26

I know that hard disks have faster access times and transfer rates and when I searched why it appeared that:

  1. Hard disk drives spin much faster

  2. If filled with helium this can decrease the energy needed for spinning them and thus faster speeds

  3. You can store more data on a hard disk than on a similar sized optical disc so higher data density so the head won't have to travel as much distance as it has to with optical discs

  4. Optical discs (not all) store their data in spiral tracks

But why do HDDs spin faster, or why do their arms move faster? Why don't they manufacture the arm inside the optical disc drives to move at the speed of HDDs? Even magnetic tape drives are fast. (At least transfer rate, not access time.)

Also, can't the head go to a specific part of the spiral track and read from there?

3
  • 7
    ODDs and HDDs use two completely different head types (and technologies in general) - ODDs use a gear track, whereas HDDs use a voicecoil, and there's no way a gear track could ever compete with a voicecoil due to physics.
    – JW0914
    Sep 9, 2021 at 0:01
  • 8
    An apples to orange comparison. Both fruit, but why aren't apples as spherical as oranges? Two storage devices with very different capabilities, and they use different technologies. So why should any attributes be similar? If they did have similar capabilities, then the more expensive and/or less convenient medium would fall out of favor. Do you remember floppies?
    – sawdust
    Sep 9, 2021 at 5:07
  • 4
    @JW0914 For completeness, optical drives have a gear track for coarse control and voice coils for fine control (about +/- 2 mm).
    – Nayuki
    Sep 9, 2021 at 22:26

9 Answers 9

55

There are two separate questions: latency and throughput.

Seek time

Optical drives have random seek times around 100 ms, whereas hard drives are around 10 ms - why is this?

An optical read/write head consists of a laser, detector, mirrors, lens, and multiple voice coils to position the head/lens for tracking/focusing. This entire assembly has a number of parts and is relatively heavy. The entire head assembly moves on a worm gear.

enter image description here

A hard disk head consists of a tiny GMR sensor for reading and coil for writing. The rest of it is the plastic slider and metal arm; the whole assembly is relatively light. The head is moved by a powerful voice coil. This is why seeking is fast.

enter image description here

Transfer rate

CD drives top out at about 10 MB/s. DVD drives top out at about 30 MB/s. Blu-ray drives top out at about 70 MB/s. Hard drives routinely hit 100~200 MB/s.

The two main factors that determine transfer rate are linear speed (metres per second) and linear density (bits per meter).

Desktop hard drives in the 3.5-inch form factor have a platter diameter of about 90 mm. Optical discs have a diameter of about 120 mm. I would rate these as "close enough" for comparison purposes.

Hard drives usually spin at 5400 RPM or 7200 RPM, with old enterprise models going up to 15000 RPM. Optical discs are spun at various speeds, depending on how quickly the host wants to read/write data (e.g. bulk read vs. streaming audio/video), how long the drive has been actively used, how much noise is desired, etc. But optical discs can spin up to about 10000 RPM in real drives without problems. So this is also in the same ballpark as HDD RPM.

enter image description here

Areal density is a huge factor to consider. Pretend for a moment that a CD is 1 GB and an HDD is 1000 GB and they have the same physical dimensions. Clearly, the HDD has 1000× the data density per area. As for linear density, the HDD is √1000 = 32 times denser than the CD. So if you position a head over the discs and make one full revolution, the HDD should read 32× more data than the CD, simply because more data is packed on each track. As we can see, this is why DVDs and BDs have higher transfer rates than CDs. But a 25-GB single-layer Blu-ray disc absolutely pales in comparison to even a cheap, basic 1 TB HDD. Though however, multi-layer optical discs and multi-side multi-platter hard disks make this calculation more complicated.

Bonus

To add insult to injury, my experience shows that after inserting an optical disc into a drive, it takes about 20 seconds to begin reading any user data on it. This start-up time is far worse than flash drives and even hard drives, and is especially painful when sifting through many discs.

9
  • 5
    I have to say this is the only good answer IMO. It doesn't misattribute the speed difference to RPM and is well explained. However, your 32x higher density is not entirely correct: HDDs have multiple platters. Edit: heh, stopped reading too soon because you mentioned it. However to be more realistic maybe you should include the number of platters in a your-average-joe 1TB HDD
    – aross
    Sep 10, 2021 at 8:33
  • 1
    Good answer, except the multi layer/platter part is missing.
    – Nobody
    Sep 10, 2021 at 10:25
  • 1
    The density argument seems incomplete. Just going by the density figures you’ve provided it seems like HDDs should be dozens of times faster than Blu-Rays rather than 2-3x faster.
    – jl6
    Sep 10, 2021 at 19:24
  • 1
    @jl6 Good points. Partial explanation: Optical discs can rotate at ~10000 RPM while most hard disks top out at 7200 RPM, factor of 1.4× benefit for optical. Optical discs have a diameter of 120mm vs. 90mm for hard disks, 1.3× benefit. Shingled magnetic recording HDDs increase the number of tracks without necessarily increasing the density of bits per track. And finally, remember that linear density is approximately proportional to the square root of areal density.
    – Nayuki
    Sep 10, 2021 at 19:51
  • 1
    I’ve already upvoted, but I would love to see that last comment explaining the apparent discrepancy make its way into the answer, too.
    – KRyan
    Sep 12, 2021 at 4:46
32

