Earlier sections to be written: The Wire Recorder, Magnetic Tape
Hard Disks for Speed and Reliability (1956)
After 3 years of passionate experimentation, IBM's RAMAC (Random Access Method of Accounting and Control) 350 was the first production HDD; storing 5 million characters (about 3.75MB) on a whopping 50 platters, 24 inches in diameter spinning at 1200rpm. Interestingly (in contrast to the collossal growth in density and dramatic shrinkage in form factors), platter edge speed has remained surprisingly constant ever since – 38.3m/s then, 36.8m/s on the Seagate Elite's 130mm platters at 5400rpm, 35.8m/s on the Barracudas' 95mm platters at 7200rpm and 34.0m/s on the Savvio 10K's 65mm platters at 10krpm.
Rented out for US$3,200/month in 1956 (use your own inflation calculator), this ancient clunker gave way to ever more-capacious and ever-faster successors, and despite regular challengers, enshrined IBM's leading position in HDDs since (until 2003 of course, after the Deathstars lost it all
)Floppy Disks – A Halfway House between Tape and HDDs (1971)
But smaller computers didn’t yet need the transfer speed provided by HDDs, and head crashes were morbidly‑feared back then; so it was easier, cheaper and safer to retain the flexible medium and contact heads as used by tape recorders, only in a slowly‑spinning disk form for reasonably‑quick random access.
As a further economy measure, the disks were removable in standard sizes and formats, interchangeable between drives from any manufacturer.
The first floppy disks were of course IBM's big old (and truly floppy) 200mm (“8 inch”) models, followed by Shugart Associates' mostly‑direct shrink to 133⅓mm (“5¼ inch”) and then the iconic Sony‑developed 90×94mm (“3½ inch”) type (nowadays immortalized as the Save icon in countless programs) with a much‑sturdier casing, protective shutter and improved usability. Various competing formats came and went, but Sony’s standard persisted into the mid‑2000s.
Even so, floppy disk drives were still an expensive luxury in the 1970s and early 1980s (in many parts of the world) – Compact Cassettes could store sufficient data for earlier microcomputers, and their loading times were bearable for the era’s small programs (and RAM capacity).
It was only in the late 1980s that floppy drives really became essential (followed by HDDs in the early 1990s)…
Although floppy disks (just like tapes) inevitably wear out after enough usage (given the heads’ contact with the spinning disk), those from the 1970s and 1980s were generally well‑made and many have survived beyond the bravest predictions given back then. Inevitably however, their commodity status (and lack of updates since the rare 3.5″ 2880KiB ED variant, most users sticking with the 1440KiB HD standard) brought down quality standards together with their prices through the 1990s, and the last batches in the 2000s were complete and utter junk.
On a side note, a few novelty gadgets (notably the Seiko Executive Egg) actually contained a (non‑removable) floppy disk for analog sound recording.
The Microcomputer Age: Enter Seagate (1980)
1977’s Apple II computer was a smash hit and a half, bringing all kinds of traditionally-corporate computing tasks into the home and office, despite the MOS 6502 CPU’s meager processing power. Although Apple Computer themselves never quite outdid its success, IBM broke from character and worked on a little side‑project; the engineers had to meet IBM’s stringent quality standards, but otherwise had free-rein to do as they pleased. Hence emerged the IBM Personal Computer (model 5150) in 1981; it succeeded where microcomputers from other “Big Iron” companies failed miserably (largely because IBM didn’t deliberately cripple its computing power), and despite strong subsequent competition from 1982’s Commodore 64, the IBM PC had one compelling advantage – its individual peripheral cards (while intended just to ease repair) made for very easy and flexible upgrading too, an invaluable asset during this time of frantic hardware advances. The IBM PC quickly became more‑successful than all previous IBM products combined…
But IBM’s own HDDs were still squarely in the Big Iron category – humongous all around (in capacity, physical size and emphatically cost). The time was high for start‑up companies to make smaller HDDs, particularly in the 5¼″ form factor currently occupied by the floppy drives. Seagate Technology made the move with their 5MB ST‑506 in 1980 – using two 130mm platters (hence 4 heads) with 153 cylinders (which can be formatted with either 32×256‑byte or 17×512‑byte sectors per track – the latter providing a 6.25% capacity boost at the cost of greater “slack” after small files), it found success in DEC’s Rainbow computer.
