The clock is wrong, the memory profile is off again, and nothing you set last night survived the power switch. What the coin cell holds, and what it does not
This page is for a PC that starts normally and then forgets: right time yesterday, wrong time this morning; the memory profile off again; a prompt at startup telling you the BIOS settings have been reset. ASUS states what that memory holds in one sentence — "Clear CMOS clears the Real Time Clock (RTC) random access memory in CMOS, which contains the date, time, BIOS password and BIOS setup parameters" — and a board losing exactly that list every cold start is losing the thing the coin cell is there to keep alive.
If your PC does not start at all — no POST, a debug LED stuck on CPU, DRAM, VGA or BOOT, a blank screen — that is a different symptom with a different page. Go to motherboard not POSTing: what the debug LEDs mean instead. Nothing on this page is a diagnosis for a machine that never gets as far as forgetting anything.
There is a flat silver battery on your motherboard, about the size of a five-pence piece or a nickel. On almost every desktop board it is a CR2032. Energizer's datasheet for the cell gives its designation as "ANSI / NEDA-5004LC, IEC-CR2032", its chemical system as "Lithium / Manganese Dioxide (Li/MnO2)" and its nominal voltage as 3.0 volts.
What it is for is narrower than most people assume, and ASUS names the scope precisely in its support article on clearing CMOS. The thing being kept alive is "the Real Time Clock (RTC) random access memory in CMOS", and its contents are "the date, time, BIOS password and BIOS setup parameters".
That is the whole list. A clock, and the settings you changed in firmware. It is not a backup battery for the system, it does not hold the BIOS itself — the firmware lives in flash memory on a separate chip and survives with no power at all — and it has nothing to do with your files, your Windows install or anything on your drives.
The reason a battery is involved is that this particular memory is volatile: it needs a trickle of power to keep its contents, and when the machine is unplugged there is nowhere else for that trickle to come from. That is why the failure shows up as forgetting rather than as breaking.
Work backwards from what ASUS says is in there. If the coin cell can no longer hold that memory, the things you lose are the things on that list, and they go together.
The pattern that matters is all three moving together, and moving with mains power rather than with restarts. One setting that will not stick while the clock keeps perfect time is not this.
Boards word it differently, and the phrase readers most often search for is a checksum error. We have not sourced the literal message strings that any particular vendor prints, so this page does not put words in your board's mouth — what ASUS documents is that a prompt appears, not what it says. Read what is on your screen.
This is the half of the subject that gets people ordering a battery they did not need.
The machine does not POST at all. No display, no progress, possibly a debug LED stuck on one of CPU, DRAM, VGA or BOOT. What the vendors document about the coin cell is what it holds — a clock and a set of firmware settings. Neither Intel nor ASUS describes a flat cell as something that stops a board POSTing, and we are not going to assert it in either direction, because boards differ and we have no source that generalises it. What we can say is that this is a different symptom with its own diagnostic path: the debug LED guide works through what each light means.
It was fine until you enabled EXPO or XMP, and now boots take forever. On AM5 in particular, a long boot with the DRAM light on after a memory-profile change is usually memory training, not a battery. That has its own page: why your AM5 board takes a minute to boot after enabling EXPO.
Windows itself is slow, unstable, or throwing errors. Nothing in the list ASUS gives is a Windows component. A coin cell that cannot hold the date is a coin cell that cannot hold the date.
The clock is a few minutes out over weeks. That is ordinary timekeeping drift, and an operating system that syncs time over the network will correct it. The battery symptom is a clock that stops, not one that runs slightly fast.
Before you conclude the cell is flat, rule out the possibility that something cleared CMOS deliberately. ASUS's article lists three ways to do it, and every one of them produces the same symptom you are troubleshooting.
Intel documents a fourth route on boards that have it: a jumper rather than a plain pin pair. "In general, the CMOS jumper is three pins located near the battery", and it "has positions 1–2 and 2–3" — so a jumper parked on the wrong pair has the same standing effect as one bridging the CLRTC pins.
