Something incredibly curious...
entropy
Yah-Der-Hey (Wisconsin)
Ok, I happened to be watching the Windows clock when something didn't seem right. Took me awhile to figure out what it was, but when I found it, I was stunned. Every fifth second it pauses for more than a second! The amazing thing is that it's been dead on for months, almost as if it is counting too fast to begin with...does this happen to anyone else??
Note: It's a .rar inside the .zip because .rar isn't acceptable and .zip was too big
Note: It's a .rar inside the .zip because .rar isn't acceptable and .zip was too big
0
Comments
This could be load of things running at once in Windows, especially if it skips a second right after each pause. BUT, since it has been going for month like this, lets cross-check time against the Cesium Waterfall clock in Colorado if you are in the US. http://www.time.gov/ will open a page that might take few seconds to load. When it loads a picture of the US, with a small picture of Alaska and Hawaii as outlines to one side, click on area where you are. IF you get nothing, then look up at top of green block that has a JavaScript error (Sun Java is used, not Jscript) for a link that has HTML in it. Click that link. You will get a clock that will start counding seconds. IF you get no error, the government time clock site will just load a small Java routine that will show a digital clock in browser text. Ignore the accuracy shown below the time counting away. See if over a 30 second period the Widnwos clock stays off in time. If it is way off, this time tick display can be used to set the clock in Windows. Right click your time clock, choose to adjust time and date, drag the time settign window partly out of the way so you cna see the time, and set it. Do it agian in 24 hours to check time.
Windows might be pausing the time count it shows due to intermittent load cycles, then catching up, or might be losing time. Let us know what is off from Time.gov's time as to time and date if way off, after first check, that will give us a better idea as to what to look for.
It doesn't load a bunch of crap, just checks the time when I tell it to.
Some of my machines gain, and some loose. None are even close. They are all off by minutes/month.
WTF?
Clocks here on computers are within a second all the time (Linux also NNTP syncs time, within the O\S). Essentially, Atomtime does this for you also in Windows, and syncs to NNTP servers fed by the cesium clock in US through network of time servers every once in a while on a schedule if you run it all the time (or when you run it and tell it to sync if your computer clock is reasonably stable)-- when I first used atomtime, I got a clock server in Europe that did not feed time zone it was based at, had a clock that was suddenly running at German time and dating (time zone on this box was set right) but that happened way back in the EARLY 1990s when Atomtime was in its software infancy-- I ended up finding the NNTP docs and server list, and then played with server naming until I got a 2nd layer NNTP server here on east coast that did work right for my older version of Windows and with Atomtime, and Linux now uses that time server also here and now (Shorty, at the University of Delaware). The Cesium waterfall clock in Colorado measures atomic flow also, so the time is atomic flow based from that still. Now we have radio sync atomic clocks for walls here, wild how older tech goes into our small appliances these days.
Atom Time for Windows was neat and still is a neat program, though it does not have to run all the time unless you have a box with a dead or dying CMOS battery (which also powers clock in computer when computer is not running, so if you have a box that is not running a lot, clock can be wildly off at every cold start boot). I recommend Atomtime with a sync at boot and frequently after that while box is up and a CMOS battery check and\or replacement for boxes that have clocks that are wildly off and fold or do things that need timestamps in logs or do tasks that need accurate time and dating for files. VERY good utility now, has been for quite some time...
Linux syncs back to hardware clock on motherboard by default at reboot or halt\shutdown time, so if time is off by a lot it might be good to also "restart" (aka reboot) if you run Linux, after first and possibly second time sync. Then the clock will get settled in, Linux applies a correction automatically if you run the NTP daemon when box starts up, and after each time sync it tunes the correction value until it is so close to zero as not to even be funny.
About the time.gov site you showed. I was within 5 seconds. Always. My computer's clocks have never synchronized with a site, because every time it did, it got messed up. We called tech support once (for a Dell or Compaq - forget which) and they actually told us not to synchronize. From that point on, the clock was accurate. This is just very strange.
Windows gets its time from the bios.
The "second" part of the bios clock changes every....second.
Unfortunately, windows does not update the time at 1 Hertz (or one Hert), most likely somewhere around 1.2 Hz.
