Slashdot Mirror
Western Digital Announces 200 Gig Drives
twilightzero writes "Video capture fanatics and pr0n moguls, rejoice! Today marks the official release of the Western Digital 200 GB hard drive! Never again run out of space for your X-10 video stream of the neighbor's house! See the graphic, specs, and press release. This also marks the release of WD drives using fluid dynamic bearings rather than the old BB type." The glorious march of technology continues forward, and digital video fans rejoice. Update: 07/26 03:34 GMT by M : Headline corrected. Taco's at a conference, cut him a little slack.
Liquid bearings add a little bit to the price. At New Egg, for example, a 40GB ATA133 Maxtor is $3 more with liquid bearings and an 80GB ATA133 Maxtor is $8 more with liquid bearings.
Allegedly they operate with less noise than standard bearings. I haven't verified this personally, but the online reviews I've read seem to indicate that this is true.
"...always new atoms but always doing the same dance, remembering what the dance was yesterday." -Richard Feynman
I've installed a couple different drives with the fluid bearings, and they do run quieter than the older style bearings. Very nice!
Instead of using ball-bearings (for the disc plates) it uses a liquid-suspension system. It's more reliable, you'll have a higher durability, possibly it's less noisy and can run faster. Since it's a new technology (and cool) it'll cost higher too...
You'll need an ATA133 controller, or a RAID controller that can address drives beyond the current limitation of most ATA100 controllers.
Promise makes one, I'm sure. Maxtor 160gig drives are sometimes bundled with a controller.
I'm in the market for a new machine, and I've been spec'ing out different parts for my budget...These drives are nice and big, but what happens when you lose a 120 gig drive...I've pretty much decided that I'm going to have to get an IDE RAID card and highly recommend them...the RAID cards at work have saved me hours and hours of restoring from backup...Check out the 3ware Escalade, the Promise SuperTrak, or the Adaptec 2400A. RAID 5 is the way to go (with or without removable drives). I've been watching the prices for 120 Gig drives drop and now it's just about the price where I can afford to spend 150 clams to buy an extra drive that would be used to protect myself from a drive failure.
- grunby
Now I hear people say this all the time. But in my nearly (oh wait, this is 2002...we can skip the 'nearly' part now :-P) 20 years of computing experience, I have only had 2 drives that ever died on me (like totally dead, not just developed a few bad sectors or whatever) and they were both Maxtor drives. And I have owned drives that were manufactured by Maxtor, Western Digital, IBM, Seagate, Kalok, Fuji, Quantum, Toshiba and another company that I can't remember the name of right now... :)
My journal has hot
Western Digital already announced 200GB drives a few weeks ago, so this is probably a Maxtor announcement.
What I want to know is how they made a 200GB hard drive with 60GB platters. Doesn't seem to add up.
Computer Science is no more about computers than astronomy is about telescopes. --E. W. Dijkstra
It's WD, I submitted the article with a different title and it got edited...to the wrong company LOL! I laughed so hard I almost wet myself when I read it...
"Christ what a design! I could eat a handful of iron filings and PUKE a better emergency pump than that!"
The press release is June 25, but the drives just SHIPPED today, THAT'S the news.
"Christ what a design! I could eat a handful of iron filings and PUKE a better emergency pump than that!"
I bought a Maxtor 160GB drive that came bundled with an ATA/133 card. I ended up putting the drive into an ADS Firewire enclosure, which seems to be just as capable as addressing the whole drive as the bundled ATA/133 card.
Perl - $Just @when->$you ${thought} s/yn/tax/ &couldn\'t %get $worse;
3 platters = 6 sides.
5 sides for dats (40 gigs each) and the bottom of the bottom platter for positioning.
I just asked my dad who's an engineer at WD about this and he said in fact it uses 3 (which is the max they can use) "60GB" platters. I put 60 in quotes because they're not exactly 60GB, really they're ~67GB platters they just round down to the nearest 20GB increment.
dd of=m00z4k.mp3 bs=1 seek=1874373 count=0
so that no disk space (except the dentry and the inode) is wasted... Oh, you have just bought a shiny 200gb disk, haven't you?
