EEVblog #395 - World's Most Expensive Hard Drive Teardown

[EEVblog]

Dave.

[wastrix]

IBM 3390 series: http://www-03.ibm.com/ibm/history/exhibits/storage/storage_3390.html

Introduced 1989, probably 5-10GB.

http://www.computerhistory.org/collections/accession/102649110

[David Aurora]

Probably my favourite teardown to date, awesome one mate!

Just emailed the link to my Dad, he was in the computer industry around the time this stuff was approaching end of life I think. I remember going to his office and seeing the rooms equipped with halon, data on tapes, hard drives measured in megabytes and considered huge... Crazy to think how far we've advanced so far in my lifetime.

[nitro2k01]

15:45 *Bangs head against monitor* That's an 8! (Isn't it?)

When you said it's the most expensive teardown to date, I immediately when to check what the Agilent 90000 cost. (Even though that wasn't really a teardown.) Nop, only $140,000.

That sticker, in the thermal chamber. Make it happen!

Here's a slightly less big hard drive that belongs to the student electronics club at uni.

[madworm]

Nice! Spin it up ;-)

Some more related photos. All of it was trashed many years ago. Too bulky, too heavy and of no real use anymore.

[notsob]

I used to repair disk drives in the 80's ( for a different manufacturer), both removable & fixed disk versions of approx 60+MByte, we repaired them to component level, the ones I worked on had individually replaceable heads, but no asics, all individual ICs. They had temp compensation built in and a full head alignment took some time.

Dave, you can get those stick on temp guages from the local home brew shop, they are put on plastic beer brewing containers to indicate the brew temperature whilst frementation is happening.

[Fliz]

Excellent, just excellent, I really enjoy every minute of this teardown 
I wish I had one of those plates  perfect material for wall LED clock
with tiny smd led's, no spinning parts... blend of old and new technology

Thanks Dave

[ElectronicTonic]

Dave! How can you not take apart the linear actuator? I've been tearing down hard drives since I was a kid, and any old drive that uses a moving coil (some had stepper motors) for the linear actuator is gonna have some bad ass super strong neodymium magnets. I've taken apart some smaller hard drives in the past (mid 80's vintage with ten 5.25 inch platters at 20 MB total) and they had much larger and much stronger magnets than the little rinkidink magnets in modern drives. Careful though. If you pinch your fingers with those huge magnets, the tear down gods will be quite pleased!

Also, it looks like that three terminal device inside the locking motor is a hall effect sensor. What is that little black colored block on the end of the semicircular toothed locking mechanism? A magnet? What causes the whole thing to move to become unlocked?

Anyway, awesome video, Dave! One of my favorites. I love tearing town hard drives, but I've never had such a ginormous one as this. Thanks for the porn!

[ElectronicTonic]

Also, what's with all those holes in the disk spacers? Is there some way for air (or halon) to flow into the center of the spindle and be propelled outward across the disks much like in a Tesla turbine?

[bitwelder]

Thanks Dave for the great teardown.
It was just pure p0rnography to watch the innards of such old engineering beauty!   

[SeanB]

I tore down a old hard drive that was a 2 platter 20M SCSI drive, that was used to run a mailing list ( moved now to a small file on a server) and which had a s100 bus Northstar Horizon and dumb terminals to operate it. Was a later version of CPM on 5 1/4" floppies which stored the program on disk and used the drive to store the database. Backup was on 2 boxes of 360k disks............

Still have the very heavy CVT that was used on the hard drive, the whole system used around 2kW.

[bxs]

Hi Dave,

That drive is too small, about 15 years ago I remember seeing one huge harddrive in a lab, I will try to contact a friend to see if it's still there (probably not...) and get some pictures

[Winston]

Dear Sir,

It has come to our attention that you are in possession of one of our Model 3390 397x hard disk
units that is still under a maintenance contract and which should have been returned to IBM upon
removal from user service.  Please return this unit to us in original, fully functional condition or
send us a check for US$250,000.

