Layman's Guide to Computing - Season 10

Issue 130: Power limits

2021-07-24T08:00:00+08:00

Previously: The larger the surface area, the faster an object loses heat. The larger the temperature difference between an object and its surroundings, the faster the object loses heat. Heat is bad for computers, and CPUs will need cooling to be able to process computations quickly. A mobile phone thus typically uses no more than 4 W of power, a laptop can use 25–45 W, and a desktop can usually use 65 W and more. Two popular ways of increasing the cooling capacity of a device is to attach a larger piece of metal to the chip (passive cooling), or use a fan to force air over the heatsink (active cooling).

Point 1: A powerful device produces lots of heat.
Point 2: A device that produces lots of heat needs a large surface area (directly in contact with the heat source) to stay (relatively) cool.

These are two of the primary factors determining how tiny a computer can be. Can something the size of an iPhone be as powerful as something the size of a Macbook? It depends on how much cooling is available to it!

One more factor to add in this issue: power. Without power, none of your devices would work … and that is one more source of heat to be dissipated, incidentally.

Batteries and DC power

Some devices need to carry their own stored energy (in the form of batteries); the devices are powered by direct current (DC) from batteries. In this form of power transmission, electric current only flows one way—this is why it is important to put batteries in the right way! Electric current comes out from one end, and re-enters from the other end, marked with + and – symbols.

Batteries seldom provide power at the voltage required by devices and CPUs. For example, smartphone batteries usually have a voltage of about 3.7 V, even though the CPU usually requires about 1 V to operate. This allows the batteries to power the device with a low current¹, so as to minimise the heating effect of current in the wires.

That means some power conversion has to take place on the smartphone’s mainboard. Fortunately, DC-to-DC conversion is highly efficient (though not 100%), so it doesn’t contribute much heat in the device.

Wall sockets and AC power

On the other hand, alternating current (AC) from wall sockets has electric current flowing both ways—the current switches directions 50 or 60 times a second (see this PDF of Worldwide AC Voltages & Frequencies). Connecting this directly to a device that needs DC is looking for trouble! This AC power source has to be converted to DC through an AC-DC converter (a.k.a. “power brick”, “power supply”, “AC adapter”, …), and that process currently only goes up to 90% peak efficiency².

That means if you have a desktop running at 100W (maybe while gaming or encoding video files), the AC-DC converter alone draws 111W at the wall socket, wastes 11W (in the form of heat), and provides 100W of power to the desktop.

And heat is the enemy of CPUs.

External vs internal power supplies

Ughh, power bricks … so many different types, with different connectors, and we never quite know if we can use one laptop’s power brick on another laptop (at least, until USB Type-C power for laptops came along; more in Issue 127))

In a laptop, you really do not want the power supply dumping this heat into the laptop! The laptop already has enough work to do getting heat from the CPU out of that cramped space into the surroundings. You don’t want to give it more heat to remove, and risk throttling the CPU’s performance (Issue 129)). It’s better that the AC converter/power brick remains external to the laptop, dumping that heat into the surroundings directly without heating up the laptop’s internal space.

The internals of the Apple AC converter (i.e. power adapter)
Source: Ken Shiriff

In larger devices—the Mac Mini, game consoles (e.g. PS4 or Xbox One), and larger desktops, there’s plenty of space in the device’s internals, and they have sufficiently powerful cooling systems that can remove this heat. In desktops especially, the power supply may be large enough that it has its own cooling fan!

The internals of a desktop power supply.
Notice the cooling fan mounted on the back with 4 silver screws, and the two silver heatsinks mounted vertically
Source: Wikipedia

So for large devices, it makes sense to hide the power supply within the device for a sleeker look.

In fact, for high-power devices, a power adapter is a poor option. Since power adapters don’t have their own cooling fans, they have limited cooling ability, and are liable to overheat easily if they have to provide >100 W to a device (remember that this means they release at least 11 W of heat, while passive cooling can typically dissipate up to 8 W).

Device categories

Putting together the information from Issue 129) and this issue, we can deduce that devices seem to sort themselves into form-factor categories depending on how much power they draw, and how much heat they put out:

Devices drawing >100W at peak, and putting out >80 W of heat:

Generally large, directly powered from the wall (by AC), with power supply within the device.

In rare cases they do use external AC adapters (such as high-power gaming laptops).

Devices drawing 12–65W at peak, putting out 25–45W of heat:

These devices cannot be passively cooled, and thus require active cooling (i.e. a cooling fan and heatsink).

To avoid adding heat to the device from the AC-DC conversion process, they usually use external AC adapters.

Even in devices as small as the Nintendo Switch, you can usually spot the cooling vents where the cooling fan blows warm air from the device into the surroundings.

These may be safely powered by USB Type-C.

Devices drawing <12W at peak, putting out <10W of heat:

These devices can be passively cooled.

Large devices, such as tablets, have a larger surface area to dissipate heat and can afford to draw as much as 8–10W, while smaller devices such as smartphones typically have to remain under 5W.

These are usually powered by USB (at a voltage of 5V), though some may draw power at 9V.

The M1 Macbook Air is passively cooled and thus in this category because its M1 processor is configured to limit its maximum heat output to approx. 10W or less; the M1 Macbook Pro has a cooling fan and a higher max heat output configuration, which allows it to perform at greater capacity.

If you spot a device in the wild that claims to have performance much greater than its form factor—its shape, size, footprint—suggests, you would be wise to suspect over-optimism or a scam! At least until it is clear how they plan to provide that power and get rid of that heat …

Issue summary: AC power from the wall uses electric current that alternates directions, while DC power from batteries uses electric current that flows in one direction only. All electronics are DC-only, and require an AC-DC adapter to be powered from the wall. The AC-DC conversion produces a significant amount of heat; AC-DC adapters are usually external unless the device has sufficient space or cooling capacity for it.

I’m noticing a pattern: issues where I explain concepts tend to be shorter than issues where I explain limits due to engineering and the physics of reality. I hope I can shorten the latter without sacrificing practical knowledge. Let me know how you’re finding these issues :)

What I’ll be covering next

Next issue: [LMG S11] Issue 131: What do early CPUs and startup founders have in common?

This season was focused on firmware and computer components; it is part 1 of a set of concepts I need, so as to explain why the Apple M1 processor is a game-changer for personal computers. I explained what a graphics card is and what it does, I explained why some laptops are upgradeable and why some are not, I explained why some devices need cooling fans and others don’t, and I summarised the relationship between device form factors and their power limits.

Part 2 will extend this exploration inside the computer. I noticed that layfolks’ mental concept of a computer typically includes the idea that there is a CPU, memory, a hard disk/solid state drive, and maybe a graphics card inside a computer. That’s plenty good enough for everyday life; it’s like understanding that all the employees of a company are in a particular building. But it is insufficient for understanding why the M1 is so much faster; you’re going to need to know where the employees are situated, and what their workflow is like!

It’s a tempting but misleading story to imagine that Apple simply has much better engineers; I would say that their engineers were instead under the influence of incentives that allowed them to imagine a more coherent architecture. Let’s get into it starting next issue, again beginning from first principles: how exactly does the CPU, memory, and storage disk work together?

