In message , Jim GM4DHJ ... writes
Although I thought I was slow and clumsy, it remembered the time that I set up before taking the power off.
The machine is 2007 vintage and the old battery reads 0.83 volts.
They used to be but I'm talking 80286 or 80386 machines...
Not had to replace a backup battery for years and that includes this roughly 15 year old 1 GHz single core Athlon...
*Any* CR2032 cell will do, even pound shop ones. The main board manufacturers switched from those troublesome 3 cell nicad battery packs to CR2032 cells some two decades back.
I wouldn't worry too much about those 'expiry dates'; that just means they only have 90% or more of their original capacity left (assuming they'd been properly stored somewhere where the ambient temperature had been maintained at a reasonably constant 20 deg C for the whole 9 or 10 years since they were shipped in from the factory).
The "CMOS RAM" was originally the 70 odd 'spare registers' in the RTC chip used in the IBM AT machine and its clones. The RTC functionality, along with its registers, was eventually incorporated into one of the main chips on the later 486 and Pentium MoBos.
During this transition from 80286, 80386, 80486 and P50/60/66 AT class machines to the P75 and beyond ATX class, the 3 cell rechargeable NiCad packs got upgraded to the humble (but over-priced) CR2032 cell. Also, during that period, the original EPROM chips used for storing the BIOS code (the "BIOS ROM") were upgraded to flash ram which permitted in-place BIOS updates without the need to physically swap out the old BIOS eprom chip with another updated BIOS eprom chip.
The battery only provides backup for the CMOS technology based RTC functions and its collection of registers used to store the hardware system configuration settings. It doesn't (and never ever has) provided backup to an actual cmos memory chip being used in place of a ROM or eprom chip to store the BIOS code.
IMHO a high volume seller on eBay with decent feedback is going to shift far more than Boots or a supermarket.
Correct BUT the 'registers used to store the hardware system configuration settings' are now usually in the Flash, not in CMOS on the RTC chip.
As I said, the only thing you lose with the battery power these days is the time and date.
And, quite possibly, before TNP as well! :-)
All joking aside, that problem largely disappeared with the advent of ATX for the vast majority of computer users who were happy to accept the modest 1 to 4 watts standby consumption after letting the OS shut the computer down.
Unless one was in the habit of shutting the mains off after the computer had switched itself off at the end of a session, the 5VSB was used to relieve the CR2032 cell of its battery backing task so could last several shelf lifetimes in principle (stored under proper conditions, that 10 year shelf life just meant that it would still have 90% or more of its original capacity left at the expiry of that deadline).
Before the advent of ATX, the prime warning symptom of an EoL CR2032 was a slowing of the RTC maintained time, often to a standstill during over- night shutdowns. The registers in the RTC (or its on chip integration) could maintain their contents (the cmos settings) on a voltage below the specified 2 volt minimum.
The actual clock circuitry, crystal oscillator and the chain of dividers, would lose accuracy when the battery voltage hit the 2 volt lowest limit. Initially as the result of the oscillator slowing down losing several seconds an hour during overnight shutdowns but worsening as the voltage reduction caused miscount events in the divider circuitry.
If you were slick enough about swapping a replacement CR2032 into the socket, you could save the cmos settings (the register contents were checksummed and verified during the POST on every reboot to prevent misconfiguration arising as a result of data corruption). If a cmos checksum error was detected, the system would halt and report the error to screen (and send a beep error code to the PC Speaker if fitted) and give you the option to either enter the cmos setup menu or else load defaults and continue the boot process.
If you were lucky, the only thing you'd have to deal with after a coin cell change would to correct the date and time which could be done from the desktop environment (the OS, whatever it was, would update the RTC settings for you, neatly avoiding a visit to the cmos setup menu by the user).
If you were *unlucky*, you'd simply be faced with that error message on restart so no worse off than having to hit the F1 key to load defaults and boot the machine (assuming the defaults didn't cause any problems in the choice of boot device on account of some customisation of the hardware - if this was the case, the user would very likely be aware and know just how to sort the problem out anyway). Whatever the result, no serious consequences could arise if any of the cmos settings had suffered corruption, that fact would be detected and the user warned and given options to deal with the problem.
Most likely because the 5VSB has been relieving the coin cell of the RTC load (only works if you've had it plugged into the mains for most of that
15 year period).
As I've already pointed out, this won't make a huge difference unless they've been badly stored. A Lithium primary cell, when stored under the proper conditions, will still have 90% or more of its original capacity left at the end of the ten year shelf life (or 80% or more after 20 years on the shelf).
Unless extraction of more than 90% of a CR2032's energy content is a critical requirement, such profligate spending on "Only the best that money can buy" will be a false economy.
Checking out my recent CR cell purchases from pound shops in recent years, I noticed a partially used 8 cell pack, under the brand name of FusionMax, with 3 of the 4 CR2032 and a CR2025 removed with an expiry date of Dec 2015 which I seem to recall happily buying in January 2016, knowing full well that they'd still contain 90% of their original capacity *because* they had reached the end of their 10 (or possibly 8) year shelf life (if based on an 8 year shelf life, they'd have had 92% or more of their original capacity left).
