Closing a laptop lid looks like a single action, yet Windows can respond to it in at least four separate ways, and each one asks the hardware to do something completely different. Sleep, hibernation and hybrid sleep are often treated as three flavors of the same idle state, but on the chip and circuit level they touch different components, draw different amounts of current and expose the machine to different risks if the power cord gets pulled at the wrong moment. Understanding how they diverge below the interface explains why a laptop sometimes wakes instantly, sometimes takes fifteen seconds, and sometimes loses an open document entirely.
The ACPI Specification Defines Six System Power States
Every modern PC follows the Advanced Configuration and Power Interface, a specification jointly developed by hardware and operating system vendors so that the firmware, chipset and OS agree on what "off" and "on" actually mean. ACPI splits the life of a machine into global states labeled G0 through G3, and within the working group G1 it further defines sleep states S0 through S5. S0 is full operation, with the CPU executing instructions and every bus active. S5 is a soft shutdown, where the motherboard still receives standby voltage from the power supply but no context is preserved anywhere. Between those two extremes sit S1 through S4, each one shutting off progressively more silicon.
S1 is rarely used today; it only halts the processor while every other chip, including the cache, keeps its state. S2 additionally drops CPU and cache context but is likewise obsolete on consumer hardware. S3 and S4 are the states that actually matter for Windows users, because Sleep maps to S3 and Hibernate maps to S4. Hybrid sleep is not a separate ACPI number at all; it is a Windows-level policy that combines the S3 transition with an S4-style disk write, which is exactly why its behavior sits between the two.
Sleep Mode Keeps Memory Alive While Almost Everything Else Powers Down
When a system enters S3, the chipset cuts power to the processor cores, the discrete graphics card, most of the PCIe lanes and the majority of onboard controllers. What remains energized is the DRAM and a small support circuit around it, along with the real-time clock and enough logic in the platform controller hub to detect a wake signal from the power button, a lid switch or certain USB devices. The memory chips are switched from normal read and write operation into self-refresh mode, a state built into DDR memory controllers in which the DRAM periodically recharges its own capacitor cells using an internal timer instead of commands from the CPU. This is the same physical mechanism that keeps a phone's RAM alive when the screen is off, just implemented with desktop or laptop grade modules.
Because DRAM cells are essentially tiny capacitors that leak charge, they need this refresh cycle roughly every 64 milliseconds regardless of whether the system is awake or asleep; the only change during S3 is that refresh becomes the memory's sole job, and everything that would normally read or write to it is switched off. The current draw at this point is small, typically low single digit watts on a modern laptop, which is why a machine left in sleep overnight still shows meaningful battery the next morning but is not entirely free of drain. If AC or battery power disappears completely while the system sits in S3, the self-refresh circuit loses its supply, the capacitors discharge within milliseconds, and every open document, browser tab and unsaved file is gone the moment power returns, because the hardware never wrote that data anywhere durable.
Hibernation Writes the Entire Memory Image to the Disk Before Shutdown
Hibernation, or S4, takes a fundamentally different approach: instead of keeping RAM powered, Windows copies its entire contents to a file on the system drive called hiberfil.sys before cutting power to the memory modules altogether. The kernel captures the full working set, including running processes, open handles and driver state, compresses it, and streams it to storage, after which the platform can drop to the same power draw as a full shutdown. On an SSD-equipped machine with 16 or 32 gigabytes of installed memory, this write typically takes several seconds to a couple of minutes depending on how much of that memory is actually in use, since Windows only needs to save pages that hold real data rather than the entire physical address range.
The size of hiberfil.sys is not fixed; Windows calculates it from the amount of physical memory installed and from the compression ratio achieved on the current workload, and administrators can adjust the reserved percentage through the powercfg command if they need more free space on a small system drive. On a machine with fast NVMe storage the write speed easily exceeds two gigabytes per second, so even a heavily loaded sixty-four gigabyte configuration rarely needs more than a few tens of seconds to flush, while an older SATA SSD or a spinning platter drive stretches that same operation out considerably and is part of why hibernation feels sluggish on aging hardware. Because nothing needs continuous power once the file is on disk, a hibernating laptop draws essentially zero current beyond what the motherboard uses in a normal soft-off state. Resuming means reversing the process: firmware performs a partial boot, the Windows loader recognizes the hibernation signature, reads hiberfil.sys back into RAM and hands control back to the kernel at the exact instruction where it left off. This restore is slower than waking from S3 because it involves spinning up storage, moving gigabytes of data across the memory bus and reinitializing hardware that was fully powered down, but it survives a dead battery, a pulled plug or a swapped power supply without losing a single open window.
Hybrid Sleep Combines Both Approaches to Protect Against Power Loss
Hybrid sleep, enabled by default on many desktop installations of Windows, tries to capture the benefits of both states at once. When triggered, the operating system performs the same memory-to-disk write used in hibernation, creating a full hiberfil.sys snapshot, but instead of then cutting power to the modules it leaves the platform in the ordinary S3 configuration with RAM in self-refresh. The practical effect is a system that wakes as quickly as a normal sleep because the CPU can usually resume straight from memory, while still holding an up-to-date backup on disk in case something goes wrong.