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]

Power = Voltage × Current, so to provide the same power with a lower current, you have to provide it at a higher voltage ↩
AC-DC converters typically have a range of input power they can convert (e.g. 0–65 W for laptop adapters, 0–500 W for desktops). The efficiency is highest at about 50% of that load, and efficiency drops as the load increases or decreases from that point. ↩

Issue 129: Cooling

2021-07-17T08:00:00+08:00

Previously: Upgradable parts need a slot or socket to be inserted into; these slots/sockets need to be made robust enough, causing them to take up more space than a soldered part. Devices which were designed to be small and portable generally eliminate these as far as possible, opting to have parts directly soldered to the board instead.

Why do computers need power?

Other home appliances I can understand. They need to heat up air/water, move air/water around, or extract heat from air/water to move it elsewhere. These things all need energy. But a computer … all it does is move electrons around! All the information in a computer that changes is just electrons moving; that should not need so much power, should it?

As it turns out, energy-information equivalence theories posit that manipulating information does increase entropy which does involve energy—but this is the Layman’s Guide to Computing, not a physics newsletter. Let’s just say that managing information, in its abstract sense, needs very little energy.

Energy usage in computers

What happens to all the energy that a computer uses, then? Some miniscule amount of it goes to manipulating information. A tiny amount goes to lighting up LEDs (these devices somehow always have LEDs), and maybe running the cooling fans. The rest of it is wasted as heat.

The history of computing is also a history of less and less wasted heat. The Cray 2 supercomputer, in 1985, needed 195 kW to produce 1.9 gigaflops (1.9 billion floating-point operations per second; more context in Issue 123)) of computational performance. The iPhone XS, in 2016, needed less than 1 W to produce 1 gigaflop.

No, I’m not going into the environmental concerns and carbon footprint of computing. We have more tangible and immediate concerns here.

Because heat is bad for computers. When microprocessor chips heat up, the semiconductor material they are made of no longer behaves as it should; it gets “leaky”, allowing electrons to go where they shouldn’t. The data it handles starts to get corrupted, and it eventually crashes.

Thermal throttling

In the past, microprocessors were simpler, and had no thermal controls whatsoever; if you allowed them to run as-is, without any cooling assistance, they would just run until they began smoking and sometimes even catch fire (enjoy this 2005 video of a CPU doing just that).

Today, CPUs are somewhat more sophisticated. Once they start heating up to their thermal limit, onboard thermal control circuitry will attempt to throttle the CPU’s performance to keep it at a safe temperature. So you will have a laggy and unresponsive CPU, but at least it’s not on fire!

Still, most CPUs are going to need some sort of heatsink that helps to draw heat away from it, and dissipate the heat into the surroundings.

Getting rid of heat

There are two things you need to know about heat loss to the surroundings:

The larger the surface area, the faster an object loses heat.
The larger the temperature difference between object and surroundings, the faster the object loses heat.

CPUs are kind of at a disadvantage here.

They are really small; the part that is in contact with the heatsink is usually about 400 sq mm (approx. 2 cm by 2 cm, or 0.64 sq in)¹
The thermal limit for most CPUs is only about 100 °C (212 F), compared to most metals which have theirs in the hundreds or even over a thousand °C

Heatsinks

In practice, this means that a bare chip can only run about 4 W before it starts to run into its thermal limits (we call this overheating). Anything more powerful than a basic home router or smart device (such as the Amazon Echo or Google Home Assistant) will need some sort of heatsink to avoid overheating.

The simplest way to cool a chip is to slap a piece of metal on it to increase the surface area (factor 2). This is known as passive cooling. A paste called thermal paste is applied between the heatsink and chip to improve the transfer of heat. To pack as much surface area as possible into a tiny space, this piece of metal usually has long, thin fins, giving the characteristic look of heatsinks:

A passive heatsink on the northbridge chip of a computer mainboard
Source: found on Superuser

The effectiveness of passive heatsinks depends on the ambient airflow around it. Some creative setups that manage to get the metal case itself in contact with the CPU can readily cool up to 45 W, with zero fan noise!

For mobile phones, tablets, and laptops, such heatsinks would add too much to the device thickness. Instead, the CPUs are usually in direct contact with a larger metal surface, sometimes even the metal back of the device; this is why they feel warm to you in the first place. This allows tablets to use up to approx. 8 W of power.

The logic board cover in an iPad Pro 11 has copper inside; it helps to spread heat to the rest of the device instead of concentrating it all in one spot
Source: iFixit

Where there isn’t this luxury of space, another option is to use a slim heatsink, and increase its cooling ability by forcing air through it. This form of cooling is called active cooling, and usually done with a fan of some sort, a popular option for thicker laptops. This allows laptops to run between 25–45 W, and desktop computers to run 65 W and hotter (with larger heatsinks, of course)

An active heatsink on the CPU of a computer mainboard
Source: found on Superuser

These numbers are for stereotypical mobile devices, laptops, and desktops; stranger or hybrid designs may have different cooling capacities (e.g. a tiny cube desktop might only have 25–35 W of cooling capacity).

Issue summary: The larger the surface area, the faster an object loses heat. The larger the temperature difference between object and surroundings, the faster the object loses heat. Heat is bad for computers, and CPUs will need cooling to be able to process computations quickly. A mobile phone thus typically uses no more than 4 W of power, a laptop can use 25–45 W, and a desktop can usually use 65 W and more. Two popular ways of increasing the cooling capacity of a device is to attach a larger piece of metal to the chip (passive cooling), or use a fan to force air over the heatsink (active cooling).

If you walk away with only one thing from this issue, it’s that heat dissipation constrains the max performance for any device, and heat dissipation is usually constrained by the surface area available for cooling.

What I’ll be covering next

Next issue: [LMG S10] Issue 130: Power limits

Computers turn electrical power to heat, drawing a miniscule amount off for computation. But where does that power come from? In the last issue of Season 10, I’ll touch on a highly underrated hardware component: the power supply.

This and more, next issue!

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]

Multiple CPUs are produced from a single semiconductor wafer. Keeping CPUs small maximises the yield from manufacturing, and reduces the chance of a manufacturing error on any single CPU. This is going to take a whole ’nother season to explain … ↩

Issue 128: Upgradeability

2021-07-10T08:00:00+08:00

Previously: USB Power Delivery is a specification that describes how much voltage and current can be supplied by different categories of USB cables. It allows power delivery at different levels for all kids of connected devices, up to 100W. This should help to simplify cable setups that otherwise require multiple kinds of cables between two closely interconnected devices (such as a laptop and an external monitor).

At some point in the past, computers could be upgraded with all kinds of parts: you could upgrade to a better network card, a better processor, or add more memory, without changing out the entire computer!

The short history of personal computers

After IBM released the IBM 360, the first mainframe that was meant to be used in both large-scale and small-scale applications, it quickly realised that providing service support for it was going to be a nightmare if each version required its own specialised support. In the 1970s, it was already thinking of a family-of-parts concept that allowed its products to use a set of interchangeable parts to reduce the number of these parts.

But Intel, founded in 1971, had made a major breakthrough in manufacturing low cost microprocessors. The price of computers dropped steadily in response, allowing the personal computer (PC) to become affordable to home users, and the family-of-parts concept came along for the ride.