In 1981 Seagate doubled to 306 cylinders, releasing three new models – the 10MB 2‑platter ST‑412 being most popular by a wide margin, but there was also a flagship ST‑419 (15MB, 3 platters) and entry‑level ST‑406 (5MB, 1 platter) accompanying it. Even though other companies released solid competitors (some even of superior quality – particularly Disctron), Seagate’s model won out for acceptable reliability at the right price point.
Despite already manufacturing their own higher‑quality 5¼″ drives (the 10MB 2‑platter WD‑12 and 20MB 4‑platter WD‑25 – WD standing for “Winchester Disk” and unrelated to Western Digital, who then only manufactured controller cards and other electronics, not the drives themselves) in 1982, IBM graciously welcomed Seagate’s ST‑412 into their own model 5160 PC/XT (Personal Computer/eXtended Technology) in 1983, and Seagate’s business grew and grew.
In 1984 their ST‑225 combined another capacity doubling (615 cylinders for 20MB) with half the height (the same height still used by optical drives) and a faster rotary swing‑arm actuator, and 1986’s ST‑251 squeezed in a third platter (40MB over 820 cylinders) and added the much needed auto‑parking feature (standard in voice‑coil enterprise HDDs, but tricky to implement with the stepper motors in consumer models) – although the related ST‑277R (60MB with RLL 2,7 instead of traditional MFM encoding) was pushed too far and gave errors like nobody’s business…
Before much longer Seagate grew wealthy enough to aim above their humble MFM/RLL drives, and even acquire traditional enterprise maker Imprimis Technology in 1989. That same year they launched the scorching Elite series of 5400rpm, 5¼″ full‑height giants; in 1992, the famous 7200rpm 3½″ Barracuda (accompanied by the 5400rpm 3½″ Hawk); and in 1997 the ferocious 10,000rpm Cheetah. For some reason Seagate ran out of new product names by 2000 – calling their 15,000rpm pioneer just the “Cheetah X15” – but it maintained Seagate’s tradition of providing first‑of‑its‑kind performance nonetheless.
Even more‑valuable than their SCSI leadership, however, was Seagate’s commitment to raising the ATA bar ever‑higher – in 1995 they made the first 5400rpm ATA drive, the slimline‑but‑speedy Decathlon 850 (ST5850A) matched only by Quantum’s Fireball 1080 soon after; and in early 1996 they briefly made a “Hawk ATA” (albeit under the regular Medalist name), the 2.1GB 5400rpm ST32140A. In 1998 they went even further with the Medalist Pro 9140, the first 7200rpm ATA drive and the first ever with fluid‑dynamic bearings (FDBs) – although its high heat (11W at idle!) did cause some catastrophic failures when inadequately cooled, so Seagate discontinued its ATA models (continuing only the SCSI variants) for the year after (so they could further refine the technology).
In mid‑1999 the Medalist 17240/17242 (the last of four mainstream 5400rpm Medalist generations) gave way to the excellent (and huge, up to 28.5GB) 7200rpm Barracuda ATA, delivering every bit of the traditional SCSI Barracudas’ performance. While the Barracuda ATA II was well‑matched by Maxtor’s superb (and unusually, more reliable) DiamondMax Plus 40, the Barracuda ATA III was much more‑sturdily built with a mostly‑mature FDB motor option (as advertised in the datasheet); it’s just a pity Seagate’s decision to enable AAM by default (back when almost nobody else even knew of it) curtailed its place in benchmarks, back when speed was still their main selling point.
