It's the oldest working Seagate drive in the UK Seagate reckons it has found the oldest working Seagate disk drive in the UK: a 28-year-old ST-412 disk drive from 1983. It is the drive for an old IBM PC, which booted up when it was brought down from owner Mitch Hansen's attic in his Ruislip house. The 5.25-inch disk has four platters, eight read and write heads, spins at 3, …
i want the geek crown because it was an Amstrad PC was the first PC that was ever built with the millennium bug in mind..
other pc's at the time were compliant, but was more an accident . Amstrad built the a pc that in its design spec was that it must be Y2k compliment. other machines at the time and some even afterwards were designed still not giving a rats arse about Y2K as they believed the machine would not be in use that far in the future !
just knowing this makes me king of the court of geeks !
Excellent story, which raises an intriguing issue with disk drives, ie. capacities have increased exponentially, whereas access speeds only linearly. Will the same be true of flash ?
More generally, Moore's law continues apace. However, due to vast increases in software complexity, a 2011 PC doesn't do *that* much more than a 1995 PC, aside from playing games. I mean, they both run word processors and spreadsheets. They will both produce a letter to your solicitor or do your accounts. The 2011 PC does accomplish more, but not a thousand times more.
Not a thousand times more than 1995?
Well, apart from the fact that you can pretty much just plug in a scanner, printer, tablet or what-have-you and it will self-configure the thing instead of your having to muck about with IRQs and drivers for hours.
I guess it depends on what you are counting when you make the measurements. If you take a sensible measurement, like how much the computer is really doing while you scoff crisps and watch the install screen for your games, or what has to happen for your WoW Wizard to cast his spells with appropriate visual FX, then yes it does, rather.
> Not a thousand times more than 1995?
>
> Well, apart from the fact that you can pretty much just plug in a scanner, printer, tablet or what-have-you and it will self-configure the thing instead of your having to muck about with IRQs and drivers for hours.
>
That is a function of bus design that has very little do do with whether or not the clock on the CPU is at 60Mhz or 3Ghz.
Would this be one of the classic jet engine turbine drives? i.e. one that whirs, clunks and whines itself at startup until finally it somehow manages to settle into an amost tolerable noise?
Had to love those old Seagate drives, and just old drives in general. With their assortment of interface types (MFM anyone), disk capacity IDs, and all the general pain of IRQ assignment - want your MFM adaptor card to work at the same time as the serial cards and the keyboard? Get your graph paper out, cross reference the IRQs that each will use and try to find a best fit!
Hmmm,
I thought MFM wasn't an interface type... There was ST-506 (SASI?), ESDI, the old narrow SCSI that I used during my early career.
But I do remember bodging of IRQ's, including the use of 16bit IRQ's on an 8 bit ISA card... I'm fairly sure little bits of wire and solder were utilised. You could do things then that you probably couldn't get away with now because of the much higher clock speeds.
wasn't an interface type, it was an encoding type. Back in those days, there wasn't room on the drive for the actual controller/encoder so the add-in card you got (usually with the drive) was the equivalent of the board on the bottom of your modern drive plus the interface circuitry on the motherboard (there being, in general, no built-in HDD interface on motherboards back then). ST-506 was the interface, I believe. MFM and RLL were the encoding types for the drives. My dad and I had hours of fun trying to reformat MFM drives on RLL controllers to get that extra capacity. I think we even succeeded once...
[<Back in the day Icon?>]
I had two 40 MB MFM drives and reformatted them both on an RLL controller and both worked perfectly. One was a Seagate and the other was a Miniscribe. The Seagate bumped up to just over 61GB and the Miniscribe only made it to 55 or so. Ran a BBS on those two in an old homemade 286 box for a couple years, and never had an error.
Any storage medium is going to be limited by how fast you can get data in and out of it. Flash's limits are pretty high though, so it's not such an issue.
More of a problem for Flash is the temp file thing. Each time you write to Flash, you wear it out slightly, and eventually it stops working. The same is true of hard drive sectors, of course, so the same solutions will work, but the number of write cycles is a lot better for magnetic media. When your OS and all your apps dump a shitload of temp files onto your drive every time they start up, this is not a nice place to be if you're a Flash drive. Likewise the whole RAM paging thing where the PC uses the hard drive as a temporary RAM extension is not going to be good for a Flash drive.