Best regards,
Winston Smith
President (imaginary)
IBM Service Department

[alanb]

I've seen larger units.

In the 80's I used a system with exchangable disk packs. The drive was about the size of a typical washing machine and each disk pack that was loaded on top was about 20 inches in diameter and 2 inches deep. The capacity of each disk was 10Mb.
I had a carrying case to hold one of the disks and had to take a disk off site every evening.

edit - It was a CDC Hawk and the capacity of the removable disk was 5MB there was a fixed disk of the same size internally.

[calin]

Indeed .. great. Used to have a similar at the university - came in really handy when we learned peripherals and storage. Not as sexy as this one but good learning tool.

BTW, u know IBM still calls the mainframe storage DASD up to these days.  I have a guy around me that is old enough to have seen these in use - he was overjoyed when I showed him the video .

[SeanB]

BTW, the removal of the heads off the media is called parking........

Those disks are not sputtered units, they are spin coated. Pour a small slurry of ferrite and binder onto the inside while it spins at high speed and it will be evenly distributed over the disk, like in modern CD production when they are coating the optical pit layer with the top coating. No fancy magnetics at that low density, just finely ground iron oxide and chromium oxides that are sieved then blended in a ball mill to a fine submicron powder then mixed with a solvent based carrier to stick it to the aluminium platters. During final test there would have been a special jig that moved the head assemblies a precise amount as it wrote servo tracks on the one platter.

As to the magnetics, the writing direction is in x, the material in the gap forces the magnetic field to spread out into the magnetic media in the Y direction ( and a little internally as well) while the Z is inter track coupling. the data is written by the magnetic field at the trailing edge of the head gap ( it retains the magnetic orientation it has as it left the field in the gap) while the reading is of the material in the whole gap. That is why on later units you will have had 2 coils in a line, the first with a very small gap does the reading off the disk, while the back one a certain precisely controlled  time behind ( actually distance travelled as radial velocity of the disk) can write the new data onto the sector block after the read head has read the preamble and been servoed into alignment on the track. this compensates for offset due to heating and expansion of the platters and heads, yours has a servo platter that does the same function.

Older ones had a big enough head ( same as a floppy drive size wise) that they could ignore it mostly, as the tracks were set by a stepper motor.

As to the air lines, those are not Halon, but for pressurised clean dry air from a built in compressor and filter drier pack in the drive bay, used to circulate clean air into the drive case and remove the hot interior air. The cool air is aimed via the microfilters inside to pass over the heads, then pass through another filter before circulating through the disk pack and then exiting via the spindle motor to cool it. Thus the holes in the spacers so air can go down. All designed so there is no turbulent flow on the disk pack where the heads are flying just above the disk, on a thin cushion of air compressed by the wedge shape on the bottom. This keeps the heads almost in contact with the platter so the magnetic field does not spread out and weaken, but reduces the friction to almost zero, and reduces wear to almost nothing. Compared size wise to flying a 747 at full speed 6 inches over the ground.

[kyndal]

as for taking the linear actuator apart.
looks to me like there is a "hinge like" mechanism on each side with a pin going length wise
down / through the side ?

what do you mean you dont have pneumatic impact tools in your shop ??

[SeanB]

The magnet holds it together, you need a set of jigs that pry the poles apart to get it open.

[free_electron]

i just posted some technical info under the video . not going to repost here. now, as far as cost.. THis is not the worlds most expensive drive. That would have been The RAMAC ( the original harddisk... )

http://en.wikipedia.org/wiki/IBM_305_RAMAC

Below a picture fo the entire drive. left the platters ( 52 of them , right the cabinets with electronics to tame this beast. )
-50 user disks (dummy disks at end to reduce turbulent buffeting)
-100 sides, 100 user tracks per side (2 test only tracks on inside and outside)
-5 sectors per track, 100 characters per sector
- Grand total of 50 disks x 2 sides/disk x 100 user track/side x 5 sectors/track x 100 char/sector = 5,000,000 characters

The bootloader of a modern OS wouldn't even fit on this thing... Take a picture with a 10 Megapixel camera of it in JPEG and you can't store it on this thing ...

