Eliminating "CMOS" battery

[pitz]

Seems to me that the "CMOS" battery for real-time clock and for storage of settings is an unnecessary part of the BOM on computers.

After all, it would be relatively trivial to write UEFI code to retrieve time from the Internet from various public clock servers. Or even use-configurable ones. And "CMOS" settings could be stored in persistent "flash" ROM instead of in NVRAM.

So why does every "x86" PC built these days still include a CR2032 cell and the RTC circuitry? With system longevity growing all the time, there's bound to be a lot of unnecessary service calls relating to this fundamental vulnerability in the platform. Especially for embedded platforms.

I know the RTC has been a fixture of the PC world since the time of the original IBM PC AT (the PC XT didn't have a RTC, as I remember being forced to enter the time/date every time I booted one up in the mid 1980s!!). But is the battery backed RTC a concept that should have been buried years ago? I think so personally. Or at least engineered to be optional, so that a missing/dead battery doesn't generate a failure/service call. Nor condemn the user to a ridiculous "zero date".

 

[JeffMD]

So.. grats on thinking about step 1838, and completely forgetting step 1. OS install is usually offline. Also UNTIL that point that the OS can connect to the net for the current time, how does it handle marking file last access dates?

 

[mikeymikec]

And if the computer doesn't have a permanent Internet connection?

Why make the BIOS more complicated (for example it would require a DHCP configuration option), which in turn makes it more vulnerable to attack as well as inviting more flaws in its design.

The RTC allows the computer's clock time to be maintained even if the CPU is fully occupied with another task(s). Why on earth would you want it otherwise? Guaranteeing CPU time within real-time constraints isn't something most computers/operating systems are designed to do.

It seems to me like you're suggesting removing an extremely reliable and basic feature and replacing it with a much more complicated alternative with zero benefit. So it may become necessary to replace a battery every 6 or so years and that's it? I think many people would kill to have the rest of their computer to exhibit that level of reliability.

 

[nerp]

A CMOS battery costs $2. It takes two seconds to replace. They last for 5 to 10 years. I think I'm OK with sticking with CMOS batteries.

 

[pitz]

So.. grats on thinking about step 1838, and completely forgetting step 1. OS install is usually offline. Also UNTIL that point that the OS can connect to the net for the current time, how does it handle marking file last access dates?

The UEFI can include a module to connect to a ntp server. Not a big deal to implement.

 

[pitz]

So it may become necessary to replace a battery every 6 or so years and that's it? I think many people would kill to have the rest of their computer to exhibit that level of reliability.

The every 6 years (which actually becomes 3-4 years in some cases) is a problem on machines which are in datacenters or even office PCs. Computers are lasting a lot longer these days, and a $2 battery can easily cost $100-$200 for a service call + lost productivity.

I have no problem with the batteries being offered, but failure of the battery needs to be a *lot* more graceful rather than greeting the user to a machine which won't even boot up and has no time.

 

[DigDog]

A CMOS battery costs $2. It takes two seconds to replace. They last for 5 to 10 years. I think I'm OK with sticking with CMOS batteries.

i'm so lazy i just stole his quote.

also, although i like the way you think, OP, can you suggest a reason WHY we should eliminate the cmos battery? Does using is have any downsides?

 

[pitz]

A CMOS battery costs $2. It takes two seconds to replace. They last for 5 to 10 years. I think I'm OK with sticking with CMOS batteries.

i'm so lazy i just stole his quote.

also, although i like the way you think, OP, can you suggest a reason WHY we should eliminate the cmos battery? Does using is have any downsides?

Reduced TCO and service costs, especially when the battery inevitably fails. I don't mind if the computers continue to ship with battery-backed RTCs because the BOM cost can't possibly be more than a couple bucks. I just would like to see the failure modes be far more graceful and not require intervention in an otherwise managed IT environment.

 

[DigDog]

you are talking servers maintenance now - it's not the same thing. (i don't even think they are called the same).

having worked in the call center for IBM, i know batteries (cache, raid, etc) fail often enough that eliminating them might actually be a good idea .. but on consumer pcs? not worth it. a cmos battery on desktop runs longer than the practical life expectancy of the motherboard it is attached to.

