The hard drive form factor used to be 5.25 inches and then got reduced to 3.5 inches.
I assume we want the highest linear velocity that the platter can deliver.
If the linear velocity is the radius times the angular velocity, then for a constant
velocity drive, a 5.25 inch platter delivers 50 percent higher linear velocity than a
3.25 inch platter.
Stated differently, a 5400 rpm 5.25 inch platter delivers the same linear velocity to the read head
as a 3.5 inch 8100 rpm platter.
So why don't the hard drive manufactorers make their drive platters with larger diameters rather
than smaller diameters ?????
For the nitpickers..I don't know the actual platter diameter of 5.25 and 3.5 inch drives,
just trying to illustrate a point.
The smaller the platters, the better your random access/seek time. SCSI hard disks come with an even smaller platter.
It makes better sense to have 2 small platters rather than have one mammoth platter.
There have been attempts at going back to 5.25" form factor. Quantum once did it for cost reasons and introduced the single-platter "Bigfoot" series. As bendix demonstrated, they had quite good linear throughput although rotating slowly. Seek times were awful though.
Seagate in turn long sustained the capacity lead in SCSI drives by staying with the 5.25" full height form factor - that's two drive slots by today's standards! That was the "Elite" series, which went up to 47 GBytes.
As we speak, further shrinkage of SCSI drives is being done, to increase density in rack servers, and also because the fastest spinning 3.5" SCSI drives have 2.5" platters anyway (for seek time and stability reasons).
What would be even better is a 250 gig ram drive.
no platters = super fast IO
__________________
will not vote for or support any candidate for Congress or President who does not make a speedy end to the war in Iraq, and preventing any future war of aggression, a public position in his or her campaign.
Ram drives are expensive and I/O isnt blazing fast like you would think. then again they do have a .01ms average access time. which in this case makes it quite faster. until they get more reliable and stable and of course when the prices come down from teh ceiling. I will then think about purchasing one.
vicflo
__________________ My polished beauty... Stupid Space - join the rest of us Stupids.
3800+ x2@2.5ghz, asrock dual sata2, 2gb gskill, 2x160gb seagates (non-raid), agp 6600 non-gt @ 480/615 (make me offer for anything better pci-e or agp...)
RAM based storage drives do exist and did so for a long while. They come in Ultra320 SCSI, with an actual HDD in the background for non-volatile storage, and a RAM array doing the front end business. They go at full blast 300 MB/s throughput over the SCSI bus.
Prepare for an equally huge bill if you want one.
I remember a while ago I heard about a laptop with a 100GB flash memory based (or was it solid state? - can't remember the term) -anyways, it had 100GB of uber-fast storage coupled with a 16-hour battery life !!!
Originally posted by: oldman420
What would be even better is a 250 gig ram drive.
no platters = super fast IO
A Mercedes or BMW. I'm guessing they should cost about the same.
__________________
post count = post count + 0.999.....
(\__/)
(='.'=)This is Bunny. Copy and paste bunny into your
(")_(")signature to help him gain world domination.
If the drive manufactors keep kicking up the size of the on board cache,
then we may end up with a ram/platter drive as a standard.
Don't know how much the cache contributes to the overall speed of the drive.
Maybe they figure if Intel can sell more CPUs with larger cache size then
it should work for them too.
Actually, its none of the above reasons. The reaosn why HD manufacturers went to smaller diameter platters is because the flatness of a smaller diameter substrate can be controlled to a more uniform degree then a large diamter substrate. Its a stability issue basically. A hard drive platter spins around really fast. If the substrate is not as flat and uniform as possible, micro-waviness in the substrate will cause the substrate to wobble at high speed. This wobbling causes the distance from the magentic head and the platter to vary. This is a major problem because as you all know, magnetic flux decreases with distance. thus, the flux from the magentic media that is detected by the head will be constantly varying if the substrate wobbles as it is spun.
Ever wonder why really high speed drives have smaller storage density then a lower speed drive (i.e. a WD raptor 10,000 RPM fdrive comes in only 37GB and 74GB varieties)? Its for this reason. These drives use smaller platter sizes so the RPM's can be cranked up without this microwaviness issue. In other words, smaller platter=increased speed at the cost of storage density.