While the other answers are correct, there is more. Optical media is manufactured to be cheap and lightweight. Its just a small piece of plastic. It it not designed to spin at high speeds. The faster it spins, imperfections in the plastic start to cause the plastic to warp. The faster it spins, the more it warps. This warping will make it unreadable by the drive. Spin the media fast enough and it will fail... violently. Here is a video showing the warping and failing at the extreme.

The fastest CD/DVD/BD optical drives spin at ~10k RPM at most.
Really high-end magnetic drives spin at 15k RPM, although consumer HDDs (and bulk-storage enterprise HDDs) typically spin at 7200 or 5400 RPM.

9
  • 5
    This just isn't true, modern opticals disks actually spin faster than standard modern hard drives.
    – Nobody
    Sep 11, 2021 at 14:04
  • 5
    @Keltari See Nayuki's answer, as well as various Wikipedia links. Optical drives routinely go up to 10000 RPM - though there are tradeoffs involved, especially noise, and a tiny risk of disc breakage.
    – Nayuki
    Sep 11, 2021 at 15:31
  • 3
    "This just isn't true" is plenty of a rebuttal if the details are available by just scrolling down on the same web page. Furthermore one minute of googling tells me that a blu-ray drive supporting 10krpm can be had for around $160 from stock.
    – Nobody
    Sep 11, 2021 at 16:08
  • 3
    15k drives have been dying out for at least 5 years. Do more research next time.
    – Nobody
    Sep 11, 2021 at 18:56
  • 2
    @Keltari I doubt that the sort of exotic 15k RPM hard drives, despite being noteworthy, should be a consideration in the comparison. They're outliers. Someone probably has a 50k RPM HDD in a lab somewhere. Also, the beyond the press video is showing a CD (not even a DVD) being spun at 30k RPM until it fails, so it could probably go up above 10k RPM without warping, even if only a bit. But disregarding even that, the speed of a 3.5" 15k RPM HDD works out to 68 m/s on the edge, and a 10k RPM CD works out to 63 m/s. Tiny difference.
    – aross
    Sep 13, 2021 at 9:37
19

Hard drives are precision assembled in a clean environment and sealed. They can be (and are) made to very high tolerances.

CD drives are open and designed to accept media that are clean but may have dust and fingerprints on the surface. These cannot be built to the same tolerances as sealed hard drives.

CD drives turn slowly. Hard Drives turn at 7200 rpm (and some expensive drives at 15,000 rpm) in order to provide faster data access and transfer. Tolerances have to be different. The HDD head moves over the platter surface in as little as 3 nanometers. Dust (cannot be introduced except by opening the drive) or stopping the drive rotation while the head was over the drive would damage the platter. Particulate matter can be created in a crash.

They are very different devices used for very different purposes and are not designed to be the same.

13
  • 7
    All HDDs have an internal filter medium To catch any foreign particulates that enter the drive (the rotational g-forces auto-cleans the platters, flinging any particulates into the filter). Unless someone has opened an HDD, it's unlikely there would be damage due to foreign particulates, as there's simply no way for foreign particulates to enter the drive other than by the case being opened/compromised.
    – JW0914
    Sep 8, 2021 at 23:53
  • 4
    There have been isolated incidents of optical media with impurities and imbalances literally "exploding", in the early days of higher-speed CD drives. imgur.com/a/exqSj Sep 9, 2021 at 17:15
  • 8
    CD drives do not turn slowly. They can operate in the neighborhood of 7200 RPM. en.wikipedia.org/wiki/CD-ROM#CD-ROM_drives
    – Nayuki
    Sep 9, 2021 at 22:29
  • 7
    Sure, playing an audio CD at 1× is slow. But if you're bulk-reading a data CD in a 52× computer drive, it will likely spin up to around 10,000 RPM.
    – Nayuki
    Sep 9, 2021 at 22:41
  • 8
    The reason CD sequential-read data rates are slow at the same RPM as HDDs is the much lower areal density (actually linear density; spacing between tracks of the spiral isn't relevant). Fewer bits per unit time are flying under the read head. DVD improves that, and Blu-ray further, but still not to modern HDD levels of ~200MiB/s. For random access, seek times are also much better for HDDs because of precision tolerances in HDDs. Sep 10, 2021 at 3:01
12

Don't forget, the CD-ROM was an extension of the audio CD. The underlying technology was originally designed with music in mind. It was designed to be fast enough for audio playback, and then later Sony piggybacked the CD-ROM on top of it. This isn't a format designed with high-speed data transfer as the target use case.

