxtors with model numbers ending in 0, check the second character in the serial number), spindle speed, bearing type (BB – ball bearing or FDB – fluid-dynamic bearing), contact start/stop (CSS) or load/unload (L/U), interface (including command queueing if supported), cache size, form factor if it isn't 3.5″*/1″ high)
Comments (transfer rate, access time, noise, etc.)

*The form factor is really misnamed. The width of the drive casing is 4″ and the disks in most 7200RPM and slower drives are 95mm. It was derived from the 3.5″ floppy disk drive which had the same external dimensions (though most 3.5″ HDDs have different mounting holes)…which was itself misnamed, as the floppy disk in question was actually 90mm. 5.25″ HDDs had 130mm disks, and the floppy disk bearing the name was 133 and a third. The drives had an outer width of 5.75″. The pre-PC 8″ floppy disks were actually 200mm. “Full-height” 5.25″ drives were 3.25″ height and “half-height” are 1.625″ high (and while long dead for floppies and HDDs, 5.25″ HH lives on for optical drives). Quantum Bigfoots were 0.75″ or 1″ high. 3.5″ drives are usually 5.75″ deep (the exception being Seagate's mid-1990s Decathlon family, later renamed Medalist SL, which in addition to being 0.75″ high was 5″ deep) and 5.25″ drives were originally 8″ deep, but modern optical drives are much shorter.

“Ultra ATA” was a mixed-up marketing term - there are PIO, DMA, and Ultra DMA transfer modes, which have the following bandwidth (ignoring overhead):

PIO mode 0: 3.33MB/s
PIO mode 1: 5.22MB/s
PIO mode 2: 8.33MB/s
PIO mode 3: 11.1MB/s
PIO mode 4: 16.7MB/s

Multi-word DMA mode 0: 4.17MB/s
Multi-word DMA mode 1: 13.3MB/s
Multi-word DMA mode 2: 16.7MB/s

Ultra DMA mode 0: 16.7MB/s
Ultra DMA mode 1: 25.0MB/s
Ultra DMA mode 2: 33.3MB/s (“Ultra ATA/33”)
Ultra DMA mode 3: 44.4MB/s
Ultra DMA mode 4: 66.7MB/s (“Ultra ATA/66”)
Ultra DMA mode 5: 100MB/s (“Ultra ATA/100”)
Ultra DMA mode 6: 133MB/s (“Ultra ATA/133”)

(There was also single-word DMA, but no-one really used that.)

Now for some of mine:

WD800JD-00LSA0 (40GB/surface, 2 heads, 7200RPM, FDB, CSS, SATA 300MB/s, 8MiB cache)
One of the “black top” generation. Access time is slow (≈19ms) with AAM on. With AAM off it's ≈15ms. Nothing too special there but small variations in access time can easily be overshadowed by differences in caching (StorageReview's tests show a big difference in application performance between otherwise-equivalent 2MiB and 8MiB drives, and many high-speed server drives are ill-suited to desktop use). Seek noise is just audible in use (with AAM off). I had 2 of these in my previous main PC – one for system and one for user data; I occasionally connect a few others.
(Technically, it ought to be called a “WD800JS” – not that it makes any real difference, of course…)

ST31000528AS (firmware CC38) (250GB/surface, 4 heads, 7200RPM, FDB, CSS, SATA 300MB/s NCQ, 32MiB cache)
The third drive in my previous main PC, which was used primarily for multimedia files. It wasn't heavily used by any means (it spun as long as my PC was on, but was seldom accessed); it held up okay for a while, but eventually reallocated quite a few sectors and also developed at least one uncorrectable error. It's a disappointment in general – spindle vibration is low, but seeking substantially vibrates the computer case and seek noise has an obtrusive tonality. Oddly enough, this drive reported itself as being AAM capable, but changing the setting did nothing. I can't recommend this firmware version – some of you might be lucky enough to have one with a different, quieter firmware, but buying one of these drives without knowing the firmware version is a gamble.