The early computers, when purchased, needed a network card to connect to a network, needed a sound card to produce sound, needed a graphics card to view graphics … you basically needed a card for everything!

Upgradeable parts

On a desktop, you could upgrade a number of components.

Desktop CPUs were purchased separately, and had to be installed into a socket on a mainboard (also called a motherboard).

A CPU socket on a mainboard
Source: Digital Trends

Desktop computer memory was likewise purchased separately, and slotted in:

Memory slots on a desktop mainboard
Source: Bleuwire

On top of that, you still had to buy your own DVD/Bluray drives, storage disks, and other peripherals, to be plugged in laboriously to the mainboard. Many of these features made it into laptops as well.

Laptop upgrades

Because of the smaller space available, the CPU socket had to go. Most laptops have their CPUs soldered directly to the mainboard, and the CPU cannot be upgraded separately. Buying a laptop with a better CPU often means buying it with a different configuration, mainboard and all.

But laptops retained the ability to upgrade memory through the use of memory slots. These were oriented flat along the board to reduce space.

Memory slots on a laptop
Source: Laptopmag

And they also retained the ability to swap out or replace their disks, whether hard disk or solid state drive.

The hard disk slot on a laptop
Source: TweakTown

The SSD slots on a laptop
Source: EaseUS

Limitations of upgradable parts

A common limitation here is that replaceable components take up more space than directly-soldered ones, because the sockets and slots cannot be manufactured too thin if they are to be robust. As the industry moves towards thin-and-light designs, and hardware support moves to a system that replaces the whole device rather than individual parts, these replaceable parts are already on their way out.

On the Apple Macbook, Apple has already soldered down the CPU, memory, and SSD, allowing for no upgrades. This is also the case for many other thin-and-light laptop manufacturers.

Issue summary: Upgradable parts need a slot or socket to be inserted into; these slots/sockets need to be made robust enough, causing them to take up more space than a soldered part. Devices which were designed to be small and portable generally eliminate these as far as possible, opting to have parts directly soldered to the board instead.

There are of course other reasons for this transition, economic as well as financial, which I will not tackle here. I just wanted the chance to show some upgrade slots so if you ever get to see an old laptop opened up, you know what they are for!

What I’ll be covering next

Next issue: [LMG S10] Issue 129: Cooling

Next issue, another dive into another issue: why is my laptop running so hot? Why is it so loud? Why can’t I get a more powerful processor in this tiny PC? Why is my i7 slower than my friend’s i5?

This and more, next issue!

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]

Issue 127: USB Type-C Power Delivery

2021-07-03T08:00:00+08:00

Previously: USB is a (licensed) technical standard that describes how devices connect to each other through a cable. USB Type-C is a new connector standard that supports USB 3, DisplayPort, HDMI, and Thunderbolt. It is able to carry multiple types of data simultaneously, in limited combinations. In a USB connection, one device acts as the host while the other acts as the device; the host initiates all communication.

Last week, I differentiated the USB Type-C cable specification from the USB3 data specification; the former describes how the cable and connector should be, while the latter describes how to transmit USB data over a Type-C cable. Remember that in addition to USB3 data, the Type-C cable can also transmit HDMI or DisplayPort video data!

What else can a Type-C cable do? Oh, right—provide power to devices. In other words, charge them.

What’s wrong with USB2 charging?

Absolutely nothing! At this point, it was not expected that USB would be needed for powering anything more than some basic peripherals, like keyboards, mice, and anything less power-consuming than a small external hard disk. 5V (volts) of voltage with 0.5A (amperes) could provide 2.5W of power, and that was considered plenty enough.

The USB Implementers’ Forum (USB-IF) didn’t want to set too high a standard for USB devices and peripherals (including USB-certified cables); original equipment manufacturers (OEMs), the ones who actually had to manufacture the goods, would complain about the high cost. Nobody likes being undercut by cheap knockoffs that don’t bother applying for the license and following the specs (do you notice the official USB logo when buying?). So for a long time, we had up to 2.5W of power from USB ports.

And then tablets came along, drawing 8–12W of power to do whatever they have to. Aside from the iPad, which used its own connector, these were charged over USB. So their manufacturers had to come up with kludges to work around USB limitations. They had Quick Charge, Dash Charge, and all kinds of other standards which were not approved by the USB-IF, just to allow their cables to provide up to 12W of power (3A of current, more in some cases) to their tablets while charging.

USB 3.0 bumped the limit up to 0.9A (providing 4.5W), which was nice but far from enough. The hard limit was the cable itself though: drawing anything more than 5A over the usual thin USB cables would cause them to heat up to unsafe levels. Clearly, something more was needed.

USB Power Delivery

From 2012, the USB-IF finally added a Power Delivery (PD) specification, which allowed power to be delivered over USB cables at different voltages.

In addition to 5V, which is used by phones, tablets, and their power banks, the PD spec also allows charging at 9V (fast charging for some devices), 15V (for higher-power devices like the Nintendo Switch), and 20V (what most laptops use for PD charging). With a current of up to 5A, this technically allows up to 100W of power to be delivered—sufficient for pretty much all laptops.

Which voltage is used?

The actual voltage to be delivered by the charging host is negotiated with the host. When a charging device is connected, it communicates the voltages it can support, the host compares it with the voltages it can supply, and power at a supported voltage is delivered.

Can any cable be used?

Greater current requires a thicker cable to be used, as thinner cables have more resistance and will heat up to unsafe levels. A cable following the USB-PD specification will negotiate the correct voltage and current in any case, so if your cable is not charging at a level you know is supported by both your devices, do check the cable rating. You may have to buy a higher-rated cable.

Are there any advantages to buying a more expensive cable?

Users of external monitor screens and docking bays often have to connect multiple cables from those devices to their laptops for power, (USB) data, and video output. With USB Type-C and USB-PD specifications unifying these three requirements into one cable, we will (eventually) be able to connect a laptop to a Type-C monitor using a Type-C cable, and this cable will supply power plus allow the laptop to use all devices connected to the monitor.

It’s supposed to simplify the physical cable mess, at the cost of having to manage a more complicated specification. Let’s see how that plays out in the next decade.

Issue summary: USB Power Delivery is a specification that describes how much voltage and current can be supplied by different categories of USB cables. It allows power delivery at different levels for all kinds of connected devices, up to 100W. This should help to simplify cable setups that otherwise require multiple kinds of cables between two closely interconnected devices (such as a laptop and an external monitor).

What I’ll be covering next

Next issue: [LMG S10] Issue 128: Upgradeability

Millennials and other older computer users may remember the glory days of the desktop, when almost any component could be removed and swapped out for another. Laptops used to enjoy this upgradability to a lesser extent; the laptop memory and hard disk came as separate slotted cards that could be replaced with upgraded versions for increased performance.

What happened to that trend today? The reason mainly lies in the realm of economics, but I figured I’d use the chance to dig a little deeper and explain what is going on with the hardware that no longer allows this to happen.

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]

Issue 126: USB Type-C

2021-06-26T08:00:00+08:00

Previously: Analog formats such as VGA mostly contain the control signals that the CRT needs to operate, while digital formats such as HDMI and DisplayPort contain image data that the device must convert to control signals. Analog signals need a digital-analog-conversion (DAC) chip to be converted to digital signals, hence VGA-HDMI adapters tend to be more costly than DisplayPort-HDMI adapters. Dedicated graphics cards generally support more simultaneous output video streams than integrated graphics cards.

