Moore's Law: Hard Drive Edition?

Moore's Law, first identified by Intel co-founder Gordon Moore in 1965, states that the number of transistors on an integrated circuit doubles approximately every two years. This has held true for CPU and memory performance, but does it apply to hard drives? Hard disk drives (HDDs) have traditionally kept up with Moore's Law in terms of capacity, but their performance has lagged behind. The mechanical limitations of HDDs prevent them from achieving faster rotational speeds, and their random IO performance has not improved significantly. Solid-state drives (SSDs), on the other hand, have brought storage back in line with Moore's Law, with continuous performance gains and greater capacity points. While HDDs have increased in capacity over time, their mechanical limits and physical constraints on disk density impact their ability to keep pace with Moore's Law.

Hard drives and mechanical limitations

Hard drives have mechanical limitations that prevent them from achieving faster rotational speeds. This has resulted in an exponential gap between the advancements of CPUs, memory, and networking, which continue to follow Moore's Law, and hard drive performance.

Hard disk drives (HDDs) are electro-mechanical data storage devices that store and retrieve digital data using magnetic storage with rapidly rotating platters coated with magnetic material. The performance of a hard drive is determined by factors such as seek time, rotational latency, and data transfer rate.

Seek time refers to the time it takes for the read-and-write heads to move to the track of the disk that contains the data. Rotational latency is the time incurred for the desired disk sector to move under the head when data transfer is requested. The data transfer rate is the speed at which the data is transmitted once the head is in the right position.

While advancements in HDD characteristics have led to increases in capacity and performance over time, the mechanical limitations of HDDs have constrained their rotational speeds. In contrast, solid-state drives (SSDs) have emerged as a solution to overcome these limitations. SSDs offer higher data transfer rates, higher areal storage density, better reliability, and lower latency and access times.

The mechanical limitations of HDDs have led to a lag in performance compared to the exponential advancements predicted by Moore's Law for CPUs, memory, and networking.

Kryder's Law and hard drive density

Moore's Law, first identified by Gordon E. Moore, co-founder of Intel, in a paper he wrote in 1965, observes that the number of transistors on integrated circuits doubles approximately every two years. The law has continued to hold true for CPU and memory performance, allowing them to scale to meet the demands of data-intensive environments.

However, Moore's Law does not seem to apply to traditional storage devices like spinning hard disk drives (HDDs). While HDDs have kept up with Moore's Law in terms of capacity, their performance has slowed as capacities have increased. This is due to mechanical limitations that prevent hard drives from achieving faster rotational speeds. As a result, an exponential gap has emerged between the advancements in CPU and memory and the capabilities of storage.

The observation, made in the mid-2000s, noted that the capacity of magnetic hard drives was increasing at a rate faster than that predicted by Moore's Law. In just over a decade, hard disk capacity had increased 1,000-fold, a rate that Intel founder Gordon Moore described as "flabbergasting." In 2005, commodity drive density reached 110 Gbit/in2 (170 Mbit/mm2), a significant increase from the 100 Mbit/in2 (155 Kbit/mm2) of circa 1990.

Kryder projected in 2009 that if hard drives continued to progress at a rate of about 40% per year, a two-platter, 2.5-inch disk drive would store approximately 40 terabytes (TB) and cost around $40 by 2020. However, by 2014, it was clear that the actual rate of areal density progress was falling short of this forecast. The validity of Kryder's projection was questioned, and the term "Kryder rate" was coined to refer to the observed rate of areal density progress, which was significantly lower than the predicted 40% per year.

By 2019, it was acknowledged that Kryder's law had become outdated as the cost of media storage was decreasing at a slower pace and stabilising. Nonetheless, Kryder's Law has had a significant impact on the digital world, contributing to reduced costs and increased capacities for data storage, enabling the development of data-heavy technologies such as cloud computing, big data analysis, and multimedia applications.

SSDs and flash memory

Moore's Law, first identified by Intel co-founder Gordon Moore in 1965, states that the number of transistors on integrated circuits doubles approximately every two years. This law has driven innovation in various fields, including smartphones, the internet, and big data.

