Thanks to extensive timing testing and end user reports, PST has uncovered an anomaly in some computer hardware that will result in experiments having extremely poor timing results. These problems range from simply faulty components on the machine, improper machine design or meeting of specifications, and interactions with power management services such as (but not limited too) Intel SpeedStep and AMD Cool'n'Quiet. In some of the cases when the problem is found, the timing error is so large that an end user can use a stop watch to determine it. E-Prime offers time auditing features to assist experiment authors in determining when experiment design or the operating system causes any timing issues. These time auditing features do not help in determining this problem because the problem is that the physical clock is not reporting accurate values to E-Prime.
PST has identified a method to diagnose a machine that is having these problems. This involves comparing the delta values between the E-Prime clock and the clock used to track the system's date and time. On systems where the hardware timing problem occurs, the deltas will report excessive values (greater than 1000ms per hour). This method will produce false positives because the clock used to track date/time will not be accurate, many systems sync date/time to an Internet service such as Time-Windows.com, and comparing any two clocks over a large amount of time will show drift values. PST offers an update to the RefreshClockTest experiment that can be run for a length period of time (hours) to be able to find machines that may exhibit this problem. When the diagnostic offerings determine that a machine may be susceptible to a hardware timing problem, an alternative hardware clock can be sought and used in many cases. In the cases of false positives, the diagnostic can help determine if the problem is related to the hardware clock or the system date time clock having excessive drift characteristics. An E-Prime experiment script provided by PST can also be used to collect triggers from a frequency generator or another E-Prime computer to assist in this diagnosis. Although this problem is most likely to occur on laptop machines and those that have aggressive power saving features, this problem has been seen on desktop machines and higher end CPUs.
Using properties on the E-Prime clock, PST will evaluate this diagnosis procedure within E-Prime 2.0 BETA. Because of the potential for false positives, PST has chosen not to implement these properties in E-Prime 1.x.
E-Prime and Psychology Software Tools provide the tools not only in diagnosing the problem, but in many cases can correct the issue by finding an alternative hardware clock that will provide stable timing values.
Diagnosing the Problem
The following actions should be provided on every lab computer...
Run the RefreshClockTest experiment for at least two hours.
To download the RefreshClockTest experiment, click here.
NOTE: The RefreshClockTest experiment will specifically indicate during the startup screens that it is going to check for this timing issue. Older versions of this experiment will not!
Upon completion of the experiment, a warning message will appear if the problem may exist.
If you receive this warning message, attach the generated .edat(2) file to PST Web Support for evaluation.
From the values in the .edat(2) file it can be determined if the problem is within range or additional tests should be considered.
The additional test involves running an experiment on the computer attached to either a frequency generator or another E-Prime computer that is known to not exhibit this timing problem.
It can then be determined if the problem is a false positive in that the system date clock has the issue or if there are serious problems with the computer and should be flagged as not being used.
Under many circumstances, PST can provide a utility that will query and configure the system to use an alternative chipset on the motherboard that will provide acceptable timing results.
NOTE: The diagnosis should occur on a routine basis. One end user found this problem to be intermittent and the faulty chipset would exhibit the problem upon aprx 1 out of 4 reboots of machine.
Steps to prevent timing issues
- Run the diagnostics (see above) to ensure the lab machine does not have an inherent timing problem
- Use as few background applications as possible. This can be achieved under Windows XP with MSCONFIG, using diagnostic mode, adding the Plug and Play service, and any other essential services and applications, (INFO: How to use MSCONFIG to troubleshoot machine configuration and reduce background applications ).
- Disable any screen savers
- Disable all power management features
- Connect the AC adapter to the computer at all times
Machines and Configurations Known to Exhibit This Issue
The following are machines that have been found to exhibit this issue and their diagnosis.
This does not imply that all machines of this type will have the same result.
This table is to outline machine classes that may be more likely to exhibit the problem and should further be tested if they are the same type in your lab.
Please report your findings to E-Prime Web Support so that your machine classes can be added to this listing as applicable.
|Machine Class||Operating System||Diagnosis|
|Fujitsu 2.4 GHz P4 laptop||Windows XP||Chipset problem determined. End User Opted to not use machine for data collection.|
|Dell Precision 670x64||Windows XP||Chipset problem determined. Utility found stable alternative hardware clock|
|Fujitsu Siemens Desktop, P4, 2.80 Ghz||Windows XP||Drift problem found by RefreshClockTest. Alternative stable clock not found. End user did not report results from additional tests. Remains pending.|
This topic applies to:
All versions of E-Prime