This week, I attempt to untangle the confusion around USB Type-C, informally also referred to as USB-C.

What is USB Type-C?

It is a connector standard. It sets standards for this connector:

USB-C plug
Image via Wikipedia

What does a connector standard do? It determines how many pins the connector should have, and what each of the pins should be used for, how the connector should be shaped, how the docking port (where the cable gets plugged into) should be designed, and other similar details. It’s all about the docking.

Pinout diagram of a USB-C plug
Image via Wikipedia

But won’t somebody think about the data?!

Ah, now we’re going back in history …

Universal Serial Bus (USB) is a (set of) industry standards that sets requirements and protocols for—well, how data is transferred from one device to another. It is maintained by the USB Implementers Forum (USB-IF). The first version of the standard was released in 1996, second version (USB 2.0) in 2000, and third version (USB 3.0) in 2008.

While USB 2.0 (or Hi-Speed USB) supported a transfer rate of up to 60 MB/s, USB 3.0 supports up to 625 MB/s, allowing for much faster transfers from external disks and other devices.

We don’t have to worry so much about these versions, because USB is designed to be backward-compatible. That means all devices that support USB2 also support USB1, and all devices supporting USB3 also support USB2. The primary advantage that each successive USB version has over previous versions is higher throughput, more supported features, and more connectors to confuse (okay, that last isn’t an advantage 😛).

Comparison of USB connector plugs, excluding USB Type-C plugs
Image via Wikipedia

As you can see, this makes for a lot of confusion, especially when compatibility is mixed: USB2 ‘A’ connectors are meant to go into USB3 ‘A’ receptacles, but USB2 ‘B’ connectors aren’t meant to go into USB3 ‘B’ receptacles …!

So USB Type-C is meant to be the one connector standard to rule them all. It even has rotational symmetry, so it shouldn’t matter which way you plug it in!

So what is a USB device?

Technically speaking, it is a device whose manufacturer has paid for a USB license, sent their device for certification, and passed the USB-IF’s certification requirements, allowing the manufacturer to put the USB logo on the packaging.

Practically speaking, it is any device that has a USB port, allows USB devices to be connected to it, or allows itself to be connected to other USB devices, and basically behaves like a USB device. (If it walks like a duck and talks like a duck …)

Furthermore, it is important to differentiate between a USB host and a USB device. The host acts as the controller of the device, and initiates all communication between the two. For instance, if you attempt to connect two Android phones to each other with a USB cable, one must act as the host and the other as a device, even if both are capable of acting as hosts. The host decides what can be done through the connection.

This helps to explain why a USB-charging battery pack cannot also be an external storage device at the same time, i.e. you cannot combine a phone battery pack with an external hard disk and hope to charge your phone + access the external hard disk at the same time. When your phone charges from the battery pack, it acts as the USB device (in charging mode); when it accesses a hard disk, it is acting as the USB host. It cannot do both simultaneously!

Data supported over USB Type-C cables

In addition to USB3.0 data (and later versions of USB 3), the USB Type-C specification also allows the Type-C connector to carry other kinds of data, if supported by the device:

DisplayPort video data for monitors and computer display devices (Issue 124))
HDMI video data for monitors and consumer electronics devices (also covered in Issue 124)
Thunderbolt 3 data, for high-data-transfer devices

The Type-C specification even supports modes that allow a Type-C cable to carry multiple types of data simultaneously. For instance, a Type-C connector that connects a laptop to a monitor can:

carry DisplayPort video data, enabling the laptop to use the monitor as an external display,
carry USB 3 data, enabling the laptop to use USB ports and other features (such as audio inputs/outputs) on the monitor

… All this comes with the caveat that one must read the manufacturers’ fine print to see if these features are supported on the respective devices. Just because the USB spec allows it, doesn’t mean that a particular device implements it!

For example, some laptops may have two Type-C ports, but only one of those ports will support Thunderbolt and DisplayPort; the other port sometimes only supports USB3 data.

Issue summary: USB is a (licensed) technical standard that describes how devices connect to each other through a cable. USB Type-C is a new connector standard that supports USB 3, DisplayPort, HDMI, and Thunderbolt. It is able to carry multiple types of data simultaneously, in limited combinations. In a USB connection, one device acts as the host while the other acts as the device; the host initiates all communication.

What I’ll be covering next

Next issue: [LMG S10] Issue 127: USB Type-C Power Delivery

This issue was about how data is handled over USB Type-C. Next issue, how power is handled over USB Type-C. After all, every day, millions of devices are getting powered over Type-C: from smartphones, to Internet-of-Things (IoT) devices, to full-size laptops. How is a single type of cable able to cover such a wide range, when earlier cable types could not?

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]

Issue 125: Analog and digital conversion

2021-06-19T08:00:00+08:00

Previously: The VGA video format originated in the time of cathode-ray televisions (CRTs). It was superseded by HDMI, a video format standardised by consumer electronics companies. DisplayPort, on the other hand, is a video format standardised by computer display companies.

The bulk of the story has been written in Issue 123), so this issue will be short.

Why two digital formats? HDMI vs DisplayPort

HDMI is a consumer electronics standard, and is thus heavily focused on broadcast and home video needs. HDMI primarily supports video and audio data. It also carries some control signals through CEC (for Consumer Electronics Control) capability, enabling a video game console or set-top box to send remote-control commands to a television set via the HDMI connection.

DisplayPort is a computer display standard, focused on computing needs. DisplayPort supports video data, optionally audio data, and additional data (such as USB). Since 2014, compatible devices can also transmit DisplayPort signal format over USB-C, provided both the transmitting and receiving devices support it.

Analog vs digital formats: a recap

Digital formats differ from analog formats, because they do not carry the raw signal for the device. Instead, they carry information about the image, encoding the image data (Issue 43)) into video form; after all, video is just a series of moving images! The device takes on the responsibility of figuring out how to make the images appear on-screen, which is why digital TVs require significantly more electronics than CRT TVs.

Analog to digital conversion

An analog signal does not easily convert to a digital signal! Analog-to-digital converters, such as the VGA-HDMI adapters that seem to be needed universally, have to figure out how to process a wave-like (analog) signal, and convert it into the pixel data that constitute an image. These adapters need a digital-analog conversion (DAC) chip to carry out that conversion.

In contrast with analog signals, digital signals usually carry uncompressed video data. Digital-to-digital converters thus do not need to carry out any conversion—it’s the same image! Most of these converters merely need pins to be mapped to each other, which makes them cheaper (e.g. DisplayPort-HDMI converters).

A chip that is able to handle multiple formats and produce a robust output is costly, and it is expected that a good adapter will cost quite a bit. That is no guarantee, however, that a costly adapter will always be a good adapter.

Annddddd … back to graphics cards

The graphics card is in charge of converting the final rasterised signal (Issue 122)) to a video signal, depending on the video format that is required. Naturally, this requires additional chips. Most integrated graphics chips support VGA and HDMI, while DisplayPort support is usually reserved for higher-end devices.