But Seagate quickly turned that quietness into their greatest asset with the outstanding Barracuda ATA IV – the first 7200rpm drive to 40GB/platter; FDB as standard, quieter than any then‑current 5400rpm model (including their own U Series 6); as fast as most competitors (substantially-outperformed only by Western Digital’s 8MiB‑cached Caviar Special Editions) – and more reliable than IBM’s best (built rock‑solidly; with the famous Quantum‑originated AirLock® and upsized headstack bearings so that despite the fairly‑normal rated 350G, at least while non‑operating it was easily as tough as most laptop drives of the era; top‑quality PCBs, with soldering so beautiful it would make NASA proud; and implied by the POH counter to really have a 10‑year design life, with many ATA IVs through 7200.8s easily surviving well past even that). Seagate was thoroughly committed to beating the (now fallen) Deskstars at their own game…
The Barracuda ATA V was the industry’s first model with SATA variants, and by July 2004, Seagate backed‑up the Barracuda 7200.7’s continuing bulletproof reliability with an enterprise‑standard 5‑year warranty on all of their internal drives (showing everyone what they thought of Maxtor’s “MaXLine” upselling ploy)…
As density growth slowed while demand refused to relent, Hitachi (new owner of the former IBM Storage division) brought back their 5‑platter design with mid‑2004’s Deskstar 7K400 – now with FDBs, it was a huge, fast and expensive beast. Seagate’s response was unprecedented again – they finally beat IBM/Hitachi to a new head technology, Tunneling Magneto‑Resistive (TMR). By late 2004 (although it only reached general availability by early–mid 2005) this enabled Seagate to match Hitachi’s 400GB with just 3 platters, in the extraordinary Barracuda 7200.8 – and that augmented by thermal flying‑height control for an uncompromised 0–60°C operating range (a year before Hitachi took credit for it, at that). This was also, however, when Hitachi and Seagate went separate ways – previously the entire HDD industry adopted new head techs (MR and GMR) within a year or two of IBM; a bit after Seagate’s TMR, Hitachi introduced their IrMnCr (enhanced GMR) which is more‑stable in principle, but they still have yet to outdo the old Barracudas’ durability in practice, so interpret that how you will…
Another year, and Seagate was the first to ship perpendicular recording in the Momentus 5400.3 (and then Barracuda 7200.10).
Another late‑2004 release was commissioned by Hewlett‑Packard – since 10K and 15K drives like the Cheetahs have reduced platter diameters anyway, why not shrink the whole drive to 2.5″ (like a laptop drive, only thicker for robustness)? This produced the Savvio 10K.1 – super cute, even faster than the Cheetah 10K.7, and undoubtedly more reliable too (given its conservative platter edge‑speed, and the benefit of being premium‑priced as the first‑of‑its‑kind again).
If not for Seagate growing increasingly aloof (and losing all pride in mainstream (S)ATA after acquiring Maxtor’s mis‑management…), we would have surely seen a “Savvio SATA” (with desktop‑optimized firmware) a few years later (in 2008 Western Digital brought out their VelociRaptor which handily delivered the performance, but was initially encumbered by major firmware bugs)…
Despite Seagate’s Maxtor‑acquired ruthless penny‑pinching in modern “mainstream” models, however, they retain skilled engineers and occasionally still make an honorable product – 2010’s “Barracuda Green” (ST2000DL003 & ST1500DL003) was a surprising improvement over 2009’s unreliable CSS‑encumbered Barracuda LP, finally adopting ramps but not abusing them with the Caviar Greens’ suicidally‑short (8 second) parking timer. It was also explicitly rated for 24×7 operation (which Western Digital subsequently pigeonholed into the “NAS” market with their Red series), and even though calling it a Barracuda was very optimistic, it was a solid revival of the U Series tradition (barely‑adequate, but not designed to fail deliberately nor catastrophically). The 7200rpm 2.5″ nearline Constellation drives (channeling IBM's Star theme