 

[pitz]

you are talking servers maintenance now - it's not the same thing. (i don't even think they are called the same).

I am talking about both server and client PCs. Client PCs, especially, as they tend to be more geographically dispersed.

having worked in the call center for IBM, i know batteries (cache, raid, etc) fail often enough that eliminating them might actually be a good idea .. but on consumer pcs? not worth it. a cmos battery on desktop runs longer than the practical life expectancy of the motherboard it is attached to.

Not really, especially in this era of people getting 5-10 years out of their desktops routinely. Just because the 'average' battery lifespan is 7-8 years doesn't mean that a significant chunk of the battery population won't fail after 4-5 years.

 

[DigDog]

idk man, maybe you have more experience in this field, but out of ... a dozen computers which i have used, counting those owned by me and by my friends, not a single battery has failed. The only occasion in which i have seen a failed cmos battery was on a old-as-dirt work pc in a restaurant, and this is due to them having a contract with some scammer to come and fix it for a ton of money - so, literally someone using old cmos batteries in order to get more money from the stupids.

 

[VeryCharBroiled]

I cant even count how many computers Ive setup completely off line with no wifi signal or network jack anywhere.

later they may go online but keeping the battery RTC is no brainer. add what the OP wanted but keep the RTC.

guess some people think the internet grows on trees.

 

[Sheep221]

RTC/CMOS battery is crucial in every computer. And its RTC crystal is very precise clock.
Computers were designed with universal use in mind, that also counts being deployed in environments out of reach of telecommunication infrastructure. I don't think this will ever go online.

 

[Deders]

It's not just the time that it keeps, it also keeps all you settings and overclocks. Iagine having to re-do them every time you booted up.

 

[PhIlLy ChEeSe]

Seems to me that the "CMOS" battery for real-time clock and for storage of settings is an unnecessary part of the BOM on computers.

After all, it would be relatively trivial to write UEFI code to retrieve time from the Internet from various public clock servers. Or even use-configurable ones. And "CMOS" settings could be stored in persistent "flash" ROM instead of in NVRAM.

So why does every "x86" PC built these days still include a CR2032 cell and the RTC circuitry? With system longevity growing all the time, there's bound to be a lot of unnecessary service calls relating to this fundamental vulnerability in the platform. Especially for embedded platforms.

I know the RTC has been a fixture of the PC world since the time of the original IBM PC AT (the PC XT didn't have a RTC, as I remember being forced to enter the time/date every time I booted one up in the mid 1980s!!). But is the battery backed RTC a concept that should have been buried years ago? I think so personally. Or at least engineered to be optional, so that a missing/dead battery doesn't generate a failure/service call. Nor condemn the user to a ridiculous "zero date".

You think Asrock wouldn't love to save a few cents on a watch battery, look at cars they used to have full spares n factory rims. Now its temp tire n the rim is made of cheap metal your old car was reduced to.
What if the power fails before these Flash ROM chips copy what you need saved so the board can boot, why you think there is UPS units in a server room, why you think there are deiseal motors backing up the power for theses servers and UPS units? In a prefect world I'm sure yer theary would work, we have bad weather here cause power outages spikes, etc.
Why don't you solve the fuel crisis first, baby steps first!

 

[StrangerGuy]

I cant even count how many computers Ive setup completely off line with no wifi signal or network jack anywhere.

later they may go online but keeping the battery RTC is no brainer. add what the OP wanted but keep the RTC.

guess some people think the internet grows on trees.

Gotta love "solutions" that are far worse than the problems that they claimed to solve.

 

[mikeymikec]

Also, the OP complains about having to visit a data centre once every six years (I strongly suspect that's not the only reason that most people have to visit them, as my past employment has included visiting data centres!), I'd say if a computer is reaching that kind of age then an excuse to give the insides a quick once-over once every say 5 years isn't that bad an idea: I'd much rather find out about say a bulging capacitor, dust build-up or some other physical factor before functionality is affected.