Higher-end graphics cards offer support for more video formats. Furthermore, they also have the capability to rasterise and output video streams for multiple screens, enabling multi-screen support for those who need it. If you find that you need more than two screens for work or play, you are likely going to need a dedicated video card that supports three or more simultaneous video output ports.

Issue summary: Analog formats such as VGA mostly contain the control signals that the CRT needs to operate, while digital formats such as HDMI and DisplayPort contain image data that the device must convert to control signals. Analog signals need a digital-analog-conversion (DAC) chip to be converted to digital signals, hence VGA-HDMI adapters tend to be more costly than DisplayPort-HDMI adapters. Dedicated graphics cards generally support more simultaneous output video streams than integrated graphics cards.

I hope this sufficiently explains a question I hear so often: why do VGA-HDMI adapters cost so much? I’m also glad this issue ended up much shorter than I expected.

In general, if your adapter/cable needs a chip to carry out signal conversion, it’s going to cost more than a plain cable.

What I’ll be covering next

Next issue: [LMG S10] Issue 126: USB Type-C

Next issue, let’s zap some common questions about the latest USB standard! This is going to stretch over two issues. First up: USB-C for data.

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]

Issue 124: Video formats

2021-06-12T08:00:00+08:00

Previously: Graphics cards contain lots of tiny cores that are much better at performing the same calculation for lots of decimal numbers. These cores are organised into compute units; a graphics card with more compute units can perform more calculations every second. Graphics cards have their own onboard memory, separate from the CPU. GPU memory is different from computer memory; it is configured for much higher data throughput. Integrated graphics are GPUs that are integrated into a CPU chip; these do not have their own onboard memory, and share memory with the CPU.

Ah, the esoteric, tricky, complicated art of shooting electromagnetic radiation into the eyes of humans … entire tomes have been written about this. And I will attempt to summarise the pertinent parts into a single newsletter issue. The hubris!

It’s really something when you suddenly remember that television has been around since the 1930s, while computers in some recognisable form were a 1970s invention. The first part of the computer to be invented was the screen!¹

How did screens work if computers weren’t invented? A crash course:

Cathode-ray tube (CRT) televisions

Early colour television screens had primary-colour (red, blue, and green) phosphor dots embedded in the user-facing portion of the screen. These dots emit coloured light when struck by electrons. At the back of the television, cathodes made of barium oxide are heated, causing them to emit electrons.

These electrons, when emitted, fly in all directions, but they are shaped into a beam by an electric field (hence the name “cathode ray tube”, which you might have seen in the form of the acronym CRT).

1: Barium oxide cathode, which is heated to emit electrons (labelled “2”), which are shaped into a a beam by an electric field (labelled “3”)
4: Deflecting coils, discussed in the next paragraph
8: Coloured phosphor dots, arranged on a flat screen (labelled “7”).
Image via Wikipedia

This electron beam could be aimed at any of the phosphor dots by a set of electromagnetic deflecting coils mounted along the sides of the TV, on the inside surface. One set, oriented vertically (mounted left-right), controlled the horizontal deflection of the electron beam, while another set, oriented horizontally (mounted top-bottom), controlled the vertical deflection.

The electron beam produced by the cathode is deflected by (electromagnetic) deflecting coils.
Image via Wikipedia

To produce an image, the electron beam is manipulated to scan across the screen, one line at a time. Each pass across the screen causes it to strike phosphor dots, emitting light in a line. A variety of techniques (microdeflections, masks, and filters) are used to ensure the correct dots are struck.

Each line of the screen is laboriously scanned with this technique, about 60 times a second. This means that the screen “updates” with a refresh rate of 60 Hz.

To make this happen, a varying voltage is applied across the two sets of deflecting coils. The required pattern for the deflecting coils has to come from the television signal source; the television signal from the broadcasting station, therefore, closely resembles the pattern required by the deflecting coils. The television itself applies almost no processing to this signal! (Remember that the chips used to do this kind of processing had not been invented yet.)

These are what we call analog signals. Phonographs and early telephones also used analog signals.

Video Graphics Array (VGA)

So when the computer was first invented, and these screens were widely available, there was no need to reinvent the screen. Graphics cards (Issue 123)) simply had to figure out how to emit analog signals that would work with CRT screens.

The graphics standard for doing so is called VGA (Video Graphics Array), and was first released by IBM in 1986. An organisation, the Video Electronics Standards Association (VESA), was quickly formed in 1989, spearheaded by Nippon Electric Company (NEC), to extend this standard and allow it to support higher resolutions (up to 1080p!).

LCDs replaced CRTs

As CRTs grow larger, they ran into a few problems. CRTs were big, bulky, and heavy. The larger you made them, the longer you had to make the cathode ray tube, which made them immensely heavy!

By this point, LCD technology had been developed. Instead of using a scanning electron beam, it consisted of a backlight² behind a liquid crystal layer (hence the term liquid crystal display, LCD).

The liquid crystal layer consisted of pixels of each primary colour. Each pixel had an adjustable transparency, which depended on the voltage applied across it (high voltage = transparent, low voltage = opaque); a cluster of red, green, and blue pixels formed a single image pixel on screen. By applying different voltages across each primary-colour pixel, we can put an image together.

Digital signals

CRTs controlled the voltage plates directly to deflect an electron beam, through an analog signal. But LCDs use an internal processor to determine what voltage to apply across each liquid crystal pixel. As the technology improved, lower voltages could be used to reduce power usage. So LCDs need a different kind of signal: a digital one, consisting of the raw image data (Issue 43)).

The exodus to digital formats

As digital television became more feasible due to decreasing microprocessor and LCD screen costs, digital formats sprung up to replace VGA.

An early competitor, Digital Visual Interface (DVI), was launched by a working group convened by some computer makers (Intel, Silicon Image, Compaq, Fujitsu, HP, IBM, NEC). It was very quickly superceded by High-Definition Multimedia Interface (HDMI), an interface which implemented standards set by consumer electronics companies (Hitachi, Sanyo, Silicon Image, Sony, Technicolor, Toshiba).

HDMI and mini-HDMI connectors displayed from top to bottom. Both support the HDMI video standard
Image via DataPro

This was followed by the DisplayPort (DP) standard, developed by PC and chip manufacturers and standardised by VESA, and released in in 2006.

DisplayPort, mini-DisplayPort, and Thunderbolt Type-C connectors displayed from left to right. All three support the DisplayPort video standard
Image via DataPro

Issue summary: The VGA video format originated in the time of cathode-ray televisions (CRTs). It was superseded by HDMI, a video format standardised by consumer electronics companies. DisplayPort, on the other hand, is a video format standardised by computer display companies.

Phew. This issue is much longer than I would like; there is so much history to these things! The HDMI-vs-DisplayPort question/complaint I keep hearing is one that only made sense for me in the context of the respective industries they sprung from, and this is something I think most layfolks could definitely understand.

What I’ll be covering next

Next issue: [LMG S10] Issue 125: Analog and digital conversion

Now that I’ve laid out the key differences between VGA, HDMI, and DisplayPort, we can talk about … video adapters.