ST3120026A (firmware 3.06) (40GB/surface, 3 heads, 7200RPM, FDB, CSS, UDMA 5, 8MiB cache)
Still going strong after 34,398 hours of runtime. This was meant to be in an extra PC of mine, but I couldn't use that PC at the time (ran into a design flaw with Windows Product Activation) so I then removed it and copied some files across; I'm not sure what will become of the PC in question. Seeking is audible but neither sharp nor tonal; interestingly, it seemed louder when cold.

Quantum Fireball LCT20 10 (10.25GB/surface, 1 head, 4400RPM, BB, CSS, UDMA 5, 512KiB cache)
This drive runs without errors but the spindle bearings are quite loud (so much for “reducing noise” by slowing the spindle). The seeks are very quiet (even with AAM off), but very slow (19ms access), which is made worse by the small cache. I've been on a system with the 20GB version and it was unbearable.

WD360GD-00FLA1 (18.5GB/surface, 2 heads, 10kRPM, FDB, CSS, SATA 150MB/s TCQ (bridged), 8MiB cache)
One of the heaviest drives I have (748g), tied with the WD3003FZEX-00Z4SA0. Idling is quiet enough but the seeking sounds like thunder. (I guess I shouldn't be too surprised – speed was the number one priority with these drives.) A pity, as I was thinking of running a system from that drive to check out the performance increase. NOTE: Do not confuse this drive with the WD360GD-00FNA0, which was the first-generation Raptor with ball bearings.

Maxtor 4R080L0 QuickView (40GB/surface, 2 heads, 5400RPM, FDB, CSS, UDMA 6, 2MiB cache)
Came from a video recorder. Has 106 reallocated sectors. With AAM off the seeking is loud and sharp, but the access time is still 17.5ms. With AAM on it jumps to 22ms. Basically, the DiamondMax 16 line is the spiritual successor to the Seagate U6. (See also the results for the 4R160L0 in StorageReview's database. Compare with Samsung SV1604N and the 7200RPM + 2MiB drive of your choice.)

[LongRunner]

Fujitsu MPE3064AT (3.4GB/surface, 2 heads, 5400RPM, BB, CSS, UDMA 4, 512KiB cache)
As mentioned in another thread, the runtime counter fails S.M.A.R.T., and the bearing noise is unbearable. So much for Fujitsu's “new top cover that cuts sound levels” – it doesn't do any good if you run the drive for the equivalent of over five years non-stop. I don't remember how loud the seeking was, but I really don't feel the need to re-test. Fujitsu lied about the transfer rates on these drives (as did Maxtor at the time) – they say it's good for 17.2—30.4MB/s, but in reality it measures 11.7—18.9MiB/s (12.3—19.8MB/s in powers of 10) in HDTune.

ST32140A (270MB/surface, 8 heads, 5376RPM, BB, CSS, PIO 4/DMA 2, 128KiB* cache)
As mentioned in that same thread, this drive makes weird seek noises even when sequentially reading, and it also emits a lot of coil whine even after reaching full spindle speed. It weighs approximately the same as ST3120026A (the Desk Reference says a bit more (680g vs. 635g), but that same value is also given for the 5-disk ST36450A so it's probably just laziness on Seagate's part). Measured STR of 2.6—4.9MiB/s; access time of 17.5ms which is on the slow side, and not as good as claimed. Predates S.M.A.R.T. so I can't tell how good it is in every regard.
*Seagate made versions of this drive with either 128KiB or 256KiB of cache, under the same model number; so to tell the two apart without connecting them to a computer, you have to look up the part number on the RAM chip.

ST32122A (541MB/surface, 4 heads, 4500RPM, BB, CSS, UDMA 2, 128KiB cache)
Surprisingly quiet bearings for a drive that old (though does have a bit of coil whine) but loud seeking. No problems with S.M.A.R.T. IIRC. Access time again falls a bit short of claimed. This drive is from a line with a unique appearance – to be specific, with no screws holding the top cover on, only aluminium tape.
(No doubt as a cost-cutting measure; initially on the Conner CFS270A, then it carried over when Seagate bought out Conner.)