Seagates recent (May 28, 2002) US patent entitled "Antiwear lubrication coating for bearing surfaces for hydrodynamic fluid bearings" explains very well the advantages, the structure and the operation of fluid disc drive bearings. Go here: http://patft.uspto.gov/netahtml/srchnum.htm and enter this patent # : 6394654
Everybodu knows that a "real" gigabyte is 1024*1024*1024 bytes, or 1073741824 bytes, while a "hard drive" gigabyte is 1000000000 bytes. If you put 3 60 "real" gigabyte platters together, you get 193273528320. That is close enough to 200. Especially if you insret some marketing-speak there to fill the 7 remaining gigabytes.
Of course, the 60 gigabyte platters are "hd" gigs, so you are really getting 167 "real" gigs, but who's counting...
This would be funnier if it weren't true :-) Of course they are 66 GB platters and 198 GB drives.
I am a science fantasy fan
The limit is due to having only 28 bits in the IDE registers to selecting the address. There are four 8-bit registers, and the "head" register uses 1 bit for master/slave selection, one bit to select CHS/LBA addressing, and two bits are "reserved" (originally used to select sector sizes, but in modern times sectors are always 512 bytes).
ATA-6 kludges this 28 bit LBA limit to 48 bits by specifying that the host is to write 20 bits twice!
But for the forseeable future, 32 bit computers will only really use 32 of those 48 bits, which turns out to be only 2 terabytes. If the operating system uses a signed integer (common practice, including the linux kernel until only recently), you only end up with 31 bits of sector addressing, or just one terabyte.
Of course, there are probably even more limits lurking. Doesn't linux ext2/ext3 use 32 bit numbers? FAT32 uses 28 bits for cluster numbers, but clusters can be as much as 32k in the standard (apparantly larger in some systems, though Microsoft doesn't document that in the FAT32 specification).
PJRC: Electronic Projects, 8051 Microcontroller Tools
In my last 5 years I regularly worked on the servo controller block of a drive controller ASIC so I'd like to point out some much under appreciated issues and correct some recurring miss understandings: a) the servo system, b) the real benefit of fluid bearings, c) error correction, d) why an unused surface, e) spin speed
Some short answers are:
b) Worn ball bearings seriously disturbs the servo system from keeping the read head or the write head on true center of data track. This puts a ceiling on increasing track densities.
c) Very strong error correcting codes are applied to every data block (about 512 bytes) and not on any unused surface.
d) As for unused surfaces, there are multiple issues in this decision, but a new feature from some manufacturers is to reserve one outside surface for a template servo pattern & BIOS code so that the drive can self write its own servo patterning and more cheaply load its BIOS code. This reduces / obviates the many hours it takes of very expensive capital equipment to write servo patterns to drives.
e) Spin speeds above 10K introduce horrendous resonances at the outside of platters that make servoing tracks much harder. One remedy has been to reduce platter diameter & capacity (by about 10% as I recall).
And the long version of a) & b) or 101 of disk drive servoing:
For atleast a decade, hard drives have used embedded servo patterns on every surface that are intermingled with the data areas. Using a dedicated surface for servo worked long ago only because track & data bit densities were much lower. Todays drives typically have 120 or more curved radial servo wedges that costs 5-7% of the surface area. User data tracks nestle between these servo wedges.
1) In these short servo wedge areas, servo tracks contain a few tiny fields of digital servo data followed by several analog modulations so that the servo processor can sense its fractional position within any servo track. Servo tracks actually abutt each other and the only bit change between adjacent servo tracks is in the Gray coded track no. Since IBM patented this many years ago (1980's ?), manufacturers have since added proprietry extra small digital fields to correct for read errors in the digital fields & analog modulations to continuously improve servo tracking and hense improve data track densities.
2) The user data tracks are not neccessarily pitched to be inlign with servo tracks and may be reduced to 2/3 the density of servo tracks. This provides guard space and reduces inter track symbol interference. As the disk spins from servo wedge area on into user data area, it becomes an increasing act of faith that the read head or the write head is indeed still following a track center until we reach the next servo wedge. Such miss tracking is called runout.
A major source for Non-Repeatable Runout comes from worn bearings which introduces random wobbling, and this degrades the servoing and limits tracks densities.
Fluid bearings improve upon ball bearings because they don't introduce this NRR so spinning is quiter but more importantly track densities can keep climbing.
