Why do hard drives not use larger platter sizes anymore?

Question

In reading a related question about floppy drive capacity, I thought to ask a question I've had for a long time. When I started programming, all PC hard drives used the 5¼" platter size. There were very large capacity drives available that used multiple platters / heads to reach > 2GB but resulted in large enough devices that they mostly ended up in external enclosures.

Now, I know that platters shrunk because of improvements in hardware and media; that is not my question. As the density improved it was more convenient to put the added capacity in a smaller size so it would fit in smaller cases and even laptops; that is all well and good.

My question is if you can fit X number of tracks on a 3½" platter, why not use the same density on a 5¼" platter and have even more tracks? I know there are trade-offs like latency due to head-travel or cost / complexity if you tried to put multiple heads on the same platter-side, but in any case you end up with a capacity that is unobtainable with the smaller platter.

Answer 1

My question is if you can fit X number of tracks on a 3.5" platter, why not use the same density on a 5 1/4 platter and have even more tracks? I know there are tradeoffs like latency due to head-travel or cost / complexity if you tried to put multiple heads on the same platter-side but in any case you end up with a capacity that is unobtainable with the smaller platter.

There are many factors, but I guess it all comes down to two: Technical and Marketing.

On the technical side, a wider track area (assuming that's what you meant with larger platter) also means a longer distance covered (assuming the same disk technology). This has several drawbacks:

1. The head carrier (arm) needs to extend in size (get longer), which adds more flexibility and more deflection (at the same energy) due vibration.

2. A head carrier mounted on a single pivotal arm will have to cover a wider arc over an increased track area (even though it is mounted farther away) which leads to a greater misalignment of the head slot(s) at the inner and outermost tracks.

3. Increased part size can come with increased tolerance.

4. Increased part size does come with increased temperature related shrinkage/extension in absolute numbers. If head carrier and media are of different materials their coefficients will differ.

5. If a longer distance is taken by the head carrier moving between track there is a higher chance that it will over or undershoot, thus requiring re-positioning.

6. Since heads fly, the distance varies in relation to the speed of the rotating surface resulting in a changing head surface distance.

This is not an all encompassing list of issues, just the first that come to mind. Taking them into account would mean for each a reduction in track density (1,2,3,4), increase of access time (4,5) or decrease of bit density (2,6).

Each of these issues can be addressed, but with that comes increased complexity and cost - with little return (e.g. density) or even negative results (increased access time).

On the Marketing/Sales side of things issues are:

1. Simple drives can be cheaper and therefore can be sold in larger numbers.

2. With the right production numbers, two smaller drives may be less expensive than a larger cutting edge drive.

3. The PC market especially asks for smaller drives to fit ever smaller computers. That's why even 'double' height 3.5" drives vanished despite offering more than double the room to stack platters.

Here it may be helpful to look at the range of disks offered. The majority of disks come in a middle of the road configuration - which usually hits the production price and capacity sweet spot. Increasing the capacity past this point comes with a steep increase of cost per capacity. Below that point capacity drops don't usually result in reasonable savings. The later drives tend to vanish soon from the market, while the former are reserved for markets where maximum capacity and/or access time is worth the additional money.

It might be noteworthy, that for mainframes 13" drives were developed way into 1990, and 5.25s at least until the 2010s - with capacities that might have sounded like a rip-off - unless other criteria, especially reliability where taken into account. Here drives were designed for 24/7 operations over more than a decade(!) without producing a single fault. All with low - and most importantly, repeatable - access times. Capacity was a criterion way down the requirement list.

Answer 2

Larger radius = lower RPM, which makes pretty much all timing specs worse.

The Quantum Bigfoot was limited to 3600 RPM. Why not spin faster, like a Cheetah X15 (15 kRPM)? Flutter and buffeting. Halving the radius gives four-times reduction in these effects (assuming no spindles -- real platters are on spindles and the coupling to these spindles provides more radial stiffness than just being a flat disk, so 3.5" can be a bit larger than half of 5.25").

Answer 3

Like you said, seek time and latency are longer for larger-diameter platters. Another factor higher is power usage at start-up due to inertia. Lastly, for mature technologies smaller is very often cheaper to manufacture.

Answer 4

Smaller platter sizes let you fit more individual drives into the same space. Then you can setup RAID-5 or 6 to keep the data online when a drive or two fails (this also reduces the need to restore from backup), or you can use RAID-0 if your primary objective is performance. And because you can join multiple drives together to achieve larger volume sizes, there's really no need for larger platters, so why not take advantage of economies of scale by choosing a platter size that can be used for both business and home use?

Answer 5

Let's assume, for the sake of argument, that all timing issues are the same -- large (diameter) disks have the same latency and data transfer rate as small disks. So why aren't larger disks used?

If you have two smaller disks, each with half the capacity of one larger one, you can have two I/O operations going on at the same time with the smaller disks, as long as the data being accessed is on different disks. With one larger disk, the I/O operations have to queue up, one behind the other.

If two smaller disks cost the same as one larger one, the smaller disks will be faster for that reason.

Plus all of the other reasons that have already been mentioned.

Answer 6

It simply doesn't make sense.

You can slightly-more-than double the capacity by using a 5.25 instead of a 3.5 inch form factor. Doing so can be achieved by putting in one extra platter, too, but at a fraction of the technical problems introduced. Also, consider that in terms of density, during the last decade we're talking about improvement rates which make "double" look like a silly joke.

Nobody really needs double-digit or triple-digit terabyte drives, but those who do need them will use a redundant network of servers with RAID on each server anyway. Still, single-digit terabyte drives (which nobody truly needs either, if you are honest) are readily available in 3.5 inch form factor at neglegible cost.

Further, SSD is evolving much more rapidly than HDD, so you can expect HDD technology to eventually die out anyway. Terabyte SSDs are not precisely a bargain, but they do become kinda affordable. If thought and R&D goes somewhere, it'll be there, nobody will want to spend a lot of BD re-establishing 1960s technology. Go forward 10 years, nobody will be using a spinning, mechanical disk any more. Things need to be small and smaller, and they need to consume less and less energy (because virtually everything is going towards mobile one way or the other).

Among the technical problems with larger platters are not only the ones elaborated very well by Raffzahn and some others which are rather obvious (increased angular momentum in everyday situations, to name one). You also need to consider that in addition to wear being much higher and seek times being much more unreliable than they are already on 3.5 inch, the sequential data rate differs much more in a larger disk.
That, or data density, if you are willing to give away most of the benefit of a larger disk. But then, if you are willing to give this away, why would you use a larger disk in the first place, that doesn't make sense.

So, in addition to having several very real mechanical problems, you also have some practical problems. For example, you need to be capable of writing out data at much higher rates (or spin the drive slower overall) because more area moves under the head in the outer regions. You also need a lot more energy because the head "touches" (well it hopefully doesn't touch!) the same area for a much shorter time.
The larger a disk (assuming the same speed), the more pronounced everything gets. For example, the technical specifications of your drive get a lot more unreliable. It's nice when you can claim a (average) sustained rate of 200 MB/s, but if the actual rate varies from 80 to 300, that's not quite so cool. Same goes for access time. If you write 8ms on the box, and actual times go from 6 to 18 ms, that's not cool PR-wise.
Access time is one of the most dominating factors nowadays, anyway. Throughput is rarely an issue. Anything that makes (or may make) access time worse is bad mojo, even moreso as SSD kinda guarantees a mostly-constant, low access time.