Example:
windows checks the time @ 0s, bios @ 0, displays 0
windows checks the time @ 1/1.2s, bios only @ .833333, displays 0 (truncates because the bios second is only at "0")
windows checks the time @ 2/1.2s, bios only @ 1.6666666, displays 1
windows checks the time @ 3/1.2s, bios only @ 2.5, displays 2
windows checks the time @ 4/1.2s, bios only @ 3.333, displays 3
windows checks the time @ 5/1.2s, bios only @ 4.166, displays 4
windows checks the time @ 6/1.2, bios @ 5, displays 5
windows checks the time @ 7/1.2, bios @ 5.83333, displays 5
every 5th windows "second" will happen twice.
I've never really paid my clock much mind before.
Ok, that pause is there to keep it the same as the CMOS clock. Watching the clock at http://www.time.gov the windows clock gets almost 1 second ahead of the time at the website, it will then pause for that amount of time to keep in check with the CMOS clock.
Why couldn't Microsoft code the windows software clock to stay as accurate as the CMOS clock and not run fast and have to pause every 5 seconds?
Shelf life on the modern Lith CMOS cells is 5 years ON SHELF. Use life, if after 1-2 years on shelf the battery is installed in a computer, can be 2\3 normal to 58% normal USAGE LIFE of 3-4 years if you do not reset the computer's CMOs often as when setting up OCing. Ship half dead?? YUP, ALMOST....
Same cell, in an electric watch, lasts about 3 years. So why would a computer be different for a clock???
Yes, sync rates of a display clock in windows can affect this also, but CMOS cells of Lith type do NOT have a flat curve for voltage versus remaining charge pool. They instead have this kind of curve-- slightly positive until 92-95% drained, then a sharp drop to zero. The positive curve has, on many computers, yielded a FAST clock for me, then a clock that radically loses time, then a clock that is YEARS off after every cold boot. This happens irregardless of OS on same hardware, for many kinds of hardware.
Thus, primary cause is NOT OS, period. seen it in DOS, Linux, Windows of many versions. Here, XP and Linux sync automatically, no problems, but display clock staggers when Windows polls performance variables in the system management console or for performance logs, and did in 2000 also-- and stagger frequency changes with performance poll interval, if the box is doing many other things, else NOT.
In Linux, stagger happens when system is real busy, ONLY. But clock accuracy patterns overall follow strength of CMOS cell and voltage curve it has. Linux does not like being off 15-30 min after two-three months of runtime between cold-boots.
So I timesync my clocks on both boxes here. And change CMOS cells every 2.5-2.9 years(when it is convenient), and EVERY TIME I change I voltage output and LOAD test old, and they are typically 90-95% discharged. I date them when stuck in. Been doing this for DECADES. So doubt if you want, and prove its right yourself. With simple records.
Also, out of the new Lith computer cells I have purchased, not one in ten has arrived outputing 3.3 Volts DC (which is full charge NOMINAL)-- normal average is about 3.0-3.1 VDC. That is 90.9-93.9 of full charge to begin with. So---- 100-7% predrain due to not fully being charged (mfr date STAMPED on batteries has been a month to two months prior, from the national distributors I buy from, once bought from a local computer store and RETURNED the battery when I showed them the 1999 mfr date during calendar year 2002, and then proved battery was 2\3 dead WHEN SOLD) on arrival, -1\3 after one year, that's 33% MORE, that's 40% after one year in use (best case, add 3% more drain for the 90% case if you want pessimistic), then add another third for second year, takes us to 73% drained and feeding slightly HIGHER than normal voltage, and guess what?? There IS no 33% more for next full year of use. There's 27-24% left of charge, less the part of charge that is next to useless, last 5% of charge or so, and that means 22-19% of usable charge left. That is half to 2\3 of a year's worth. So, I maintainance change batteries in my and my mom's computers and my watches and the radio front door bell button about every 2.5-2.9 years (I did say only one out of ten arrived real close to fully charged, high end of range allows for that case).
Number one, use life NE shelf life, number two, lith cell effectivenss DECREASED when cells get HOT, number three this is a decades long pattern for hundreds of boxes. Everyone??? No, _just_ 90% of folks, out of 300-400 cells I have R&R'd _myself_ over the years. Most folks change thier own CMOS Cells, if they are MY customers. Sometimes I have to then help them figure out how to get CMOS settings right again afterwards, BUT....
http://www.gregorybraun.com/StayLive.html
BTW shwaip that 1.2 Hertz table is really cool.
Have you done a virus check lately?