ASUS also names the situation in which you would have done this to yourself on purpose and forgotten: "When CPU/memory are overclocked or changed the frequency/voltage settings, Clear CMOS can be performed to restore the BIOS settings if unstable/crash/freeze problems appear." A board that recovers itself from a failed overclock is doing the same thing, and the settings it discards are yours.
The other routine settings-wipe is a BIOS update, which resets what you had configured as a matter of course. We cover exactly what goes and what to put back in the BIOS update guide, so there is no reason to repeat it here.
Intel publishes a procedure for this, and it starts before you touch the board. Two steps that readers skip, in Intel's words: "Turn off all peripheral devices connected to the computer" and "Disconnect the power cord from the AC power source." People do try this with the PSU live. A desktop supply keeps a standby rail energised whenever it is plugged in, so "I shut Windows down" is not the same thing as "there is no power in there".
Then, per Intel, "Find the battery on the board." There are three physical arrangements, and knowing which you have before you start saves a lot of prodding. Intel: "The battery may be in a horizontal or vertical battery holder, or connected to an onboard header with a wire."
Orientation matters and is easy to lose track of once the cell is out. Intel's step is explicit: "If the battery is in a holder, note the orientation of the + and – on the battery." The marked face on a coin cell is the positive one, and on a horizontal holder it faces up.
Intel's procedure includes a wait, and it is there because the memory does not empty the instant power is removed: "Wait 1–5 minutes, then reconnect the battery." Putting the same cell straight back in is how people conclude that removing the battery "did not work". If you are fitting a new cell rather than clearing anything, the wait is not doing anything for you.
One consequence worth knowing before you clear a modern AMD board: wiping CMOS discards the stored memory training, so the next boot is a long one again. That is normal, and it is covered here rather than a sign anything went wrong.
Whatever cleared the memory, deliberate or not, you come back to a machine on defaults. ASUS gives the recovery sequence: at the prompt, "press F1 to enter the BIOS setup, press F5 to [Load Optimized Defaults]", and "then press F10 to save and leave BIOS setup".
Loading optimised defaults and saving is worth doing rather than just pressing on: it puts the firmware into a known state instead of whatever it landed in. Then work back through what you had changed.
One item on that list is worth calling out because it can stop Windows 11 booting rather than merely annoying you: a settings reset can drop the Secure Boot configuration. If the machine will not boot into Windows after a clear, the TPM and Secure Boot guide covers putting it back, including why CSM has to be off first.
There is no published figure for how long a CR2032 lasts in a motherboard, and anyone quoting you one to the month is estimating. What the manufacturer does publish is the cell's own specification, and it is enough to calibrate expectations without inventing a number.
From Energizer's CR2032 datasheet:
The self-discharge figure is the useful one. A cell sitting with nothing attached loses roughly one per cent of its capacity a year. That is a component designed to be measured in years, not months — and a board that starts forgetting its settings within a few months of a fresh cell is telling you something other than "the battery is old". Look at the deliberate causes above first, particularly a jumper or cap left bridging the clear pins.
Two things push real-world life the other way, and both follow from the same datasheet. The cell is only doing work while the board is unpowered, so a desktop that stays plugged in at the wall draws on it far less than one that is switched off at the socket every night. And the rated capacity is quoted at 21°C; the stated operating range runs to -30C at the cold end, which is why a machine that has been in an unheated garage over winter is a more plausible candidate than its age alone suggests.
Every quoted sentence on this page comes from the three documents linked below, read on 15 September 2026. Nothing here is a measurement. No battery was removed, fitted or timed for this page, no voltage was read off a meter, and no board was opened. The 3.0 volt figure is the datasheet's nominal voltage, not something we put a multimeter across — if you do test a suspect cell, you are comparing against a published nominal, not against a reading of ours. We make no claim about how long any particular cell will last in any particular board, and no claim that a flat cell does or does not prevent a board POSTing, because none of these three sources says so.
www.asus.com/support/faq/1040820/ form returned HTTP 502 and the /us/ variant timed out when we checked, so we cite the mirror that served the textPublished 15 September 2026. All three documents above were read on 15 September 2026.