While traditional spinning hard disk drives (HDDs) have kept up with Moore's Law in terms of capacity, their performance has sometimes slowed down as capacities have increased. This is due to mechanical limitations that prevent HDDs from achieving faster rotational speeds. As a result, an exponential gap has formed between the advancements in CPU and memory and the performance of HDDs.

Solid-state drives (SSDs) and flash memory have played a crucial role in addressing the limitations of HDDs. SSDs have brought storage back in line with Moore's Law, with both performance and capacity following the expected curve. For example, SanDisk® offers enterprise-grade SAS solid-state drives with significantly improved data transfer rates and capacity compared to just a few years ago.

However, SSDs that use flash memory are facing challenges as they approach the limits of Moore's Law and experience diminishing returns. When the etching process moved below 25 nanometers (nm), quantum effects started causing more errors, requiring more error correction code and processing power, leading to lower speeds and higher costs. Manufacturers have temporarily addressed this issue by adopting multilayer flash technology, which returns to the 25 nm process but uses multiple layers to increase capacity.

Additionally, increasing the number of bits per cell in flash memory has been another strategy to enhance capacity. While this approach boosts capacity, it also increases complexity and reduces the lifespan of each cell in erase cycles. Despite these challenges, SSDs continue to offer advantages over HDDs in terms of capacity and performance, and further developments in flash lithography are expected to drive even greater performance gains and capacity points in the future.

HDDs and mechanical limits

Hard disk drives (HDDs) are electro-mechanical data storage devices that use magnetic storage to store and retrieve digital data. They have mechanical limitations that prevent them from achieving faster rotational speeds and, therefore, higher performance. While HDD capacity has increased over time, performance has slowed down as capacities have grown. This is due to the mechanical limitations of HDDs, which include the physical movement of the actuator arm and the rotational speed of the platters.

The performance of an HDD is determined by its access time and data transfer time. Access time refers to the time it takes for the drive to start transferring data, which includes the seek time—the time it takes for the head assembly to move to the track of the disk where data will be read or written. Data transfer time, on the other hand, refers to the rate at which data is actually read from or written to the disk.

The seek time of an HDD can vary significantly depending on the initial and final location of the head. The average seek time is determined by statistical methods or approximated as the time of a seek over one-third of the number of tracks. The first HDDs had average seek times of about 600 ms, while modern HDDs have improved to around 4 ms for high-end server drives. However, the seek time has not kept up with the increase in capacity, resulting in a distortion where the bottleneck has shifted to storage.

In addition to seek time, the rotational speed of the platters also affects the access time of an HDD. Rotational latency, or the time it takes for the desired sector to move under the head, depends on the rotational speed of the disks. By increasing the rotational speed, the access time can be improved. However, there are limitations to how fast the disks can spin due to heat and vibration. As a result, increasing areal density has become the main method to improve sequential transfer rates.

Overall, while HDDs have continued to increase in capacity and density, their performance has lagged behind. The mechanical limitations of HDDs, including seek time and rotational speed, have prevented them from achieving faster rotational speeds and higher performance. This has created a gap between the advancements in CPU, memory, and networking, which continue to follow Moore's Law, and the performance of HDDs, which has remained relatively static.

The future of Moore's Law

Moore's Law, first observed by Intel co-founder Gordon Moore in 1965, states that the number of transistors on an integrated circuit will double approximately every year. This prediction has held true for nearly 50 years, but there are signs that it is reaching its limit.

The law has driven innovation in smartphones, the internet, and big data, but it has not applied to storage in the same way. Traditional hard disk drives have kept up with Moore's Law in terms of capacity, but performance has slowed as capacities have grown. This is due to mechanical limitations that prevent hard drives from achieving faster rotational speeds.

The solution to this problem has been the development of Solid-State Drives (SSDs). SSDs have brought storage back in line with Moore's Law, with performance and capacity increasing at the expected rate. However, SSDs that use flash memory are reaching the end of Moore's law and seeing diminishing returns as quantum effects set in.

To address these quantum effects, manufacturers have turned to multilayers and more bits per cell, but these are only temporary fixes with drawbacks. It is physically impossible to continue the miniaturization process indefinitely, as there is currently no reliable technology to manufacture transistors at the subatomic level.

While Moore's Law may continue to hold for another five or ten years, it is clear that new technologies and manufacturing techniques will be needed to sustain the current rate of progress.

Frequently asked questions