) are just as adorable as their Savvio cousins, real power‑sippers and whisper‑quiet too (while matching the Barracudas’ seek speed), great for SFF desktops (as unlike the Savvios they do come in SATA), and available in usefully high capacities with the proper 5‑year warranty.It is true that IBM, with their 25‑year head‑start, avoided some of Seagate’s smaller mishaps – IBM already knew to avoid STMicroelectronics (which afflicted U4/U8/U10 and a lot of Barracuda ATA Vs and onward with substandard read/write channels – although Seagate took responsibility and upgraded their specification by late 2004 or early 2005), and had the sense to leave AAM off by default (on the 75GXP and successors). They would probably also have retained the will to fight off Convolve (who patent‑trolled and sued Seagate into dropping AAM from the Barracuda 7200.7 onward – let’s remember that in 2001 and 2002, every manufacturer supported AAM). But overall, Seagate won in raw engineering skill…
Even more importantly, Seagate was dearly loved as the last well‑rounded manufacturer (strong in both SCSI and ATA) – in the 1990s we also had IBM (renowned for their MR/GMR head leadership and near‑zero failure rates on earlier Deskstars, at least back in their day) and Quantum Corporation (their 5400rpm Fireball EL/EX/CR/CX of 1998/1999 providing best‑in‑class performance and durability, just as the 7200rpm Barracuda ATAs did shortly after), but IBM promptly fell with the Deathstar 75GXP/40GV & 60GXP (subsequently cheaping‑out their 1‑platter drives from the AVVN 120GXPs onwards) and Quantum – although their decline was relatively benign – did have some electronic faults in the Fireball Plus AS (and its succeding “Maxtor” D740X), necessitating a head‑swap.
(In common among IBM, Quantum and Seagate was that their lucrative SCSI businesses could subsidize the ATA divisions especially during times of hardship, maintaining their quality standards; whereas Maxtor and Western Digital, compelled to actually profit in ATA, were occasionally forced to loosen their QC.
Since buying out HGST, the pressure may have eased slightly on WD’s end though…) Fujitsu made bulletproof SCSI/SAS drives, but their ATA drives (desktop and laptop) were just average consumer models (and Toshiba’s laptop drives were barely‑passable at all, given their 1‑year warranty and 20,000 POH limit).
Seagate’s Momentus series were the first laptop drives ever to get that precious 5‑year warranty – and were succeeded only once Western Digital filled the post‑7200.11 void with their Caviar Black (enthusiast desktop) and Scorpio Black (7200rpm laptop) lines, both reintroducing the 5‑year warranty on all capacities.
SyQuest, Zip, Jaz and REV – More Removable‑Media Compromises (late 1980s—2004)
Even though the modernized voice‑coil HDDs (especially Western Digital's excellent early Caviars) were vastly faster and already proving more reliable than floppy disks, two key factors held them back from widespread use for removable storage – the substantial cost of the supporting components (heads, motors), and the inherent fragility of the heads and ball bearings (both in the spindle motor and headstack pivot).
This is where SyQuest came in – like HDDs, they had a fast spindle motor, and flying heads moved by a voice‑coil actuator; but like floppy disks, the media cartridges (with a single hard platter) were removed from the drive when not in use (the heads safely parked on a ramp).
Iomega decided to economize on that idea, by returning to a flexible disk with their Zip cartridge system (1994) – it seemed well enough at first, until those economists got carried away in 1998 and removed the donut‑shaped washer cushioning the head against parking impacts.
Widespread misalignment, the dreaded Click of Death and a class‑action lawsuit ensued…
In 1996 Iomega took a converse tangent with the Jaz cartridge, an attempt to one‑up SyQuest by packing two platters into the cartridge – but shock often knocked them out of alignment. Overall, Iomega’s offerings were seductive on paper, but the SyQuest EZ drives were much better‑built and (until the very end) far more reliable than the Zip or Jaz systems.