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]

To be pedantic, the first part of the computer to be invented was actually the keyboard, but they were non-electronic and were called typewriters then. ↩
Older LCDs used cold-cathode fluorescent lamps (CCFLs) for backlights, but today LEDs are used instead. ↩

Issue 123: Graphics cards: The Pixel Factory

2021-06-05T08:00:00+08:00

Previously: Computers are general-purpose machines that usually process integer calculations. The graphics pipeline requires more specialised hardware that can process decimal number calculations. This is why high-performance graphics usually requires a graphics card.

So why are gamers so agog over graphics cards (also known as video cards)? That’s because they do one thing really well! Unlike CPUs which often have to process an unpredictable workload, the graphics pipeline involves performing the same categories of calculations over and over again.

Graphics compute units

These calculations, which I gave an overview of in Issue 122), take in tables of numbers, crunch them mathematically, and spit out another table of numbers. Since the calculations are predictable, we don’t need very complicated hardware that enables switching instructions based on the input. We can used specialised cores—clusters of transistors that are custom-fit for the purpose, cram lots of them into a circuit board, and end up with much better performance for the graphics pipeline compared to the CPU.

These cores are organised into groups known as compute units, and graphics card companies often differentiate the lower-end and higher-end cards by the number of compute units they have, which indicate the computation rate (measured in flops, which stand for floating-point¹ operations per second). Better graphics require more computation, so more flops correlate with better graphics.

Graphics memory

Graphics cards often come with their own memory chips, based on a different technology (GDDR, standing for “graphics double data rate”, vs DDR for CPUs). Graphics memory chips (GDDR) are optimised for higher bandwidth (more gigabytes transferred per second), while CPU memory chips (DDR) are optimised for lower latency (lower time to response). These are soldered directly onto the graphics card to keep memory readily accessible by the compute units.

Power and heat

Unlike CPUs, which (for desktops) seldom draw more than 100W by themselves, graphics cards (for desktops) can draw up to 300W. Correspondingly, more of the space on graphics cards are taken up by components that try to keep this immense power under control. Voltage regulation modules help to adjust input voltages to what the compute units and memory requires.

A graphics card, here exposed without its shroud and cooler.
The graphics chip is in the middle, surrounded by graphics memory chips (smaller, black squares) close to the edges of the board. The larger protruding gray rectangles are voltage regulation modules.
Image from Gamersnexus

Other uses

The graphics card excels at carrying out predictable decimal calculations in a pipeline. Besides graphics, what else can it be used for?

In research, they have been purposed to perform calculations for simulations, which often involve processing the same calculation in bulk.

For consumer purposes, they have also been used for video decoding (decompressing videos for playback), and lately even video encoding (compressing videos to a smaller size). These also involve performing the same types of calculations in a predictable pipeline.

Most recently, they are being used for machine learning (also known as “artificial intelligence”) models, again because those involve predictable pipelines (Do you are see a pattern here?)

Integrated graphics

Not all computers need a full-size GPU to render graphics on screen. Intel processors, and some of the newer AMD ones, contain what is known as integrated graphics. That means that these CPUs have a graphics processor unit (GPU) integrated into the same chip. This integrated GPU provides basic capabilities which allow typical users to use a computer, and even run low-end graphics programs, without needing to buy a higher-end graphics card.

Integrated GPUs do not have their own memory. They share computer memory with the CPU. That means that they reserve a (configurable) amount of computer memory for graphics use (typically up to 128 MB), and video card drivers (Issue 120)) enable the GPU to use more system memory (up to 50% for Intel integrated graphics). This means that integrated graphics use slower memory compared to dedicated graphics cards that have their own memory.

Issue summary: Graphics cards contain lots of tiny cores that are much better at performing the same calculation for lots of decimal numbers. These cores are organised into compute units; a graphics card with more compute units can perform more calculations every second. Graphics cards have their own onboard memory, separate from the CPU. GPU memory is different from computer memory; it is configured for much higher data throughput. Integrated graphics are GPUs that are integrated into a CPU chip; these do not have their own onboard memory, and share memory with the CPU.

What I’ll be covering next

Next issue: [LMG S10] Issue 124: Video formats

I haven’t talked about the last part, because this issue is long enough already, and because the next part can fill a whole issue by itself. Next up, the last stage: actually displaying pixels on a screen.

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]
~~a video card? [Issue 113]~~

“Floating-point” is a fancy term for decimal, so-called because the way they are represented allows the decimal point to be placed differently, unlike with integers. ↩

Issue 122: The great flattening

2021-05-29T08:00:00+08:00

Previously: 3D models are represented with vertices (points), edges (line segments between points), and faces in a computer. Images known as textures can be mapped to faces to give the impression of detail.

Having a model represented in a computer as a large set of numbers is cool, but nobody does 3D modelling like that. We need something to look at! We need a way to convert our model into a flat picture, ideally displayed on our monitors. And this conversion process needs to be fast enough that as we rotate or change the view of our model, the computer can keep up, displaying the changes in real-time.

This process is called rendering.

A pipeline for rendering pictures

During rendering, data from the modelling program is processed in a series of steps; these steps together are known as the graphics pipeline. If you’ve done perspective drawing or tried to figure out how to create clever camera tricks, you already have a sense of what the computer needs to do here.

The 3D cube on screen looks different depending on where our eye is, and which direction it is looking. Our distance to the cube affects how much distortion the view undergoes. Lighting also affects how the cube appears, by making shaded areas appear darker, and lit areas appear brighter.

Finally, this 3D model needs to be “distorted” into a 2D view so it can be displayed on a screen. (We don’t have 3D holo-projectors yet … 😢)

Lastly, since our screens display images as a grid of pixels, we need to figure out the best way to convert the distorted 2D view into a pixel grid. This part of the pipeline is known as rasterisation. Here, the computer figures out what colour each pixel should be, based on which part of the model actually gets projected here. Hidden parts do not need to be rasterised, and neither do parts of the model which are outside the screen.

All these steps take place in the graphics pipeline.

The hardware

The pipeline used to be carried out by the CPU (Issue 53)), but that isn’t ideal. The CPU’s hardware is optimised for general-purpose processing: keeping track of integers (i.e. natural numbers like 1, 2, 3, …), adding or subtracting them, and resetting them. It has many more computational units that carry out this calculation.

Graphics processing, on the other hand, requires a different kind of calculation. The position of vertices do not fit nicely into integers; we have to carry this out using decimal numbers (1.46776, 2.58704, –3.57514, …). The CPU does not encounter these often, and therefore does not have many of these computational units.

For graphics, we need a different kind of processor, one that is jam-packed with decimal computational units. We need a graphics card.

Issue summary: Computers are general-purpose machines that usually process integer calculations. The graphics pipeline requires more specialised hardware that can process decimal number calculations. This is why high-performance graphics usually requires a graphics card.

What I’ll be covering next

Next issue: [LMG S10] Issue 123: Graphics cards: The Pixel Factory

So what does a graphics card do, just pop out pixels like nobody’s business?

Yep! Next week, a quick intro to graphics cards, and why they are so amazing. And, because this is recent news, some coverage on how the M1 differs from most laptops in the way it manages the CPU and GPU (graphics processing unit).

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]
a video card? [Issue 113]

Issue 121: In graphic detail

2021-05-22T08:00:00+08:00

Previously: Driver files provide information about the driver, and instructions on how to receive information from the device, and encode information to be passed to the device. The operating system may come with generic driver files for the device, but custom driver files might provide better performance or additional features.