This design becomes useful in a specific set of situations, which is why it is worth listing the cases where the extra disk write pays off:
- A desktop loses power because of an outage or an accidentally unplugged cable, so RAM discharges and the session would normally be lost;
- A laptop battery drains completely while sitting in sleep for an unexpectedly long time, past the point the charge could sustain self-refresh;
- The operating system or a driver crashes during the sleep transition itself, leaving RAM in an unreliable state that the CPU should not resume from directly.
In every one of these scenarios, the firmware detects that RAM content can no longer be trusted, so instead of a failed resume it falls back to reading hiberfil.sys from disk, exactly as a full hibernation would. The cost is the same brief disk write that hibernation requires each time the system goes idle, plus the modest amount of storage reserved for the hibernation file, which Windows sizes relative to installed memory.
Modern Standby Replaced the Legacy S3 Model on Many Recent Laptops
Most laptops built in the last several years, particularly those with recent Intel or AMD mobile processors, do not implement the classical S3 state at all. Instead they use Modern Standby, also called S0 Low Power Idle or referred to at the silicon level as S0ix. Rather than making a hard jump from full operation to a single low power state, Modern Standby keeps the platform logically inside S0 and steps individual components down through a sequence of runtime idle states, with CPU cores dropping through C-states and peripherals dropping through D-states until the system reaches what Intel calls its deepest runtime idle platform state.
The advantage of this approach is speed: because there is no ACPI state transition and no firmware handshake to reverse, a Modern Standby machine can resume in a fraction of a second, and it can briefly wake components in the background to check email, sync a messaging app or receive a push notification without the user noticing. The tradeoff is that reaching the true power floor depends on every driver in the system cooperating, and a single component that refuses to idle, commonly a network adapter, a fingerprint reader or a poorly written peripheral driver, can keep the platform pulling several watts instead of the fraction of a watt that a genuine S3 sleep would draw. This is why some machines get noticeably warm or lose a large share of battery overnight even though the screen shows the same padlock icon as a properly sleeping laptop.
Windows tries to compensate for this weakness with a scheduled fallback: if a Modern Standby device sits idle long enough without user interaction, the operating system automatically escalates from the low power idle state into full hibernation, writing memory to disk and dropping to near zero consumption even though the hardware never exposes a classic S3 option. This behavior means a laptop that appears to stay in the same sleep state for hours may actually pass through two different underlying mechanisms during that time, starting in a connected low power mode and later settling into the same disk backed state that S4 hibernation relies on, without any visible change on the lid or the power button.
Wake Latency and Power Draw Differ Sharply Between the Modes
The practical gap between these states shows up most clearly in two numbers: how long it takes to get back to a usable desktop and how much charge disappears while nothing is happening. A machine in classic S3 sleep typically resumes in one to two seconds because the CPU only needs to reinitialize peripherals that were fully powered off, while memory content was never touched. Hibernation resume commonly falls in the range of five to twenty seconds on modern SSDs, since the bottleneck becomes reading several gigabytes back from storage and rebuilding hardware state from a cold start. Hybrid sleep resumes at roughly the same speed as ordinary sleep in the common case, because the disk image is only read when RAM has actually failed to hold its content.
On the power side, a laptop in genuine S3 self-refresh draws a small trickle, often under two watts, largely accounted for by the DRAM refresh current and the standby circuitry watching for a wake event. A hibernating machine draws close to nothing beyond baseline motherboard standby power, which is why it is the preferred choice for shipping a device or storing it for weeks. Modern Standby sits in between in the best case and can exceed ordinary sleep power draw by a wide margin in the worst case, which explains why many enthusiasts and IT departments still search for a way to force the older S3 behavior on hardware that technically no longer advertises it in firmware.
Choosing Between the Modes Depends on the Hardware Installed
None of these states is universally correct, and the right choice usually comes down to what the installed components can actually sustain. A desktop with a reliable power supply and no chance of an unplanned outage gains little from the extra disk write that hybrid sleep performs on every idle cycle, so plain sleep or, on systems where it is exposed, disabling hybrid sleep in favor of straightforward S3 can reduce SSD writes without adding real risk. A laptop that regularly travels with a partially charged battery benefits from hibernation whenever it will sit unused for more than a few hours, since it is the only mode immune to a battery that reaches zero before the next charge. Systems built around Modern Standby leave the user with less direct control, because the hardware and firmware decide how deep the platform idles, but checking a device's sleep report through the built in diagnostic tool can reveal which driver or peripheral is preventing it from reaching its lowest power floor.
Ultimately the difference between these modes is not a setting Windows invented on top of otherwise identical hardware behavior; it reflects genuinely different physical states of the memory, the processor and the platform controller hub, each with its own tradeoff between how fast the machine returns to work and how much it can lose if something interrupts it along the way.