On Jul 4, 6:40*am, Paul <nos...@needed.com> wrote:
> mttc wrote:
> > On Jul 3, 12:08 am, Paul <nos...@needed.com> wrote:
> >> mttc wrote:
> >>> On Jul 2, 4:55 am, Paul <nos...@needed.com> wrote:
> >>>> mttc wrote:
> >>>>> Gigabyte EX58-UD4 Board, Clean XP SP3, disconnect from network.
> >>>>> When I play mp3 with Microsoft MediaPlayer, the XPclockstartlagged
> >>>>> about 3 sec per Minute! This happen only if EasyTune tools is also
> >>>>> running (Gigabyte board tune tool).
> >>>>> I try to play with VLC, and I not see anylagged.
> >>>>> I try it again on win7 clean install and I got the same. I try alsoto
> >>>>> replace the Power Supply or use with one Stick of memory. but all the
> >>>>> same.
> >>>>> Any Idea please?
> >>>> Pretty simple. Stop using EasyTune.
> >>>> A potential mechanism to screw up a systemclock,
> >>>> is described here.
> >>>>http://en.wikipedia.org/wiki/System_Management_Mode
> >>>> * * "and causeclockticks to get lost"
> >>>> Just a guess,
> >>>> * * Paul
> >>> Did you mean that ET6 run on System Management Mode?
> >>> Did this Explain why I gotClockslow also after Sis Sandra BanchMark?
> >>> why Sandra not cause me Lag on other boards systems?
> >>> Thenks
> >> To make the WinXP system timeclockrun slow, you need a mechanism which
> >> causesclocktick interrupts to get lost. If the interrupt service
> >> routine doesn't get to run in time, that is a mechanism.
>
> >> Clocktick interrupts come from hardware, many times a second.
> >> That particular interrupt has a high priority, and in a normal
> >> system, no other activity would prevent the interrupt from being
> >> serviced. To explain a slow runningclock(where the time loss is
> >> larger than can be explained by the accuracy of the quartz oscillator),
> >> you need a mechanism which causes the loss ofclocktick interrupts.
>
> >> Software activity at the kernel level, works with sets of rules.
> >> Such as not running at interrupt level for long period of time
> >> and the like. (The hardware drivers added to the system, also have
> >> to follow the rules.) If there was a problem with the design of the
> >> OS itself, then time loss would be seen by many users.
>
> >> There have been cases, where the problem is caused by a hardware
> >> bug. For example, some motherboards with Nforce2 chipset, have
> >> trouble with time keeping. The problem was only evident, when
> >> some of the boards were overclocked (non-canonical operating
> >> frequency for the chipset). The bug may have been in the
> >> interrupt logic.
>
> >> If you see the problem, with both WinXP and with Win7, that
> >> tends to suggest there must be a hardware or driver component
> >> to this problem.
>
> >> You have a function in the BIOS called HPET, and you could
> >> try disabling that for a test. I don't know why that would be
> >> a problem, but it is all I can see that might be related,
> >> in the BIOS.
>
> >>http://web.archive.org/web/200407290...crosoft.com/wh....
>
> >> * * Paul
>
> > I disable HPET Support on BOIS. the I boot XP again and I run EsyTune
> > (Gigabyte board tune tool). I not open any other App (no need to rum
> > also the MediaPlayer). theClockStartLagged. I not underspend what
> > cause that difference.
>
> > I not get any clue from Gigabyte that have known issues about EasyTune
> > with this board. Also I ask Sis Sandra, they answer me:
>
> >>> Sandra does not use SMM, at least not on purpose.
> >>> It is possible that by the OS calls it makes for SMM to be triggered - at least by what the article says - in >>any case it should not happen thatclockis lagging.
> >>> If updating the BIOS and resetting to defaults did not fix it, what other choice is there but replacing the >>board?
>
> Well, the basic mechanism for the system time to run slow, is known.
> It has to do with interference with theclocktick interrupt and
> its servicing. So you'd be looking for a mechanism that could do that.
> It isn't exactly going to be easy to figure out, unless you happen
> on the solution by accident. (I don't know enough of the details
> of the various mechanisms, to help further.)
>
> The existence of SMI and SMM, was enough for these users to attempt
> to change the settings. The software they want to run, is real-time,
> and interference is unacceptable. So they attempted to disable SMI,
> as a means to prevent any interruptions of real-time operation.
> The BIOS can lock the controls on SMI (SMI_Lock bit), to prevent
> a potential security issue. And that might mean, that any defined
> mechanism, cannot be changed while the OS is running.
>
> http://www.mathworks.com/matlabcentr...exchange/18832
>
> *******
>
> Just out of curiosity, there is another parameter you may wish to
> evaluate on your Gigabyte motherboard. That is DPC latency.
>
> http://forums.anandtech.com/messagev...readid=2182171
>
> http://www.thesycon.de/deu/latency_check.shtml
>
> There is nothing wrong with DPCs. A DPC is a deferred procedure call,
> a way of scheduling interrupt handler work, to run at user level. That
> shortens the time spend at interrupt level. The DPC Latency Checker
> tool, measures the response time for the DPCs to get services. (So
> a DPC is added to the queue, and the latency checker records how
> long it took before it was serviced. Too long a latency, affect
> sound playback for example.)
>
> http://en.wikipedia.org/wiki/Deferred_Procedure_Call
>
> On my cheap motherboard with Core2 Duo processor, I can run the DPC Latency
> Checker. Currently, I have a 3D game, running in the background as a load..
> The DPC Latency Checker, shows green bars, having a height of no greater than
> about 250 microseconds. If I actually enter the game (with alt-tab), the
> latency goes up to about 800 microseconds. But the bars are still green,
> implying the responsiveness of the computer is not affected significantly..
>
> You might try running the DPC Latency Checker, and see if you also have
> a problem with that. If you have an interrupt problem, it may also reflect
> itself in excessive DPC latency.
>
> * * Paul
I not sure that this effect on clock. the OS determine the Time by
counting the tick interrupt, even it come late.