This issue, let’s start from scratch with graphics: how does a machine that only processes 1s and 0s work with graphics? For starters, let’s think: what can we represent about graphics if we only have numbers?

Representing graphics using numbers

Numbers can be used to represent points (a vertex, in graphics-speak). In 3D space, a point can be represented with 3 numbers, usually written in math as (x, y, z).

A line (edge, in graphics-speak) can be represented as a segment between two points. So we can represent straight-line segments using two points. Curved-line segments are trickier, but for now a lazy way to represent them is just … using a series of points 😛 like connect-the-dots puzzles!

Surfaces are … a bit trickier. What is the minimum number of points we need to represent a flat surface? Turns out the answer is 3 points: if we pick any 3 points that are not along the same line, we can join them along their edges and they form a triangle, which is a flat surface!

In graphics-speak, we call these surfaces faces. Four points which form a flat four-pointed face are called quads, and we can do the same for polygons with more points too. But most graphics hardware just deals with triangles and quads, because everything else can be represented using triangles and quads anyway.

Making 3D models

3D modelling software (also referred to as computer-aided design software, or CAD software) helps us with the process of creating vertices (plural for vertex), edges, and faces. Any object we model digitally is just a collection of vertices, edges, and faces: we call such collections a mesh.

Textures

With just meshes, we quickly run into the limits of what can be represented. For simplicity of calculation, each face can only have one colour. To make really detailed and realistic models, we need very finely detailed meshes. These are problematic because a lot of calculation is needed to make these models appear on screen; the more faces it has, the more calculation is needed!

One way to reduce the number of faces in the mesh while still creating a decent model is to use images on the face. (I explained how images are represented in computers in Issue 43)) We will need additional information to describe the scale and rotation of the image on the face, but at least we can use the same image across multiple faces if necessary. Instead of having to model a hundred thousand blades of grass, I could just model a few stalks, and use a grassy texture to complete the impression.

Scene modelling

Just having models is pretty boring. We will usually be putting multiple models in a scene. Besides models, a scene needs to have lighting, a way to model the sky as the background, and other niceties.

We can use numbers to describe the position, luminosity (i.e. brightness), hue (i.e. colour), and other properties of lighting in our scene.

Issue summary: 3D models are represented with vertices (points), edges (line segments between points), and faces in a computer. Images known as textures can be mapped to faces to give the impression of detail.

Great, we have a way to use numbers to describe a 3D model; that’s nice progress. But how do we turn those numbers into a picture on screen?

What I’ll be covering next

Next issue: [LMG S10] Issue 122: The great flattening

Numbers, numbers, and more numbers … isn’t the CPU great at numbers? Why do gamers and 3D modellers have the hots for graphics cards instead? Stay tuned next week!

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]
a video card? [Issue 113]

Issue 120: Drivers, the glue between hardware and firmware

2021-05-15T08:00:00+08:00

Previously: Solid-state disks are much faster than hard disks because they have no moving parts, so no time is wasted waiting for parts to get into the right position. However, they are more expensive than hard disk drives.

So this issue, we finally get to a question that most folks would have asked at some point after buying a new device:

“What is a (device) driver and why do I need one?”

This issue is going to focus on USB devices, since that is by far the largest category of devices that people use. But driver files are needed for all hardware, including disks, monitors, hardware timers, controller chips, …

USB Devices

The USB specification was first released in 1996, and today it contains 21 categories of devices, each containing even more subcategories.

Each category of device has its own protocol for sending information to the computer, and receiving information from the computer.

I don’t think I am over-exaggerating things to say that there are an uncountable number of USB devices in the world today. For this reason, it makes no sense to complain to the USB Implementers’ Forum (USB-IF), which is responsible for designing the USB specification. Device manufacturers themselves have to be the one responsible for making their device work with our computers.

They do so by providing driver files with their devices. In a not-so-distant past, every device you bought came with drivers on a CD-ROM or DVD-ROM, which you had to install before use. Today, it is more likely that you’ll download these drivers from the manufacturer’s website. If this is a widely used device (e.g. monitors or input devices like keyboards or mice), it might even make its drivers (and updates to those drivers) available through Windows Update.

What is a driver file?

Simply put, a driver file tells the computer:

information about the device (its name, category, and available features)
how to interpret signals coming from the device (through the USB cable),
how to encode signals to be sent to the device so that it can understand them.

The USB-IF maintains a database of vendors and their products. Companies that wish to have their products recognised should get a vendor ID through the USB-IF; this also allows them to use the USB-IF logo (for USB-certified™) on their product packaging if the product passes certification.

When a device is inserted, it passes information including its vendor ID and product ID to the computer. Each driver file also includes the vendor ID and product ID it is meant for use with. This allows the computer to verify that the correct driver is installed for the device.

If you insert a device and Windows says it does not recognise it, you can be sure the problem has something to do with the device file (if the device is otherwise working properly).

On your computer, you can view a list of your devices and inspect their driver details using Device Manager.

Device Manager in Windows 10

Generic drivers

When you first set up your computer, it is not going to have driver files for devices that are already connected. How are you supposed to use your keyboard, mouse, and monitor, among other things?

The major operating systems come with generic driver files pre-packaged. These generic driver files support a class of devices (e.g. input devices, pointing devices, display devices, …) for basic features only. Manufacturers that wish their devices to work upon plugging in (this capability is also known as Plug-and-Play) should aim to support these basic features through the use of generic drivers.

They can then encourage the customer to install custom drivers which may improve the device performance (e.g. for wifi or LAN networking devices), or provide additional features (e.g. storing mouse button configuration settings, or programmable mouse/keyboard buttons).

Issue summary: Driver files provide information about the driver, and instructions on how to receive information from the device, and encode information to be passed to the device. The operating system may come with generic driver files for the device, but custom driver files might provide better performance or additional features.

Driver files were one of those mysterious things that made perfect sense once I took the time to think about why they are needed. I hope this issue does that for you too.

What I’ll be covering next

Next issue: [LMG S10] Issue 121: In graphic detail

One part I’ve always wanted to tackle with this season is explaining the process of getting graphics onto your monitor screen. The explanations I’ve found online are either too vague, trying to paper over the details with metaphors, or far too technical, going into exhaustive detail about the graphics rendering pipeline.

I want to strike more of a middle ground, to help those who have read the former and are trying to bridge the gap to the latter. The next two issues are going to talk about the processes involved in going from model (the computer’s view of things) to graphics (the final rendered display).

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]
~~a driver file and why do I need one? [Issue 98]~~
a video card? [Issue 113]

Issue 119: Solid-state disks, an upgrade from hard disks

2021-05-08T08:00:00+08:00

Previously: A hard disk consists of a read arm, and a set of magnetic platters which store data. To read or write data, the read arm must move to the appropriate track of the rotating platter, and detect the magnetic field (for reading), or attempt to magnetise the domains on the platter (for writing). Operations that require the read arm to access different parts of the magnetic platters intermittently result in slower read speeds.

Hard disks were, for a long time, the best affordable storage technology we had. But then something better came along: the solid state disk¹ (SSD).