But why do HDDs spin faster

Because rotational speed (and performance in general) is a design priority for hard drives. Hard drives are precision-manufactured devices. The spinning platters are almost perfectly uniform in terms of material composition, and are permanently attached to the motor shaft.

Optical drives, on the other hand, were designed for a business case where cost is more important than pure performance. They're primarily made out of fiberglass because it's cheap, transparent, and lightweight. An optical disc's density is far from uniform, however. That means they're not as perfectly balanced as the platter in a hard drive. When you spin something that's not balanced, it wobbles. Vibration is an enormous problem in optical drives since the read mechanism involves a laser reflecting off the disc surface; a vibrating disc sends reflections off at an angle, missing the detector and making it more likely that a bit will be misread. That's a big reason why drive read speeds haven't gotten much faster in decades. Ever used a portable CD player and jostled it enough that it skipped? That's what's happening. DVD and Blu-Ray drives actually spin slower than a lot of CD-ROM drives precisely to minimize vibration-related problems.

Optical discs are also removable, which means they only attach to the motor's shaft via a friction-fit system. If the shaft spins too fast or accelerates too quickly, its grip on the disc will slip and the disc and shaft will start grinding on each other.

Why don't they manufacture the arm inside the optical disc drives to move at the speed of HDDs?

The read heads of these two drives use wildly different mechanical principles. The read head on a hard drive looks somewhat like a record player. It's a long arm dangling over the spinning platter that can pivot to reach different points along the disc's radius. The arm does not contact the platter, and can whip back and forth through the air relatively quickly thanks to some strong electromagnets.

An optical disk uses a laser mounted on a rail. A motor pulls a chain that makes the laser module slide back and forth across the rail. This type of mechanism has significantly more mass and friction than a hard drive's air-suspended read head. There's no reasonable way to make it move as quickly without damaging something.

Also, can't the head go to a specific part of the spiral track and read from there?

Absolutely, that's how you can seek to the beginning of a specific song on a CD.

9
  • 5
    Per en.wikipedia.org/wiki/CD-ROM#Transfer_rates, 52x CDROM (a common top speed) drives spin the disc at a speed of 10,400 rpm when running at top speed. (Constant angular velocity, so higher linear velocity towards the outside of the disc, with higher data rates there.) This is faster than consumer 7200 rpm HDDs, and twice as fast as 5400 rpm HDDs which are becoming common again for less power-intensive (than 7200rpm) bulk storage in systems that use SSDs for performance-sensitive stuff. Sep 10, 2021 at 3:08
  • 1
    The sequential-read data rate difference is due to much lower bits-per-cm on CD (or even Blu-Ray) vs. magnetic. (You're right about the seek-time reasons for random access, though.) A fast consumer HDD can read at ~200MiB/s at the faster end of its platters, vs. 7.6 MB/s for 52x CDROM, or 72 MB/s for 16x blu-ray (en.wikipedia.org/wiki/Blu-ray#Drive_speeds). Again at the faster end of the discs. Optical disks spiral out from the center; HDDs typically map sector 0 to the outside where it's fast. Sep 10, 2021 at 3:12
  • 6
    Its not fibreglass, its polycarbonate
    – Journeyman Geek
    Sep 10, 2021 at 7:12
  • 1
    Actually the head can only seek to approximately the right area, then the head has to start reading the disk to see where it actually is at and hope that the target data finds its way under the head relatively soon.
    – psusi
    Sep 10, 2021 at 17:58
  • 1
    @user253751 Aside from physically moving the laser to the correct location and waiting for the disc to spin to the desired offset, audio CDs read in and buffer several seconds worth of information before they start playing. That way, minor bumps and hiccups don't have to impact playback. A lot of that delay is waiting for the buffer to fill. Drives with large skip buffers can take noticeably longer to seek between tracks.
    – bta
    Sep 10, 2021 at 17:59
9

Hard drives don't spin much faster.

HDD rotation speeds range from 4,200 rpm (super power friendly) to 10,000 rpm, with 5,400 and 7,200 rpm being the most common. 15,000 rpm server drives used to be on the market, but SSDs made them un-economical to produce so they haven't been manufactured since 2016.