BTW: the 74GB raptor I believe uses two identical platters and a stronger spindle motor, not a platter that is twice as large.
__________________
Believe it or not, someone patented the wheel in the Australian patent office in the late 20th century. Not a fancy wheel...not a high tech wheel... a wooden wheel with spokes..... thats it.
RAMdrives do'nt need to be done via SCSI any more-we can use PCI-express. Think about it-PCI-E has massive transfer rates even on a "4X" slot, superior even to Ultra320. All you would need is a small motherboard-like PCI-express card, contatining a cheap chipset (Think VIA Kt266 equivalent) and a few slots for DDR. Put in a few old sticks of god-knows-what, and you have a RAMdrive, which could be loaded upon startup from an actual hard drives. If you put just your swap file and OS on the RAMdrive (1GB of space estimated) you would still get a massive speed boost, and assuming today's prices for a motherboard (60$) and a gig of RAM (130$) you could sell these for 200$ and still make a profit.
Of course, I'm just making speculations. There's probbabl about a dozen reasons why I can't do this.
-Cheesehead.
The point of making them SCSI is that they're transparently usable as if they were HDDs. In a sense, they are - upon powerup, they initialize from an actual, builtin HDD, and they keep backing up newly written data to that HDD.
Implementing RAM on PCIE is rather pointless - if you want that, you're much faster by putting more RAM onto the core RAM controller(s). Having the swap file on a disk that is actually system RAM is completely pointless - use the RAM as RAM and you're not going to need this swapfile at all.
I'll buy the comments about it being silly to put a swapfile on a RAM disk. However, a PCIE RamDisk isn't - actually, the connection is immaterial - ATA, SATA, PCI, PCIE, SCSI, USB, FireWire are all essentially equivalent for the purpose I have in mind.
I have a 2Ghz P4 running with 512 MB of memory - tell me, why does it take it close to a minute to boot? Answer: a hard disk is hard-pressed to do more than 100 seeks per second. Every exe that runs, every DLL loaded, every Registry read, every data file opened and read, all cause that poor mechanical contraption to beat itself to death. I would dearly love (but aren't willing to pay for) a system that treats a 1-2 GB ram drive as a boot device. Build an image on it, let the system boot from it, keep it around using the standby power from the power supply when the machine is off, Ghost the image down to the hard disk on occasion. Maybe define a mechanism to keep a 1-2 GB partition on the hard disk in sync with the RAM disk - if the RAM gets corrupted (say, because it was unplugged for a week), boot to the hard disk and reload the RAM image.
Would this be better if implemented on the motherboard? Sure, I'd love for a manufacturer to provide a BIOS option to partition some of the installed RAM as a RAM Disk. At current RAM prices of around $100/Gigabyte, I'd happily throw an extra 2 GB at my work machine to get it to boot in seconds. I don't see it happening in the near future, but ya never know...
They've had those devices for ages. PCI card, 8 ram slots, appears to your OS as a HD controller.
__________________
There are two kinds of fool.
One says, "This is old, and therefore
good." And one says, "This is new, and therefore better."
-John Brunner, The Shockwave Rider
We've got a couple computers that run machines that have an odd setup. The OS is on a ROM, and the config/setup files are saved in a small ram-type drive(had battry backup). It takes 3 seconds from off to useable machine. There is an HDD that the machine backs up the ram-drive to on shutdown incase the ramdrive fails. The machufacturer claims the hdd will last virtually forever. It only spins up when the backup is written, and if the ramdrive has failed on boot.
__________________
"Walk a mile in a man's shoes. You'll be a mile away and have his shoes."
Jack Handey Heat eBay
Originally posted by: FrankSchwab
I'll buy the comments about it being silly to put a swapfile on a RAM disk. However, a PCIE RamDisk isn't - actually, the connection is immaterial - ATA, SATA, PCI, PCIE, SCSI, USB, FireWire are all essentially equivalent for the purpose I have in mind.