The "doesn't do much more" thing is a big problem for software companies. Why upgrade from Word 97 if it still works well enough? Hence the recent MS addition of that godawful ribbon thing, which has to be the stupidest user interface in the history of computing.
The problem with Flash cells wearing out is an issue, but the storage boffins have long since come up with a workaround: wear levelling. Simply put, the flash card tries to ensure that read/write activity is averaged out over the entire space, rather than being focused on specific cells. And with flash storage now comfortably into the gigabyte space (I have a 32gb micro-SD card: I can carry the entirety of Wikipedia several times over on something smaller than my fingertip. This really is the future...), the drives are generally big enough for this to be an effective approach, unless you're constantly writing to the entire drive (e.g. video recording). Even then, you're generally looking at a minimum of around 10,000 write cycles, which is a whole lotta data...
Back to the article. We've tripled the speed of the platter (10,000 rpm disks are commercially available - around that's 150 spins per *second*) and increased information density - from 1.25mb/platter to 300,000mb/platter - or a quarter of a million times more information. That's the equivalent to printing War and Piece onto the back of a single first-class stamp with room to spare, which I find pretty damned impressive.
The point I think the final paragraph was trying to make is that we're coming up against physical limits for information density, spin speeds and reliability. Magnetically-based storage has come a long way since someone slapped some iron oxide onto a bit of sellatape (as nicely demonstrated on an old Secret Life of Machines episode), but it's time is slowly drawing to an end - I'd guess that we're still at least 5 years away from flash storage being commercially viable in the terrabyte range, but hey. I used to have a 2mb SD card (and that was a step up from the 1.44mb floppy some cameras used at the time) and now own a class-10 16gb SD card which quite happily records 1080i AVCHD video at 60fps...
Sure, a word processor does what it does, and aside from live reformatting, live spell checking and not having to worry too much about downsampling images before you include them, there is not much benefit in having a faster PC.
What does change as processor speed up is the number of tasks which suddenly become interactive. I worked in computer imaging when JPEG was coming out, and we needed custom hardware to compress moderate sized images in a reasonable timeframe. Who these days gives a second thought to opening a large JPEG? Or processing a 12MPixel image interactively - people used to work with small proxy images and process the larger image overnight.
Didn't it take 300 Sparc stations about 6 months to render the first Toy Story? I am not saying that you could do that on a PC (yet) but Blender's wire frame mode seems a bit redundant these days. Admittedly that owes as much to the graphics card as anything else.
There is always something new to do with extra clock cycles.
> Didn't it take 300 Sparc stations about 6 months to render the first Toy Story? I am not saying that you could do that on a PC (yet) but Blender's
>
By 1997 and perhaps even 1995, PCs were already being used in render farms.
Sparc CPUs were never anything to write home about in terms of performance. That's why D2 was using Alpha machines running Linux to render the effects in movies like Titanic.
People are having a hard time trying to figure out what makes a new machine better than an old one in practical terms. A lot of the stuff that we think of as "modern and new" has been around for awhile already.
"but disk I/O has become a bottleneck at the platter surface level, and is set to remain that way."
So if this is the case, why do they keep increasing the bus speed? Surely if the bottleneck is the platter, then the bus is not full, and making it wider does nothing. This doesn't sound right.
> So if this is the case, why do they keep increasing the bus speed?
Because that raises the rate at which you can burst data to the disk (for it to be cached in RAM).
This allows the CPU to go off and do other things while the HDD processor gets on with writing all that cached data to the platters.
As long as the mean bus data rate (averaged over the amount of time it takes to fill the cache) doesn't exceed the platter transfer rate, you're left with a disk subsystem that appears to write data at full bus speed - even though it's actually doing no such thing.
If the mean bus data rate goes too high, you end up with the CPU in IOWait until the drive catches up.
Vic.
> "but disk I/O has become a bottleneck at the platter surface level, and is set to remain that way."
>
>So if this is the case, why do they keep increasing the bus speed? Surely if the bottleneck is the platter, then the bus is not full, and making it wider does nothing. This doesn't sound right.
>
You can group multiple drives together. This is also a rather old idea.
Plus there's the possibility of newer tech. Just because your average n00b Mac user can't think of a way to use the extra capacity doesn't mean it shouldn't be developed.
SSD looks very promising and could be 60 times faster than spinny disk. The extra speed has usable potential because the underlying interfaces have the means to support it.
...days to hours my DVD jukebox array rather than days. Too bad I would need a 2nd mortgage to afford the equivalent amount of SSD storage.
back in the early 80's a 5MB winchester Hard disk of any brand would have set you back £5k+ and a 10MB a whopping £8-10K+.
i know cos i used to have to occasionally fix them around london in the late 80's as a field engineer.
people forget that a basic vanila ibm pc without hard disks would start at 5K and the only ones buying them were large corporate companys for thier accounts dept.
everyone else there would have to make do with dumb green screen wyse rs-232 terminals.
"back in the early 80's a 5MB winchester Hard disk of any brand would have set you back £5k+ and a 10MB a whopping £8-10K+".
Surely not. In 1986, you could buy an Amstrad PC clone containing a 20 Mb hard disk for £499. Moore's law is good, but not that good.
We had a 1512 in the house. Before that, my father *rented* an Epson PC, including hard disk, because buying was far to expensive. Like the way people used to rent TVs. Now, I guess people just use credit cards.
Wikipedia alleges £499 including hard disk. Suspect that might refer to the floppy disk model only. However, even i-programmer states the PC 1512 20 Mb model cost £949. It is therefore difficult to imagine the same size disk costing £5000 to £8000 only 3 years earlier.
Maybe there was a price difference then, as now, between "enterprise" and domestic disks.
This post has been deleted by its author
If we're talking manufacturers here, I didn't get much mileage out of a WD "Elements" 640 gig external drive. Approx. 14 months on and it wouldn't rustle up, although, to be fair, it wasn't the drive itself. It must have been the interface electronics. Took the beast apart, extracted the drive, installed it in an independent housing and has been OK ever since. However, have heard similar stories. Not confident in the brand now. Certainly wouldn't buy another, that's for sure.
The first server I installed was an Apricot with a 40Mb hard disk, with a dozen workstaions, later increased to 20.
The thing is these were all thin clients without floppies, so all software, user data, and the local email system, were all stored on the server, as was the only printer connection.
Say that to the kids who are system managers today and they'll look at you like an Old Fogey!
Min's the one with the thin ethernet crimp tool in the pocket.
This post has been deleted by its author
Might be a typo. I don't think anyone made a 10gb SCSI drive until 1997 or 1998. Even then, it was typically 8.7, give or take.
As for SCSI, the parallel bus topped out at 320MB/sec, which isn't much better than a 3 Gb/sec serial connection (let alone 6Gb). Then again, if you're running SAS, isn't that still SCSI?
I have a prototype Conner "Chinook" IDE drive that has two separate headstacks (located at opposite ends of the same disk assembly but that share the same platter stack) and a clear plastic lid. After Seagate merged with Conner it was given to me by some guys I worked with at Seagate in the mid 1990s. They were throwing it out with some other items.
Did anyone ever see a production Chinook drive? Supposed to be super-fast read/write times owing to the multiplexing of the two headstacks.
Relatively young at circa 1991 but I think my Chinook drive is more interesting than a run-of-the-mill clunky old ST-412.
Didn't last very long, as I recall. Seem to remember there were some problems where the vibrations generated from one of the HDAs (Head/Disk Assemblies) would be transmitted through the frame of the drive, and throw off the tracking of the other HDA.
Was that a stepper-motor drive? Or was it rotary voice coil w/ servo tracks?
I suspect older drives in working condition could be found. Corvus Systems shipped 10Mb Winchester hard drives for the Apple ][, back around 1979. They were pretty tough, I bet someone has a Corvus drive that still works sitting on a shelf somewhere. I was a Corvus tech back in the day, so if any such drive appears, I would be glad to help get it up and running (such as I am able).
Jeez, now I think back to ~1980 when 10Mb of hard disk seemed like an infinite storage space, compared to Apple ][ 140k floppies. And then I recall backing up a 10Mb Corvus disk to 140k floppies, ouch.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