As densities kept growing, the alignment difficulties and limited dust protection of removable cartridges became ever‑more‑serious problems – as did rapid obsolescence of the drives. Despite people swiftly moving onto flash memory (for small-capacity shockproof storage) and true external HDDs (for full‑speed, reliable and dust‑protected capacity), Iomega made one last‑ditch attempt with their REV cartridge system in 2004; it was again one step closer to real HDDs, by building the spindle motor into the cartridge, but remained an unmitigated failure to put it lightly…
Solid‑State Drives: Threat or Companion? (late 2000s—present)
The massive advantage in random access time of solid‑state storage was already valued by big enterprises, but (like HDDs themselves were from the 1950s–1970s) prohibitively expensive for home and office users in useful capacities. Around 2008, though, ever‑growing capacity and declining costs in flash memory made it feasible for professional workstations and home enthusiasts to boot from SSDs of useful capacity (typically 30GB or 60GB).
Some cheaper SSDs were unfortunately encumbered by unbuffered controllers (particularly from JMicron) which caused occasional long pauses (sometimes up to a whole second), but enthusiasts were easily able to avoid these and opt for properly‑buffered models.
The shock resistance was also a big advantage in notebooks, but their capacity was still insufficient for bulky music and videos – the 1TB Hitachi Deskstar 7K1000 already released in 2007. Hardcore gamers, meanwhile, demanded speed and capacity to load their games as quickly as possible; so the race was on to make the fastest, biggest and cheapest SSD anyone could.
This came at a serious penalty to longevity, unfortunately; early SSDs retained their data well, as they had relatively‑large memory cells, and only had to distinguish two (SLC) or four (MLC) voltage levels. As cells shrunk and manufacturers squeezed more bits into each cell, write‑cycle endurance and data retention both suffered – the nowadays‑mainstream TLC (3 bits/cell) requires distinguishing 8 voltage levels, and QLC (4 bits/cell, increasingly abundant in microSD cards and other high‑density/low‑end applications) has no less than 16 levels. If not refreshed regularly (at least every few years, preferably yearly; quality SSDs should do this automatically during idle time, cheaper models may not), the data will fade away fairly quickly…
3D NAND (which stacks multiple layers of memory within the chips, allowing larger and more‑durable cells for the same capacity in the same chip package) provided a short‑term reprieve, but is still being pushed harder‑and‑harder in the latest (and cheapest) consumer SSDs.
Under merciless price pressure, major firmware bugs and other catastrophic failures (relatively uncommon a few years earlier) have also become rampant in new models – even the previously‑formidable Samsung and Crucial have finally fallen.
On the upside, proper enterprise‑grade SSDs (previously much too expensive for the mainstream) are finally down to quite reasonable prices (<AU$1/GB for the Exascend SE3, as I ordered via Mouser), so businesses and enthusiasts alike can purchase them and rest easy. If you’re a hardcore gamer on a budget, you can implement a 3‑tiered storage system – a modest enterprise SSD (120GB or 240GB?) for the OS and productivity applications; a large consumer SSD to install the games to; and an (even larger) enterprise HDD (or “enterprise-adjacent” e.g. Caviar Black, FireCuda) storing the installers and other bulk data.
For the moment, it’s true that (decent‑quality) HDDs are almost dead in the mainstream “consumer” market; but they remain as strong as ever in enterprises (and even among enthusiasts with massive storage requirements), and maybe – just maybe – they’ll see a more-general Renaissance once people learn that the consumer SSDs they naïvely trusted have now become just as unreliable as Maxtor HDDs were near their end (DiamondMax Plus 8/9/10/11 and their “enterprise” variants, MaXLine Plus II/III/Pro); returning to the once‑quintessential smaller (but quality) SSD + bulk HDD combination.
Back then, we had fun and took delight in putting dying Maxtors out of their misery – will that same passion return?
And however you measure it, quality HDDs still handily beat current SSDs in long‑term offline data retention – with the classic Barracudas generally still in good (often perfect) health >20 years later, under comfortable conditions I would trust them to store as well as the magnetic audio tapes (both reel‑to‑reel and Compact Cassette, although the same can’t generally be said for the clumsy 8‑track cartridges) which have survived since the 1960s and 1970s…
It remains to be seen if service lives (and warranties) >5 years will be formally negotiated and become official – but still, considering that the ST‑10 Barracudas topped out at 750GB and new air models are up to 10TB (12TB would be feasible if Seagate retained their 6-platter design), I can’t complain too hard…