What is a solid state disk?

It is, to put it simply, a small circuit board with lots of chips, that plugs into your laptop.

It has multiple storage chips, quite similar to the ones in your thumb drives/flash drives but much faster, and one controller chip to rule em all.

A solid state disk, plugged into a laptop slot (but not secured)

Hmm … like computer memory?

Computer memory also consists of chips on a circuit board, right? But memory gets wiped after the computer loses power … but that doesn’t happen with SSDs; why?

Computer memory sticks, inserted into the memory slot of a computer motherboard

Computer memory uses capacitors, which are like micro-sized batteries. They hold a charge when powered, and store either a 1 or 0 state by being charged or uncharged, respectively.

Solid state storage, on the other hand, uses gated transistors instead of capacitors. They lock a bunch of electrons behind a gate to fill a storage cell (storing a 1), and empty it by forcing the electrons out (strong a 0). This is slower than charging/discharging a capacitor, but hey you don’t lose your data when the power goes out!

Solid state disks have no moving parts

As you can see, there are no read arms or magnetic platters involved. No waiting for a platter to spin up, no waiting for a read arm to move back and forth … access is almost instantaneous².

This is a big deal when reading data from multiple locations; the response is hundreds of times quicker than a hard disk!

What’s the drawback?

The biggest drawback for now is of course price. Solid state disks cost much more, per GB, than a hard disk drive (HDD).

Another potential drawback is that SSDs have a limited lifespan. This lifespan is not measured in months or years, but in the amount of data written.

You see, each time electrons are forced through the gate (called a program-erase cycle), the gate gets weaker. Do it tens of thousands of times, and eventually the gate gets too weak to hold electrons. A budget SSD typically has a lifespan of 30,000 program-erase cycles. For a 256GB SSD, that’s 7.68 million GBs of data-writing before it fails!

This was a big concern for early SSDs, which had program-erase cycles in the low thousands. Today, with technology improving the program-erase cycle lifespan of SSDs, most users would not come close to reaching the end-of-life of an SSD.

Issue summary: Solid-state disks are much faster than hard disks because they have no moving parts, so no time is wasted waiting for parts to get into the right position. However, they are more expensive than hard disk drives.

I managed to write this issue without once mentioning write amplification or NAND, and I consider that a significant personal achievement.

What I’ll be covering next

Next issue: [LMG S10] Issue 120: Drivers, the glue between hardware and firmware

So … hard disk drives and solid state disks work pretty differently under the hood, yet when you plug them into a computer they just … work? How does that happen?

For that matter, how do devices just work when we plug them into a computer?

This next week, when I finally talk about … drivers!

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]
a driver file and why do I need one? [Issue 98]
a video card? [Issue 113]

The name has to do with the fact that its working principles are based on solid state physics, and not on the solidity of the disk itself. ↩
Feels almost instantaneous to humans … but this still takes a few microseconds, which is considered slow for a computer! ↩

Issue 118: When I run two file-copy processes at the same time, why are they much slower?

2021-05-01T08:00:00+08:00

Previously: Operating systems use a page file on the storage disk as a complement to physical memory. This allows OSes to behave more performantly than they would if they did not have a page file. Data that is rarely accessed is moved to the pagefile (“paged out”), and can be paged in when it is needed later, albeit with a performance hit.

It’s kind of funny the moment you realise how much of what an operating system (OS) does is try to mitigate the slowness of hard disks. But why are they so slow? There’s an actual explanation for that, and along the way it will help us understand a few things about why OSes sometimes behave funny.

What is a hard disk?

A hard disk is a magnetic platter that stores data. This platter contains trillions of magnetic atoms; yep, they’re atom-sized magnets! These atoms are grouped into clusters called magnetic domains; they are so tiny that 500 of these placed in a straight line would stretch the diameter of a human hair! These domains can align themselves in one of two different ways: let’s call them “up” and “down”. Each bit (as in, 8 bits = 1 byte) is stored as the alignment of a magnetic domain: up represents 1, and down represents 0.

To read data, all you need to do is move a tiny electromagnet over each domain, and use it to see which way the domain is aligned. This can be done by measuring the current flowing through the electromagnet, a detailed explanation of which is beyond the scope of this newsletter (perhaps in a future newsletter titled “Layman’s Guide to Physics” or something).

To write data, you pass a current through the electromagnet to magnetise the domain below it whichever way you want; just spin the platter and keep changing the current to write a series of 1s or 0s.

Put 3-5 platters together, attach the electromagnet to a moving arm (called the read arm), control the whole thing with some microchips, and you have a hard disk.

A picture of an opened hard disk, showing the read arms and magnetic platters

Characteristics of a hard disk

A hard disk spins at a constant speed, because it is much more complicated to figure out how to read/write stuff when the speed can vary. The larger hard disks, which go into desktops and use 3.5-inch platters, spin at 7200rpm, while the smaller hard disks, which go into laptops and use 2.5-inch platters, spin at 5400rpm.

Data stored near the circumference of a magnetic platter is faster to read than data stored near its centre. For this reason, OSes that are installed on hard disks usually attempt to partition the storage space so that the operating system is stored on domains closer to the circumference.

The read arm moves really close to the platters during disk operation! The gap between them (called the head gap or flying height) is half the thickness of a human hair. This is why you do not want to drop hard disks while they are in operation; the slightest movement of the read arm towards the magnetic platter causes it to gouge the platter surface and damage it permanently: this is called a head crash.

Read and write operations

The hard disk is ultimately a mechanical device; each operation involves moving parts.

Reading from or writing to a domain involves:

Spinning up the platter (if it isn’t already spinning)
Moving the read arm to the correct position
Measuring or inducing a current

This means that each time the hard disk needs to access data from a different region of the disk, there is significant lag time (~5ms; see Issue 57)). This is the time needed for all those movement described above. It is thus advantageous to try to put all the data you need in contiguous domains¹, to minimise read arm movement.

So when an OS tries to perform two data operations at the same time (instead of sequentially), the read arm has to move a lot more to access data from different regions. And this is why, if you have multiple data operations to perform on a hard disk, you should try to do them sequentially instead of simultaneously!

Issue summary: A hard disk consists of a read arm, and a set of magnetic platters which store data. To read or write data, the read arm must move to the appropriate track of the rotating platter, and detect the magnetic field (for reading), or attempt to magnetise the domains on the platter (for writing). Operations that require the read arm to access different parts of the magnetic platters intermittently result in slower read speeds.

This was something that took me a while to figure out; first, to notice that it was actually happening, and second, to read up enough about hard drives and take some (nonworking) ones apart to understand. And now I can explain it to you :)

What I’ll be covering next

Next issue: [LMG S10] Issue 119: Solid-state disks, an upgrade from hard disks

Next, we look at the technology that has been steadily replacing hard disks as system disks in most desktops and laptops. These have managed to eliminate the latency due to moving parts, and enable much higher read and write speeds.

Sometime in the future: What is:

XSS? [Issue 8]
a good reason developers write code and give it away for free online? [Issue 21]
OpenType? And what are fonts anyway? [Issue 42]
a driver file and why do I need one? [Issue 98]
a video card? [Issue 113]

This process is what millennials might remember as defragmentation, or defragging. ↩