The computer history museum has ony. The drive was invented here in San Jose and the were built in the Cottle road facility , where , today , they are STILL building and designing harddisks. Now part of Hitachi storageworks. ( IBM sold their drive business to hitachi ) HGST.
The old building was torn down and is now a Lowes home improvement sotre , but as part of the deal they had to decorate their building using the IBM tile pattern , put a memorial pagoda with images and a baseball field called RAMAC park. :

http://maps.google.com/maps?q=+san+jose+cottle+road&hl=en&ll=37.251451,-121.800066&spn=0.006764,0.011147&sll=37.776229,-121.98187&sspn=1.719323,2.853699&t=h&hnear=Cottle+Rd,+San+Jose,+California&z=17

[Lightages]

In the early 90s, IBM was doing the research into the materials used in the hard drives we have today. Their research center in San Jose California was using a very specifically designed materials research system designed to make giant magnetic resistive effect materials and high temperature superconductors. It is this research that resulted in the very small read heads that make the smaller drives now possible.

The picture attached shows their research system under development in 1992 at a company in Canada. If anyone is interested in what the system did or its capabilities, just ask!

[And!]

Try to sprinkle a little bit of metal dust on a plate. Maybe tracks can be seen?
At 20MB 8" disk it was possible.

And
Spin it up ! Spin it up! Spin it up!

It will be more difficult than running super high speed laser mirror motor (36m:00s):


But it will be scary

[dexters_lab]

Dave you gotta get that actuator apart just for the magnets, those in the newer disks are puny in comparison as the newer voice coils are so much more efficient

Great video though, love seeing top of the line old tech pulled apart

Would also love to see the spindle motor powered up

[pickle9000]

Too bad you didn't have some MICR toner, sprinkle a bit of the on the platter and the tracks would have shown up.

[mikeselectricstuff]

as for taking the linear actuator apart.
looks to me like there is a "hinge like" mechanism on each side with a pin going length wise
down / through the side ?

what do you mean you dont have pneumatic impact tools in your shop ??

An angle grinder always wins.

[Pentium100]

It's a shame they don't make 5.25" (or bigger) drives anymore. With modern data densities the capacity would be awesome.

[Robreeves]

Very similar to the first machine I was paid to program in 1973.  IBM 1440 (disk version of the 1401).  The disks were removable, 20000 sectors of 100 bytes = 2mb. We had 5 of them. Oh, and the big day is when we upgraded memory from 8k to 12k

[EEVblog]

Dave! How can you not take apart the linear actuator?

I tried to lever it open with a big screwdriver, and I couldn't.

Dave.

[EEVblog]

As to the air lines, those are not Halon, but for pressurised clean dry air from a built in compressor and filter drier pack in the drive bay, used to circulate clean air into the drive case and remove the hot interior air.

I am assured by the original owner that it was connected to cooled pressurised halon. Yes, it's used for cooling, I forgot to mention that.

Dave.

[EEVblog]

as far as cost.. THis is not the worlds most expensive drive.

Never let the facts get in the way of a good keyword searchable video title 

Dave.

[ModernRonin]

Re: Halon

For those of you who haven't played the classic: http://markdamonhughes.com/Halon/

[nighthawke]

My impression is that they used a manifold system subsisting of air for regular operation, then switched to nitrogen or Halon in case of a disaster while the drives went into SCRAM mode.  And yes, there's an emergency shutdown procedure that converts the drive motors into magnetic brakes, stopping the drives within a few seconds.  They used a magnetic brake on the large Burroughs 10MB fixed head drives to stop it. Hell, they had to since the hub itself was 30lb cast bronze.