It can also serve as a reminder that one has ageing equipment and that it may be time to review when it is due to be replaced.

 

[WackyWRZ]

I think that most bigger companies these days are still using a 3-4 year replacement cycle on desktop equipment that coincides with warranty expiration. Also on server/data center equipment many places that are keeping their equipment longer are also renewing the service/support agreements on them so if the battery (among other more likely failure items items) fails they just phone a tech and they are out NBD or sooner.

That said I don't think I've ever replaced a CMOS battery on any computer I've ever had - or at work. I have replaced RAID batteries but they were in servers that were a good 10 years old...

You think Asrock wouldn't love to save a few cents on a watch battery, look at cars they used to have full spares n factory rims. Now its temp tire n the rim is made of cheap metal your old car was reduced to.

Heck now days they are throwing a can of fix-a-flat in the trunk and a cheap 12v compressor instead of even a spacesaver tire.

 

[pitz]

It's not just the time that it keeps, it also keeps all you settings and overclocks. Iagine having to re-do them every time you booted up.

Those could be made persistent in Flash, for example. Over-rideable back to defaults with jumpers.

I have no issues with keeping the RTC, but why not make the battery optional? So those of us who don't want a RTC don't have to suffer having to perform maintenance on it?

 

[Red Squirrel]

I'm all for keeping the Basic in BIOS. I hate UEFI for that reason, you basically have to load an entire operating system before you can load your actual operating system. It's extra complexity and boot time for nothing.

I can only recall one time where I had to replace a RTC battery, it's incredible how long they last, usually more than the life of the system. What they could do though is use a small lithium ion cell and have it so it charges when the system is on. But I don't even think that's really necessary.

 

[BonzaiDuck]

Data centers and IT departments, if organized and run properly, should have a regular routine maintenance regimen, like changing the oil in a fleet of Hummers.

You can pretty much tell when the battery is running down before serious problems emerge: the clock speeds up. So among the many IT tasks, somebody takes a clipboard around, carrying either an accurate watch (with a battery) or a cell-phone (which has a battery), sets the time on each computer, and returns in a second pass a few days or a week later, to see if the RTC is still keeping accurate time. Clutching the clipboard and the information it contains, he then makes a third pass with a box full of Duracell CR2032 batteries, replacing those which need it.

As to the mainstreamer, if they can learn to manage their bank account with a cellphone, strong reading glasses and their thumb, then they can learn these simple maintenance tasks.

 

[Red Squirrel]

If organized and run properly, they'll use NTP. If a CMOS battery dies it will usually indicate this at bootup, but otherwise it will not impact every day performance of the system.

 

[pitz]

Data centers and IT departments, if organized and run properly, should have a regular routine maintenance regimen, like changing the oil in a fleet of Hummers.

I was actually thinking along the lines of a manufacturing plant which may have a few hundred PC's scattered about for control, monitoring, surveillance, inspection, etc. PC's that, once installed, should run for many years without maintenance.

If there's a power loss, 6 years down the road, and some of the machines, with depleted "CMOS" batteries don't fire up properly -- that's a problem.

Even consumer machines, it can be a problem. I can pull an old IBM PC XT out of the closet, connect power, and have it up and running (albeit with no RTC, but the time can be entered manually or acquired from a network source!). But with the backup battery now being "critical" for bootup on more modern machines, they ain't going anywhere without the battery replaced.

 

[pitz]

If organized and run properly, they'll use NTP. If a CMOS battery dies it will usually indicate this at bootup, but otherwise it will not impact every day performance of the system.

If the battery's not completely dead, it will indicate. If the battery is completely dead, bye-bye settings, and many boards won't even boot.

This is the behavior I'd like to see changed, so that the UEFI firmware (or BIOS), through an option ROM, can contact a NTP server and set the date. And that settings are saved to flash, not battery backed NVRAM.

 

[Rakehellion]

Even better idea. Have the CMOS run from a supercapacitor that's recharged every time power is applied to the motherboard. The Dreamcast did something similar.