I have a 2Ghz P4 running with 512 MB of memory - tell me, why does it take it close to a minute to boot? Answer: a hard disk is hard-pressed to do more than 100 seeks per second. Every exe that runs, every DLL loaded, every Registry read, every data file opened and read, all cause that poor mechanical contraption to beat itself to death. I would dearly love (but aren't willing to pay for) a system that treats a 1-2 GB ram drive as a boot device. Build an image on it, let the system boot from it, keep it around using the standby power from the power supply when the machine is off, Ghost the image down to the hard disk on occasion. Maybe define a mechanism to keep a 1-2 GB partition on the hard disk in sync with the RAM disk - if the RAM gets corrupted (say, because it was unplugged for a week), boot to the hard disk and reload the RAM image.
Would this be better if implemented on the motherboard? Sure, I'd love for a manufacturer to provide a BIOS option to partition some of the installed RAM as a RAM Disk. At current RAM prices of around $100/Gigabyte, I'd happily throw an extra 2 GB at my work machine to get it to boot in seconds. I don't see it happening in the near future, but ya never know...
/frank
My machine boots in 20 seconds, hardly unacceptable. Use hibernation if you want to speed things up.
I once had a bigfoot 5.25 size harddrive. It was designed to provide the same output speed as faster spinning drives and have more storage.
Slow as all get out. Long seek times, and only the outer rings provided faster I/O speeds, for the part of the harddrive (the majority of the drive) that rested within the boundries of a 3.5 inch disk, it was much slower.
So the top speed was only possible at the outer regions of the disk...
Also platters expand as they heat up, they are made of metal and all metal expands and warps slightly as it warms up. Smaller platters mean tighter tolerances, tighter tolerances means faster speeds are possible with the same technology.
Noise and bearing life, I suppose are also issues. Larger spinning mechanical devices exert exponentionally higher forces as they speed up faster and faster.
[i] Also platters expand as they heat up, they are made of metal and all metal expands and warps slightly as it warms up. Smaller platters mean tighter tolerances, tighter tolerances means faster speeds are possible with the same technology.
Actually, the vast majority of hard disks are made of primarily High Tg polycarbonate or glass substrate with a magnetic layer formed on the surface (via one or more metallic underlayers). Due to adhesion forces between the metal layers and the substrate the metal is substantially constrained from warping as the disc is spun.
Not to say that there aren;t known metal substrates out there. Lots of media use Aluminum Oxide which has been caoted with a thin adhesion layer of NiP asa substrate, but these substrates can have issues with the epitaxial growth of metal layers on their surface if they've been processed in certain ways. Titanium substrates are also known. However, the expansion of these substrates at high speed however is negligible. If it were not negligible they wouldn;t be able to be used as substrates because the noise induced by the strain imparted to the magnetic layer would be very high.
__________________
Believe it or not, someone patented the wheel in the Australian patent office in the late 20th century. Not a fancy wheel...not a high tech wheel... a wooden wheel with spokes..... thats it.
It's important to look at the business aspect of these transitions. No 8" hard disk manufacturer was successful as a 5.25" manufacturer and no successful 5.25" hard disk manufacturer was successful as a 3.5" manufacturer. Some large firms made the transition by buying successful smaller firms, but they didn't do it themselves. Innovator's Dilemma is an interesting look at why large high tech companies can't successfully innovate, and is based on a detailed analysis of the hard disk industry, though it also addresses the failure of mainframe manufacturers to make it in the minicomputer age and the failure of minicomputer makers in the PC era.
If I remember correctly, magnetic flux decreases with the distance squared
-you remember correctly. I was speaking in general terms, but thanks for the clarification.
__________________
Believe it or not, someone patented the wheel in the Australian patent office in the late 20th century. Not a fancy wheel...not a high tech wheel... a wooden wheel with spokes..... thats it.
There are drive manufacturers working on really small and really thin drives. Seagate has some nice small drives. They have good practicle uses in smaller products like PDA's, Cameras, MP3 payers.
The problems with this is that OEM manufacturers do not like to ever change anything like the size of a drive because of standardization and retooling. Large OEM's are against change. Also Smaller also sometimes means more expensive. OEM's want the lowest price possible to maximize on profit.
Why did we go from 5.25 to 3.50 ???