I did a dig and found the specs on the 3390 system.

Interface: Gen 1 SCSI or proprietary SSA. The arrays were interconnected using fiber optic ESCON cables.

1.5 ms for single platter access, average seek time 12ms, 23ms maximum.

Capacity was 792 MB or 1,585 MB

They had a encryption system built into the storage array, using 38 byte keys. There's tables for calculating capacity with or without these keys.
http://bitsavers.informatik.uni-stuttgart.de/pdf/ibm/dasd/reference_summary/GX26-4577-0_3390_Reference_Summary_Jun89.pdf

The following review was done by Performance Associates. They classified the storage array as a "gap filler" bulk storage system, not a DASD as IBM had touted it to be, and they found issues with the system as noted in the documentation. This was written up back in 1995 when the trend for storage was being shifted to smaller RAID 5 Fibre Channel,  NAS or SAN arrays.
http://www.perfassoc.com/pdf/390sampleprofile.pdf

[SeanB]

If it had gas cooling it would not have used Halon, more likely R11, as it is available with incredible purity. Halon is not that filtered or pure.


So as you are in AUS you will have to pay around AU$100 000 in carbon tax for that sniff of gas that was left in the housing........... Next time you get the aircon in the car regassed remember most of the price is that.

[digsys]

Hiya guys, MOAHDD donor here, and it's past history.
While I admit that years of LSD use in the 60s and 70s has pretty much fried my memory, most the details are accurate, maybe just mixed up.
My partners and I had several big-ass HDDs, some older and some much larger. So the 10/20MB capacity must have been of the others, my mistake.
My factory was next door to the site, so techs would drop in, for convenience. I visited the site on a couple occasions, and was always
apprehensive of the safety measures, so I stayed away. As for the cooling system. I recall that it was mentioned that at THIS site, running plain
cool air was not adequate, plus there was a fear of internal combustion, so they came up with a chilled halon (mix), with white colouring? and a
distinct smell (for safety). I'd cracked open 2-3 in all, but the first time I did, I was standing directly over the HDD when I cracked it. A burst of white?
gas whooshed out, and that's when I recalled the safety instructions. I held my breath and ran for it.
Maybe it was from another installation, but I don't think so, but then I could have sworn I remembered it as one of the 20MBs.
Facts are all there, just stick them together yourselves :-) It's GREAT to see members adding to the history.

NOTE: It may not be the world's most expensive drive, but note "TEARDOWN". Until then, it stands :-)
PS: Most of the others were converted to wall clocks and such, some still run today.

[Cozzmo]

Also, what's with all those holes in the disk spacers? Is there some way for air (or halon) to flow into the center of the spindle and be propelled outward across the disks much like in a Tesla turbine?

Most likely weight reduction for the rotational mass as they would not need to be totally solid for any strength reasons.

[G7PSK]

Is that actuator held shut with a magnet. There must be a powerful magnet in there somewhere to shift that mass at the speeds it has to.

[SeanB]

The centre is a massive cobalt magnet.

[free_electron]

Samarium cobalt to be precise. These things are so strong that , if you put two of them 1 inch apart and let them snap together they shatter ... NEVER EVER put you hands between two of those. If they jump at each other it will seriously hurt ! It may even break bones i. Your fingers if the magnets are large enough.

[blc]

I wish I had one of those plates  perfect material for wall LED clock
with tiny smd led's, no spinning parts... blend of old and new technology

Funny thing, I said more or less exactly the same thing to my other half last night; even more surprisingly, she actually agreed with me!

[Zad]

i just posted some technical info under the video . not going to repost here. now, as far as cost.. THis is not the worlds most expensive drive. That would have been The RAMAC ( the original harddisk... )

http://en.wikipedia.org/wiki/IBM_305_RAMAC

1955 eh. To be fair, they do say the first moving head storage device. Manchester University's Baby computer had drum storage in 1949.

http://tommythomas.org.uk/Manchester/manchester_drums.html

I remember some time around 1981 or so, going to the county computer centre (West Yorkshire, England) and being given a guided tour by a chap my parents knew. After showing me the (even then archaic) punched card reader and phenomenally fast and loud line printer, he took the back off a removable storage drive to show me the internals. You could clearly see the linear motor actuators.

At that time, I seem to remember the disk packs were about 400MB or so, which seemed phenomenally huge when home computers were usually 16KB-64KB with <100KB floppy drives. I still have a 10MB "full size" ST412 (or similar) disk drive somewhere.

[george graves]

Lol - it was funny to see Dave so puzzled by that temp sensor sticker.

That temperature reading strip is nothing more than an a aquarium type temp sensor:

[EEVblog]

Lol - it was funny to see Dave so puzzled by that temp sensor sticker.
That temperature reading strip is nothing more than an a aquarium type temp sensor:

Yes, everyone says they have seen them on aquariums and home-brew beer. I'm into neither, so have never seen them.

Dave.

[george graves]

I had totally forgotten about them.  Not a bad idea for something like that.

[labarowski]

I was going to chime in about the temperature sticker being for an aquarium but george graves beet me to it. I guess that a technician was concerned about temperature and placed it there. It doesn't seem like it would do much good though. Their peak temperature is not very high.

[george graves]

I'd assume you could get them custom made for what ever range you need.

IIRC - they are a two layers of plastic with heat-sealed little individual chambers filled with liquid.  The some kind of chemistry voodoo happens to make them change from black to a color.

[SeanB]

They actually use cholesterols ( the same as what is in your arteries) as the fill. they have a polarising front like a LCD, and the cholesterol s are different types, so that as they hit a certain narrow temperature range the optical twisting that they do of polarised light ( from the front filter) is 45 degrees. the back reflector then reflects, it goes through another 45 degree shift then it can go out and you can see the back reflector and the markings. At higher or lower temps there is either too much or too little phase shift, so it remains black. The colours come from fringing  as the polarisation approaches the 90 degree point. Totally passive, absolutely no moving parts and reversible.

the other variant more commonly used in engineering is a max indicating type, which has small  black coloured wax dots inside a carrier, so that as the wax melts it is wicked away in a surrounding reservoir and exposes a coloured backing that shows the top temperature reached. Once use only, but quite good as a log of overheating or max temp in say a heatsink during operation, where you either cannot or do not want or can use a data logger.

[Zad]

I remember seeing those thermometer strips in the RS (Radio Spares back then) catalogue in the 1980s. I tried to persuade my dad to buy some, but for some reason he refused. Probably the phenomenally high price back then.

[bitwelder]

Yes, everyone says they have seen them on aquariums and home-brew beer. I'm into neither, so have never seen them.

Dave.

I worked in a computer datacenter and I've seen them attached to several racks there, especially in the aisles where the server density was relatively high.

[SeanB]

We had a simpler one, one heatsink ran so hot that the most common fault was the BUX20's in it unsoldering themselves....... Limited space, and no way to redesign...... Would have loved to have modern MOSFETS in there, they would have handled the job with no problems.

[Hypernova]

I thought about getting some of those stickers for heat sink temp monitoring and in the course of my search found this place:
http://www.telatemp.com/p/492/
16 steps of 5C from 25 to 100 ought to do it for most people.

That's probably the true source for the majority of the sellers out there. At $40 for a pack of 20 it isn't that much, and they have variants reaching 120C. There are shops on the web that resell them at single quantities.

[jabramo]

Holy smokes this teardown is getting tons of attention! Front page Gizmodo, Extreme Tech and Hackaday

[cwalex]

It was a really nice teardown as usual by Dave but I like the more standard off the shelf kinds of tear downs as I learn more about product design (do's and don'ts) etc  but honestly I would rather see dave get heaps of exposure so he can make more out of his new full time gig! than just see more stuff that I personally like to see.

Good videos as always Dave! Thanks for all that you do for us

[EEVblog]

Holy smokes this teardown is getting tons of attention! Front page Gizmodo, Extreme Tech and Hackaday

Yeah, probably my most popular video to date in terms views in such a short time.

Dave.

[JoannaK]

Was expecting something a bit bigger... But I'd assume there's not many of those PDP-8 era (or older) hard drive units available. We (bunch of Uni students) did purchase one (with PDP-8) about 30+ odd years ago, not that we'd needed but it was kinds funky to have fridge/Freezer size HD system with removeable platters.

Not sure where those parts are these days.. Not even sure if I could find any photos of the system.

[alanb]

Posted by: JoannaK
« on: December 12, 2012, 10:19:22 PM » Insert Quote
Was expecting something a bit bigger... But I'd assume there's not many of those PDP-8 era (or older) hard drive units available. We (bunch of Uni students) did purchase one (with PDP-8) about 30+ odd years ago, not that we'd needed but it was kinds funky to have fridge/Freezer size HD system with removeable platters.

Try Googling CDC Hawk.

[JoannaK]

Posted by: JoannaK
« on: December 12, 2012, 10:19:22 PM » Insert Quote
Was expecting something a bit bigger... But I'd assume there's not many of those PDP-8 era (or older) hard drive units available. We (bunch of Uni students) did purchase one (with PDP-8) about 30+ odd years ago, not that we'd needed but it was kinds funky to have fridge/Freezer size HD system with removeable platters.

Try Googling CDC Hawk.

Something like that for sure.. Kinda scary when one starts to think about those platters turning ..

[alanb]

Kinda scary when one starts to think about those platters turning ..

The most scary part was the voice coil / actuator mechanism. I once met a service engineer who had lost three fingers on his right hand when a the voice coil was actuated with his hand in the way.

[pista24]

Hi,

which videocam do you use to record this video? Especially macro mode for magnetic heads.

Thanks.

[SL4P]

Around a decade before this drive, I was working on DEC hardware, and specifically the RK05 (2.5MB) and RL01/2 (5/10MB) drives attached to LSI-11 systems.
It's interesting to see the similarities and throwback concepts - but despite being built like battleships they were extremely fragile in transit and in use.
Don't underestimate the voice-coil positioners. They would seek sequential tracks in under 2mS, and to any track under 6mS with a LOT of force. The'd have your finger f it was in the way!.
A colleague had his flashy bling ring melted onto hois finger by shorting across the power supply to the positioner... !   An optical strip encoder provided the track position feedback to the positioning electronics.
The RL0x drives were a bit more friendly - higher data density / slightly slower seek speed, and used another strategy common back then - with one surface pre-formatted with index data (effectively hard sectored)

Great to have the memories brought up !

[aqarwaen]

does anybody knows how you recover data from theys??my workplace got 2 theys kinds hdd.only issue is ,computer what used them are dead.both hdd are located at basement..any  clue how can i get data from them and connect with pc??they mostly collect dust there

[Brumby]

My first suggestion is to find somewhere that still has one of these running ... probably a museum by now ... and see if they would be prepared to try and load it.

But I think you will have another problem.

I take it those drives haven't been accessed for some years - so I would highly doubt they would be readable.  The magnetic domains have likely deteriorated over time.

[digsys]

Now we're going back. I can't recall if it was a SCSI or MFM interface, but you may find an old PC that still has a ISA (16?) bus. If so, it should be pretty easy,
especially with a disk recovery program. ie RSTUDIO r-tools tech
In cases where the hardware had failed, we used to use a chemical to "stain" the magnetic material, then high res photograph it. Someone had a program
that just re-encoded the headers etc, Most / all of the professional data recovery companies use that technique. On those large platters and such low data
density, it is quite easy.

[Fungus]

I'm sure there'll be a USB adapter cable somewhere on eBay.