I also have an unused 8 cell pack (brand name, PowerCell) which I must have purchased from a pound shop during the past 12 months with 4 x CR2032, and two each of CR2025 and CR2016. The expiry date on this pack is given as "10/2020". In this case, based on a 10 year shelf life expiry date, those cells still have in excess of 93% of their initial capacity (otherwise it will be in excess of 95% if the expiry date is based on an
8 year shelf life).For further insight into CR lithium coin cell technology, take a look at the following pdf linked here :
They refer to an 8 year shelf life for their CR2032 cells but it's still based upon the ~1% loss per year of storage at normal room temperature meaning the remaining capacity will be 92% rather than the more usual 90% for the more common 10 year shelf life quoted by most other manufacturers.
If we're talking about AT motherboards from the 2nd millennium, almost certainly a 3 cell NiCad battery pack (typically 60mAH) soldered onto the board. These were prone to leakage of the highly corrosive alkaline electrolyte past the seals which, in extreme cases, could inflict collateral damage by destroying a thin circuit trace or two in the vicinity of the battery.
Luckily, most of these boards more than outlived their usefulness long before this state of affairs arose so was only a serious issue in a very few cases, typically legacy PCs used by small specialist businesses to control CnC machines in their factories or machine shops.
Well I can't explain it but when I used cheap lithium button cells for my car's remote hoofer doofer, they only lasted a week before the car started declaring that the key fob battery was low.
Replacing the cells with Duracells cured to problem. It seems that not all cells are created equal,
Tim
So, my cell from Morrisons cost a quid or so and not 10p - this PC that I built early last year has a £220, 24", 16:10 monitor and I've just spent nearly £300 on an SSD - where's me 90p or so fit into that?
As the PC that had the "profligate" cell will be used to try out some distros of Linux, it might have another 10 years' life - I used v. good components for it.
I do take your point that agonising over whether to spend £1 or £2.70 on a 'quality' cell versus 25p on a pound shop *replacement* CR2032 (hopefully some 5 or more years down the line) does seem rather ridiculous when you consider the several hundreds of Pounds Sterling already invested in the desktop computer setup. :-)
However, the OP was asking a serious question and I provided a serious answer. Those Pound shop Lithium coin cell packs with 2 to 4 CR2032s (even with recently matured expiry dates) are more than ample for the task of maintaining the RTC and its registers (CMOS memory) in any modern [1] desktop PC.
It seems to have been demonstrated by one contributor that pound shop CR2032 cells aren't always a wise investment when it comes to car key remote fobs. The problem in this case would seem to be a deficiency in handling low duty cycle high current pulse demands created by the transmitter circuit in these key fob remotes; a problem that simply doesn't exist in the useage case of a PC.
rather neatly describes the issue involved as it relates to such useage (internal resistance).
[1] I'm using the description, 'modern' in the sense of any ATX PC that utilises a lithium coin cell (typically the ubiquitous CR2032) that's left connected to the mains supply whilst shut down in between active sessions of use which can double or triple the effective life of the coin cell compared to a PC that's actually disconnected from the mains upon shutdown to save the 1 to 4 watts overnight draw of standby power (5VSB psu rail).
True - although if my PC failed, I'd likely want a new battery quickly.
I am fairly sure I had a PC clone that ran the BIOS from CMOS memory, but had a copy backed up in flash memory that could be reloaded if the working copy was lost or degraded. This was said to be a speed benefit. I can't remember what make it was, nothing specially interesting I don't think.
I remember that vaguely: I think however that it copied ROM into DRAM rather than CMOS
====snip====
I think you're misremembering the shadow ram technique used to speed up calls into the BIOS routines by DOS and application software. I think this became a feature on MoBos when simms replaced banks of socketed DIL packaged dram chips with the advent of the 80386 and its cheaper cousin, the 80386SX when the smallest size of simm you could start with was 256MB (8 or 9 bit wide dram modules) in systems using 16 or 32 bit wide data buses with the minimum populated memory size being either 1/2 or 1 MB using 2 or 4 simms.
Unlike the earlier 80286 based AT MoBos where you could fit exactly
640KB's worth of dram chips, leaving the top 384KB of the base memory map unoccupied for rom use, these later MoBos incorporated memory management features that could map out the 'unused' 384KB chunk (in 16KB blocks afaicr) of a 1MB simm into the extended ram space beyond the 1st MB limit of the original 8086 and 8088 cpus rather than merely disable that address range of the 1MB installed dram to allow bios roms to reside without conflict in the upper 384KB space.This meant that the bios could copy itself (and other roms) into this remapped dram then switch the rom addressing with the dram addressing so that the bios rom code could run from the copy installed into the higher performance dimms. This is basically the way shadow ram worked when you enabled 'shadowing' in the cmos setup, it had nothing to do with the 'CMOS RAM' which was just a collection (70 or so) of the registers in the RTC chip that replaced the use of dip switches which had been used previously to store hardware settings on the pre-AT MoBos.
Initially, the default cmos config setting for shadowing roms into dram was "disabled". Later on, the shadowing was enabled by default as the Power On Self Test (POST) routines became sophisticated enough to prevent shadowing into defective ram so as to prevent a show stopping crash and allow the BIOS to generate a fault code to assist in remedying any problems as a result of bad ram or MoBo chipset defects. These days, I'm not even sure whether the user still has the option of configuring shadowing of system roms in the cmos setup any more.
snip
I don't think so. I clearly remember the instructions that if a problem occurred with BIOS you should enter special 'setup' menu on boot by holding a key down and reload the BIOS, then go through ordinary BIOS setup (or set defaults) again. I wish I could remember which machine it was.
En el artículo , Roger Hayter escribió:
Gigabyte's DualBIOS?
HomeOwnersHub website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.