There are quite a few other NRR & RR terms impacting on servo tracking.
I've used several 3ware escalade cards and they are faster than my old software raid5 in everyway. Plus, they setup in their own BIOS, can use hot spares, and were seen instantly by Linux Mandrake on installation (as a scsi disk). The 4 port 6410 escalade can be found for 89MB/sec sustained in Raid5, with bursts WAY higher using WDC120's. It takes some serious U320scsi to do those numbers in the real world.
-- Whitewlf White Wolf Networks
A big clarification I would like to make here is that most cheap "hardware raid" controllers are NOT hardware raid controllers. They are clever hacks to implement RAID using a combination of a bios handler for software RAID and an OS driver implementing software RAID at the driver level. That is why linux md running raid5 is often faster - the implementation is better than the device driver provides.
Now, when comparing performance to a real live IDE RAID controller (Adaptec AAA or 3Ware, etc.), it is not as fast. These controllers have an on chip implementation of RAID 5 (ie hardware XOR etc. usually implemented on an intel i960 or somesuch) and perhaps some cache memory, and they interface with the OS using the standard SCSI drive api.
Now software raid 0 or raid 1 is often just as fast as hardware raid 0 or 1 because the implementation is so simple and the drive r/w speeds are the limiting factor.
BTW: does anyone know exactly what to call things like the promise and Highpoint "Raid" controllers that rely on BIOS hooks and software drivers to do the RAID dirty work? -- "Hardware" doesnt work and "Software" doesnt work -- is there a word for it?!?
~GoRK
For more information on how bearings work, and wear in general, do a search on the field of "tribology". It's all there.
Some people win the lottery.
Some people buy 6 substandard drives from the same manufacturer.
Some people use a 5400rpm CPU fan and 5400 rpm drives and expect they won't set up narrows bridge style resonate frequencies in their cases.
Some people do not properly cool their cases.
Some people bang their boxes around at once a month lan parties and wonder why their drives fail.
Some people overclock their machines but don't use western digital drives because they tend to behave badly.
If voting were effective, it would be illegal by now.
That 60GB is two sides at 30GB apiece, so I'd guess they've used 4 platters, but are only using seven sides to keep the phyiscal drive height down. That still leaves an error of 10GB mind you, but hey, that's only 5% for the sake of a round number.
UNIX? They're not even circumcised! Savages!
Does anyone actually have a Java program designed to control air traffic, or for the operation of a nuclear facility?
As an auto mechanic, I can clarify a few things, although you got almost all of it right:
Ball/roller bearings are (generally) much more expensive to make than most implementations of fluid-dynamic bearings.
Hydrodynamic bearings' load capacity goes up with rpm, meaning a small, fast rotating bearing works better than the same bearing at a slower speed.
Roller-element bearings have limited high-rpm applications. The higher the rpm, the more scuffing the rolling elements are subject to.
Plain hydrodynamic bearings usually require a pump to provide adequate fluid-dynamic pressure.
A "plain" bearing rides on a wedge of oil which is related to the viscosity of the oil, the velocity of rotation and the clearance between the elements. A roller bearing's capacity and max rpm are determined by the surface finish of the rolling elements, the tolerance of the cage (if any) that retains them, and the radial runout of the assembly.
In automobile use, plain (hydrodynamic) bearings are much, much cheaper to make than rolling element bearings because of their ease of manufacture.
For the love of God, when will the PC industry stop with these damned limits?
FYI, for anyone interested in reading a nice list of all such limits with a technical description for each one, I suggest this link.
Yes - most BIOSes can and do spin the hard disk down after a period of inactivity. However - most Linux distribution flush to disk every 5-10 seconds, which of course means there is never a period to spin down.
I did read about a hack to disable to bdflush on laptops somewhere, but it was a while ago.
> 40% on-disk errors before low-level (factory) format?
Coming from a firmware engineer from a disk drive company, that sentence makes no sense whatsoever.
The "stretched capacity formats" that drive companies are using to reach their 200GB or larger drives are almost purely a function of the heads used in the drive, and have almost nothing to do with the specific media. From the plots I have seen, if media had 1% surface defects I would be surprised...
More data, damnit!