A 48x CD-ROM drive spins the disc at 9,600 rpm, a 24x DVD-ROM drive spins the disc at 14,000 rpm, and a 16x BD-ROM drive spins the disc at 13,000 rpm.

So it's not a lower spin rate at all; optical discs simply have a lower data density, so the head sees less data per revolution, and that's what limits transfer rates. And they have lower data density because... well, they're very different technologies, so it's hard to make a fair comparison, but a large part of what lets hard drives store bits in such a small area is the fact that the head sits mere nanometers away from the platter surface, which is possible because everything is carefully constructed and then sealed in an airtight container. Optical discs are cheap, removable, and prone to getting dirty and scratched, so all of the tolerances have to be much bigger.

0
6

This isn't directly related to the question but it'll help you understand the answers here and why HDD has to be sealed, which is a point repetitively mentioned in the various answers here.

First, its worth pointing out how incredibly small the read/write head is. I thought it was big and couldn't understand how it can "crash" to a dust particle.

So to read and write the magnetic information on the hard disk, the read/write head of the HDD is positioned incredibly close to the platter. It floats less than 0.1 microns over the surface of the platter. A micron (or micrometer) is one-millionth of a meter, meaning that the read/write head is less than 1/10th of a millionth of a meter from the platter's surface.

A tiny piece of dust can make the read-write head bounce up and down, crashing into the platter and damaging its magnetic material. A head crash often causes catastrophic damage to a hard drive all you need to imagine is lowering something metallic onto a circular disk that is spinning at thousands of rotations per minute. Even the slightest impact is going to cause significant damage sending platter debris across the surface.

To illustrate this, this picture shows a comparison of the sizes of a read/write head, an average dust particle (which is 2.5 microns), and an average human hair (which is 50 microns). In looking at the differences in size, it is easy to see how a simple piece of dust or hair on a platter could cause the hard disk to crash, and why the internal components are sealed inside the hard disk assembly. enter image description here

Resources: https://www.dataclinic.co.uk/what-is-a-head-crash-hard-disk-drive-information/ https://www.sciencedirect.com/topics/computer-science/magnetic-hard-disk

Also from a deleted discussion room by John: Platters in small drives are often glass, not metal, but contacting the magnetic coating in either case damages it. Remember a hard drive is meant to be read and written thousands of times. No read only CD (most) can do this at all and any writeable CD is not good for many thousands of writes. Again, the two devices (CD and HDD) are as different as night is to day.

7
  • 6
    What does this have to do with the question?
    – Barmar
    Sep 9, 2021 at 18:54
  • 4
    ... so that huge block is < 0.1 microns, and that small dot is 2.5 microns? Sep 9, 2021 at 21:38
  • 9
    @Peregrino69: The labelling is certainly unclear, but the "huge block" isn't <0.1 microns, though it is 0.1 microns away from the surface.
    – psmears
    Sep 9, 2021 at 21:51
  • 1
    Yeah, I know what he means - but that ain't what he says. He's comparing "the sizes of a read/write head, an average dust particle..." :-) Sep 10, 2021 at 7:08
  • 1
    Oh, actually..you're right. But some answers in attempt to explain why HDD spins fast mentioned how sealed it is and why it has to be that sealed, that's why. Understanding that part helped me understand the other's answers better :)
    – Manar
    Sep 13, 2021 at 0:03
2

One aspect that wasn't considered so far is: They are not just "hard disks", but also "fixed disks". That means the media (as it cannot be changed) can be extremely clean and of controlled quality, allowing extremely high bit densities through minimum distances between surface and read/write head.

Also CD/DVD/etc. changeable media may be dusty more or less, and they all "wobble" more or less, forcing the laser optics to re-focus. Some make terrible noises at high RPM, shaking the whole drive (and thus maybe force reduction of RPMs in turn). Many drives detect vibration and slow down automatically.