Maxtor 90288D2 (1.44GB/surface, 2 heads, 5400RPM, BB, CSS, UDMA 2, 256KiB cache)
Very light and very noisy. Maxtor says it's good for up to 15.2MB/s. While not as outrageous as with the MPE3064AT, it still only manages 7.7—12.8MiB/s in reality. It has two reallocated sectors. For some strange reason, the manual lists two 6-head models – one with 8,400MB and one with 8,455MB. Why they would make both of those is a mystery, at least to me.

[LongRunner]

The side rails on my ST310014ACE (10GB/surface, 1 head, 5400RPM, FDB, CSS, UDMA 5, 2MiB cache, 20mm height) were noticeably "wavy". Every other drive I remember seeing has straight edges. The head landing made a loud (compared to most other drives) scratching noise. Add to that the heat-spreader-less motor driver and the slow seeking - it fell short of the specified seek time, which is already 12.7ms (+ 5.6ms latency = 18.3ms access; mine measures 19.5ms access) - and certainly justified the name "U Series X", considering how slow the U5 and U6 were. And don't forget the easily damaged tape sealing (actually, it's not so much tape as it is a sheet of adhesive foil covering the entire top of the drive, with the label on top). For the record, the solder isn't as shiny as it usually is on Barracuda ATA IVs (and yes, this model does predate RoHS; I don't believe Seagate updated their pre-RoHS series to comply). This drive was quiet and had low vibration, but was otherwise unsatisfactory IMO (except maybe for an Xbox - but unless someone before me troubled to "unlock" the drive, this unit isn't from one). And the manufacture date on this one is in the second half of 2004 - what sensible PC user would have bought a 10GB drive by then???

I also have a WD1600BB-55RDA0 (41.7GB/surface, 4 heads, 7200RPM, FDB, CSS, UDMA 5, 2MiB cache).

[LongRunner]

Apparently you can't reliably run even UDMA 2 with a 40-wire ATA cable.

I decided to measure the overhead of Ultra DMA by forcing a (relatively) fast drive to run at a low interface speed and then comparing the measured STR with the "raw" clock rate of the transfer mode in question. So I got a 40-wire cable and plugged in my ST3120026A with it.

In theory, doing so will run the drive in UDMA 2. But when I tried it...see the attachment. Windows therefore dropped it down to UDMA 1.

I measured the STR anyway (across only the first 1GB of the drive), and accounting for the 2^20/10^6 discrepancy, the measured peak is 18.7MB/s, which gives an interface efficiency of 75%. Extrapolating would therefore give a realistic throughput for the higher UDMA modes of:

UDMA 2: 25.0MB/s
UDMA 3: 33.3MB/s
UDMA 4: 50.0MB/s
UDMA 5: 75.0MB/s
UDMA 6: 100MB/s

(On a side note, I've twice had VIA chipsets drop drives down to PIO mode for no good reason.)

[LongRunner]

ST3250820A (firmware 3.AAE) (83.3GB/surface, 3 heads, 7200RPM, FDB, CSS, UDMA 5, 8MiB cache)
Quiet seeking but high vibration. Has 37 reallocated sectors despite very little runtime.

ST380013AS (firmware 3.18) (40GB/surface, 2 heads, 7200RPM, FDB, CSS, SATA 150MB/s (bridged?), 8MiB cache)
Seeking fast but loud. Unlike most other drives (including the PATA models of the same series), it won't spin up without a host connection. I think Seagate lied about the family having native SATA as there's an extra chip on-board next to the data connector.

ST380817AS (firmware 3.42) (40GB/surface, 2 heads, 7200RPM, FDB, CSS, SATA 150MB/s NCQ, 8MiB cache)
Seeking like the older version. This is a true SATA design and will spin up with power alone. The zoning was also changed, resulting in a marginal increase (of ≈1.2MB/s) in O.D. STR and a similarly insignificant decrease (by ≈0.5MB/s) in I.D. STR. (As an aside, the OEM versions of these, confusingly, used the old model numbers. But you can tell them apart by looking at the PCB. There was also an obscure OEM-only SATA 300MB/s version.) Also, as this drive uses buck converters (rather than linear regulators) for the +3.3V (SDRAM) and +1.25V (Vcore) supplies, it runs noticeably cooler than the ST380013AS.

[LongRunner]

I just found out that the ST32140A is rated for 4.1 bels at idle, so no wonder it whines like no other HDD (at least that I have).

According to the specifications, it uses identical platters to the (2-platter, 0.75" height, 5" depth) ST51080A (Decathlon/Medalist SL), however the smaller drive is rated as being less noisy (30dBA idle). I don't have one of those, though, so I can't say how it compares.

[LongRunner]

Fujitsu M1614TAU (270MB/surface, 4 heads, 4500RPM, BB, CSS, PIO 4/DMA 2, 64KiB cache)
Seeking is amazingly quiet for such an ancient drive (not that it matters anymore, considering the bearing noise), but as slow as it is quiet. Predates S.M.A.R.T. but had some uncorrectable sectors. There was a group of 3 near the end that wouldn't go away, even after zero-filling. From the incomplete performance results I got, access time is ≈19~20ms. After testing, I removed the top cover.

[LongRunner]

WD3003FZEX-00Z4SA0 (300GB/surface, 10 heads, 7200RPM, fixed-shaft FDB, L/U, SATA 600MB/s NCQ, 64MiB cache)
My current main PC uses two of these, and I have a third set aside. My intention was to set up a RAID 1, with the third unit as a separate backup drive; but mistakes made during the Windows installation left me with the double whammy of the bootloader stuck on one drive with my stuff on the other (thus preventing RAID 1 from being used), and the drives set up with þe olde MBRs (leaving the third terabyte inaccessible); reconfiguring the system will be too much trouble until I have a second PC (that works, and hopefully runs an OS modern enough to support GPT) to use as an aid (I don't know when that will be). On the plus side, the drives are actually fairly heavy — certainly by comparison to recent (or otherwise cheaply made) budget models, although that's in part due to the number of platters; so maybe they'll actually last. However, the seeking is rather loud, and as AAM is officially dead, there's no way to reduce it.

[c_hegge]

Yah. I've found that too with the WD Black desktop HDDs. They are very chattery. They seem OK otherwise and do come with a 5yr warranty (as opposed to 2 or 3yrs from others).

[LongRunner]

So far, so good: 33,592 hours and counting (with 119 start/stops, 101 power cycles, 19 emergency retracts, and 99 load/unload cycles) on my "main" WD3003FZEX-00Z4SA0, with no problems. So it appears that good, reliable HDDs can still be made at reasonable enough prices (AU$225 each for mine, IIRC).

They seem to be a decent spiritual successor to the Barracuda ATAs of old, then. And while my original ST380011A may have been knackered, my aforementioned ST380013AS and ST380817AS (manufactured around the same time) seem to still be healthy (no bad sectors, nor other faults, on either).

It's interesting to note that for the Barracuda ATA IV/V and 7200.7, Seagate not only said nothing against keeping the drives on 24/7, but even suggested using them for:

Best-Fit Applications

• Mainstream and High-Performance PCs
• Entry-Level ATA Servers, including RAID
• Cost-Effective Network Attached Storage
• PC Gaming Systems
• Home Media Servers
• Near-Line Storage

So Seagate's newer marketers clearly either don't remember their past, or might even be deliberately ignoring it.
(I mean, 2,400 POH/year for the 7200.14? That's a mere 12,000 POH in total, if you keep it for 5 years; or then <17 months continuously. )

The Barracuda ATA IV/V and 7200.7 sure look and feel like something fit for enterprise use, too – complete with some of the best soldering I've ever seen (except for the PCBs on the 7200.7, where Seagate went to fast-track the transition to RoHS compliance; still, the 7200.7 HDA retained leaded solder for a while, and was just as impeccable as before). (Seagate even made a SCSI adaptation of the Barracuda ATA IV, namely the Barracuda 36ES2.)

(The U Series were Seagate's true consumer models of the time, as used in department-store PCs; although even they were fairly reliable, if notoriously slow.
I believe a past pair of main PCs in my Mum's accounting office each contained an ST340810A, and sounded ill in their latter years possibly due to media degradation; and the office server at the same time used 2 × Maxtor (OK, technically Quantum) 6L060J3, which eventually seized. Had the builder instead paid a little more to get the ST340016A (main PCs) and ST360021A or ST380021A (server), I'm sure they'd still be spinning.)