16
  • 2
    Drives have multiple platters, but I was under the impression that they normally only read from one at once. Sequential transfer speed doesn't scale up for larger-capacity models of the same drive with more platters. The head-alignment tolerances are tight enough that keeping the head over a track requires sensing the magnetic track underneath it and using an active feedback loop to keep the head in position for the active platter. Doing this in parallel would require independent actuators for each head. (I think I've heard of some drives actually doing this, but it's costly.) Sep 11, 2021 at 22:47
  • 2
    With the heads all controlled by the same voice coil actuator, having the head perfectly aligned for one platter might not have it perfectly aligned for another. Even if the HDD was formatted initially to match the heads, thermal expansion and other factors that could make tiny changes to the position of the heads relative to each other on the actuator could throw that off, and mean you no longer have all heads perfectly aligned. Perhaps with lower areal density you'd have more tolerance for having multiple heads sufficiently lined up (esp. for reading), but capacity is king. Sep 11, 2021 at 22:51
  • 2
    Huh? HDDs already have DRAM buffers on board, and have for years. Like 8 or 16 MiB years ago, these days often 256 MiB. This allows host write commands to execute while the drive is finishing a read. Or to execute without waiting for a mechanical seek. With SATA NCQ (and earlier IDE) command queueing, it even allows the drive to optimize between requests and complete them out of order, in an order that optimizes for physical seek times. (The drive controller knows where everything phyiscally is, the host only knows linear sector addresses; C/H/S was "fake" for years before being dropped.) Sep 12, 2021 at 0:01
  • 3
    "Head Switch Time" on that wikipedia page is described as Additional time required to electrically switch from one head to another, re-align the head with the track and begin reading; only applies to multi-head drive and is about 1 to 2 ms. That clearly implies not reading from multiple platters at once, but instead switching to the next head in the same "cylinder" (set of tracks for the same actuator position). Sep 12, 2021 at 0:22
  • 2
    why the head would need re-aligning when "switching" tracks - because, like I said, tolerances in stuff like thermal expansion coefficients means that being perfectly aligned on the top platter doesn't necessarily mean it's aligned well enough on other platters. It doesn't have to move any significant distance, but it does have to verity (and possibly adjust) its position using the new active head as part of the positioning feedback loop. Remember that the distance between tracks is only tens of nanometers. Oct 1, 2021 at 6:52
1

In addition to the other answers about why hard disks have better performance than optical disks, another factor (only touched on by @WG481) is that hard disks almost universally have multiple platters that are being accessed at one time, giving parallel read/write opportunities over multiple platter surfaces at the same time. Where an optical disk only has one read/write opportunity, so all data is forced to be read/written in serial fashion.

4
  • 6
    A natural thing to think, but it's not true. The head stack only has a valid "seek" on one platter at a time; the heads for other platters aren't guaranteed to be aligned to a track at all, and aren't used for read/write until the drive seeks to a track on that platter. The tolerances are good but not that good.
    – hobbs
    Sep 10, 2021 at 9:06
  • @hobbs, then why do higher capacity drives that just add more platters but have the same areal density get higher transfer rates? The tracks aren't pre-recorded on the platter; they are put there by the head, so however the head is aligned when the disk is assembled is where the tracks will be written.
    – psusi
    Sep 10, 2021 at 17:52
  • 2
    @psusi A) the question I linked is about the fact that they don't get higher transfer rates (at least not linear in the number of platters, you might get a slight improvement in some measures due to less head travel). B) not an expert here, but the tracks are nm wide, and you can't count on every platter to have the identical thermal expansion, or for the spindle bearings to have worn perfectly evenly. You have to servo to the track you're reading, and with only one actuator that can only be one of them.
    – hobbs
    Sep 11, 2021 at 2:16
  • @hobbs, Well shoot... I can't seem to find good evidence to back my claim today, but I am sure I remember reading drive spec sheets in years past and seeing model A has 2 platters and model B has 3 platters and model C has 4 platters, and they all claimed the same areal density and rpm but the throughput stepped up markedly between each model.
    – psusi
    Feb 24, 2022 at 21:00
-1

A Blu-ray laser is rather large, several millimetres, and it's made to fit a slightly unstable optical disc and read it from further away, and the photosensor equipment is also heavy and slow-

cumbersome Ti:sapphire lasers to operate in the labs where they are being developed. To get them out of the laboratory and into consumers’ hands, the lasers powering such devices must shrink while maintaining the ability to generate picosecond-long high-energy pulses at high repetition rates. Also relevent is Density, a DVD platter is 5GB, and a HDD is currently >500GB. That also provides a 100X boost in read/write speed, without adding the higher mechanical speed of the HDD.

The HDD read-head is also smaller, narrower, lower powered and lighter to move across the disk, because it doesn't need a laser and a receiver.

2
  • 1
    0) Speed varies approximately with linear density, which is proportional to the square root areal density. 1) Heat-assisted magnetic recording (HAMR) HDDs might contain a laser.
    – Nayuki
    Sep 11, 2021 at 19:37
  • 0> the dimensionality of the density is trivial, my point is that nobody mentioned density previously did they? 2> Blu-ray lasers are about 1000 times bigger than those used for heat assisted magnetic recording, and that's without mentioning the photo sensor equipment Sep 12, 2021 at 3:55

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .