New release of sys_basher, Linux hardware stress test andbenchmarking tool - Hardware

This is a discussion on New release of sys_basher, Linux hardware stress test andbenchmarking tool - Hardware ; I've put another release of sys_basher, rev 1.0.5, on the web. I wrote sys_basher to test the stability of my new Core2 system. It does extensive memory diagnostics, disk I/O tests and floating point tests. It uses pthreads to run ...

+ Reply to Thread
Results 1 to 13 of 13

Thread: New release of sys_basher, Linux hardware stress test andbenchmarking tool

  1. New release of sys_basher, Linux hardware stress test andbenchmarking tool

    I've put another release of sys_basher, rev 1.0.5, on the web. I wrote
    sys_basher to test the stability of my new Core2 system. It does extensive
    memory diagnostics, disk I/O tests and floating point tests. It uses
    pthreads to run from 1 to 256 independent threads so it's capably of
    keeping all of the cores in a multiprocessor system running at 100% load
    using 100% of the memory bandwidth. It also provides memory, disk and
    floating point bandwidth stats.

    I've tested it on the following systems,

    PII 450MHz, 384M, Fedora Core 6 (32 bit)
    PIII, 500MHz, 512M, Fedora Core 4
    Dual Xeon (P4 generation), 2G, Fedora Core 4
    Athlon 64 3400+ (754 pin), 1G, Fedora Core 5
    Athlon 64 3800+ (939 pin), 2.5G, Fedora Core 6
    Athlon 64 X2 4400+ (939 pin), 4G, Fedora Core 6 (64 bit)
    Core2 Duo E6700, 4G, Fedora Core 6 (64 bit)

    The program is straight POSIX C so it should run on any *nix system. There
    is a PTHREADs library available for Windows so it's possible that it could
    also run on a Windows system with some work. I'd appreciate it if any of
    you who are interested in having an open source hardware stress test would
    download the program and run it. Please let me know if there are any
    problems with the program. I'd also appreciate any other suggestions and
    feedback that you might have.

    You can get sys_basher here,

    http://www.polybus.com/sys_basher_web/

  2. Re: New release of sys_basher, Linux hardware stress test and benchmarking tool

    General Schvantzkoph wrote:

    [snip]
    > I've tested it on the following systems,


    [snip]
    > I'd also appreciate any other suggestions and
    > feedback that you might have.

    [snip]

    It compiled and ran without any FAILs on an
    Athlon 64 X2 3600+ (AM2), 0.5G, Mandriva 2007.


    --
    sig goes here...
    Peter D.

  3. Re: New release of sys_basher, Linux hardware stress test andbenchmarking tool

    On Mon, 08 Jan 2007 23:04:04 +1100, Peter D. wrote:

    > General Schvantzkoph wrote:
    >
    > [snip]
    >> I've tested it on the following systems,

    >
    > [snip]
    >> I'd also appreciate any other suggestions and
    >> feedback that you might have.

    > [snip]
    >
    > It compiled and ran without any FAILs on an
    > Athlon 64 X2 3600+ (AM2), 0.5G, Mandriva 2007.
    >
    >


    Are there any features that you would like me to add?

  4. Re: New release of sys_basher, Linux hardware stress test and benchmarking tool

    General Schvantzkoph wrote:

    [snip]
    > Are there any features that you would like me to add?


    A data base of what is normal for a given processor type.

    Prettier formatting of the memory size. ("ls" can be made
    to do some pretty formatting, but I've had a look at the
    source and I can't make any sense of it.)

    Is it possible to stress the IRQ system via code? Or does
    that require physically pressing the keyboard?

    All of these are just idle suggestions. Thank you for the
    work that you have already done.


    --
    sig goes here...
    Peter D.

  5. Re: New release of sys_basher, Linux hardware stress test andbenchmarking tool

    On Tue, 09 Jan 2007 15:06:45 +1100, Peter D. wrote:

    > General Schvantzkoph wrote:
    >
    > [snip]
    >> Are there any features that you would like me to add?

    >
    > A data base of what is normal for a given processor type.
    >
    > Prettier formatting of the memory size. ("ls" can be made
    > to do some pretty formatting, but I've had a look at the
    > source and I can't make any sense of it.)
    >
    > Is it possible to stress the IRQ system via code? Or does
    > that require physically pressing the keyboard?
    >
    > All of these are just idle suggestions. Thank you for the
    > work that you have already done.
    >
    >


    I can't think of a way to stress the IRQ system, if anyone has any
    suggestions I'd be glad to add it.

  6. Re: New release of sys_basher, Linux hardware stress test and benchmarking tool


    General Schvantzkoph wrote:
    > On Tue, 09 Jan 2007 15:06:45 +1100, Peter D. wrote:
    >
    > > General Schvantzkoph wrote:
    > >
    > > [snip]
    > >> Are there any features that you would like me to add?

    > >
    > > A data base of what is normal for a given processor type.
    > >
    > > Prettier formatting of the memory size. ("ls" can be made
    > > to do some pretty formatting, but I've had a look at the
    > > source and I can't make any sense of it.)
    > >
    > > Is it possible to stress the IRQ system via code? Or does
    > > that require physically pressing the keyboard?
    > >
    > > All of these are just idle suggestions. Thank you for the
    > > work that you have already done.
    > >
    > >

    >
    > I can't think of a way to stress the IRQ system, if anyone has any
    > suggestions I'd be glad to add it.


    Present user with the prompt "Press 8 ide drives into the system
    and reboot when ready" then perform I/O in parallel to multiple
    partitions


  7. Re: New release of sys_basher, Linux hardware stress test and benchmarking tool


    General Schvantzkoph wrote:
    > On Tue, 09 Jan 2007 15:06:45 +1100, Peter D. wrote:
    >
    > > General Schvantzkoph wrote:
    > >
    > > [snip]
    > >> Are there any features that you would like me to add?

    > >
    > > A data base of what is normal for a given processor type.
    > >
    > > Prettier formatting of the memory size. ("ls" can be made
    > > to do some pretty formatting, but I've had a look at the
    > > source and I can't make any sense of it.)
    > >
    > > Is it possible to stress the IRQ system via code? Or does
    > > that require physically pressing the keyboard?
    > >
    > > All of these are just idle suggestions. Thank you for the
    > > work that you have already done.
    > >
    > >

    >
    > I can't think of a way to stress the IRQ system, if anyone has any
    > suggestions I'd be glad to add it.


    What northbridge (system?) temperatures you're
    getting? I got mine yesterday to 48C in Windows!
    2C away from the 50C cutoff. Not sure what Supermicro PDSBA
    designers were thinking.


  8. Re: New release of sys_basher, Linux hardware stress test andbenchmarking tool

    On Tue, 09 Jan 2007 18:17:32 -0800, sndive wrote:

    > General Schvantzkoph wrote:
    >> On Tue, 09 Jan 2007 15:06:45 +1100, Peter D. wrote:
    >>
    >> > General Schvantzkoph wrote:
    >> >
    >> > [snip]
    >> >> Are there any features that you would like me to add?
    >> >
    >> > A data base of what is normal for a given processor type.
    >> >
    >> > Prettier formatting of the memory size. ("ls" can be made
    >> > to do some pretty formatting, but I've had a look at the
    >> > source and I can't make any sense of it.)
    >> >
    >> > Is it possible to stress the IRQ system via code? Or does
    >> > that require physically pressing the keyboard?
    >> >
    >> > All of these are just idle suggestions. Thank you for the
    >> > work that you have already done.
    >> >
    >> >

    >>
    >> I can't think of a way to stress the IRQ system, if anyone has any
    >> suggestions I'd be glad to add it.

    >
    > What northbridge (system?) temperatures you're
    > getting? I got mine yesterday to 48C in Windows!
    > 2C away from the 50C cutoff. Not sure what Supermicro PDSBA
    > designers were thinking.



    How do I read the temps from a program? Recording the highest temp that's
    reached during the test would be useful.

  9. Re: New release of sys_basher, Linux hardware stress test andbenchmarking tool

    On Tue, 09 Jan 2007 18:15:04 -0800, sndive wrote:

    > General Schvantzkoph wrote:
    >> On Tue, 09 Jan 2007 15:06:45 +1100, Peter D. wrote:
    >>
    >> > General Schvantzkoph wrote:
    >> >
    >> > [snip]
    >> >> Are there any features that you would like me to add?
    >> >
    >> > A data base of what is normal for a given processor type.
    >> >
    >> > Prettier formatting of the memory size. ("ls" can be made
    >> > to do some pretty formatting, but I've had a look at the
    >> > source and I can't make any sense of it.)
    >> >
    >> > Is it possible to stress the IRQ system via code? Or does
    >> > that require physically pressing the keyboard?
    >> >
    >> > All of these are just idle suggestions. Thank you for the
    >> > work that you have already done.
    >> >
    >> >

    >>
    >> I can't think of a way to stress the IRQ system, if anyone has any
    >> suggestions I'd be glad to add it.

    >
    > Present user with the prompt "Press 8 ide drives into the system
    > and reboot when ready" then perform I/O in parallel to multiple
    > partitions


    I am performing IO in parallel to multiple partitions. To do it you use
    the -paths switch to give it a list of directories on different
    partitions. Each thread does I/O independently so if you set the -threads
    switch to specify the number of threads that you would like to run
    simultaneously.

  10. Re: New release of sys_basher, Linux hardware stress test and benchmarkingtool

    General Schvantzkoph wrote:
    >
    > How do I read the temps from a program? Recording the highest temp that's
    > reached during the test would be useful.


    Under 2.6.x kernels, raw temperature sensor information can be found in
    the /sys/class/hwmon/hwmon*/device/temp*_input files. You should be
    able to programmtically find and read those. (Note that there may be
    duplicates. On one of my systems the sensors can be read over either
    the I2C bus or the ISA bus, so hwmon0 and hwmon1 both refer to the same
    hardware. I suspect that it may be sufficient for your purposes to hard
    code in hwmon0, though.)

    AFAICT, dividing those values by 1000 will give you the temperature in
    degrees C. (At least that's how it looks on my hardware. No idea if
    that fact is device-specific, though.)

  11. Re: New release of sys_basher, Linux hardware stress test and benchmarking tool

    On Wed, 10 Jan 2007 18:32:13 -0500, John-Paul Stewart wrote:

    > General Schvantzkoph wrote:
    >>
    >> How do I read the temps from a program? Recording the highest temp that's
    >> reached during the test would be useful.

    >
    > Under 2.6.x kernels, raw temperature sensor information can be found in
    > the /sys/class/hwmon/hwmon*/device/temp*_input files. You should be
    > able to programmtically find and read those. (Note that there may be
    > duplicates. On one of my systems the sensors can be read over either
    > the I2C bus or the ISA bus, so hwmon0 and hwmon1 both refer to the same
    > hardware. I suspect that it may be sufficient for your purposes to hard
    > code in hwmon0, though.)
    >
    > AFAICT, dividing those values by 1000 will give you the temperature in
    > degrees C. (At least that's how it looks on my hardware. No idea if
    > that fact is device-specific, though.)



    It doesn't look like that's universal. I see those files on my Core2
    system (although they are useless because LM sensors doesn't work on the
    Abit AB9 Pro), but I don't see them on any of my Athlon64 systems or my
    Xeon system.



  12. Re: New release of sys_basher, Linux hardware stress test and benchmarkingtool

    General Schvantzkoph wrote:
    > On Wed, 10 Jan 2007 18:32:13 -0500, John-Paul Stewart wrote:
    >
    >> General Schvantzkoph wrote:
    >>> How do I read the temps from a program? Recording the highest temp that's
    >>> reached during the test would be useful.

    >> Under 2.6.x kernels, raw temperature sensor information can be found in
    >> the /sys/class/hwmon/hwmon*/device/temp*_input files. You should be
    >> able to programmtically find and read those. (Note that there may be
    >> duplicates. On one of my systems the sensors can be read over either
    >> the I2C bus or the ISA bus, so hwmon0 and hwmon1 both refer to the same
    >> hardware. I suspect that it may be sufficient for your purposes to hard
    >> code in hwmon0, though.)
    >>
    >> AFAICT, dividing those values by 1000 will give you the temperature in
    >> degrees C. (At least that's how it looks on my hardware. No idea if
    >> that fact is device-specific, though.)

    >
    >
    > It doesn't look like that's universal. I see those files on my Core2
    > system (although they are useless because LM sensors doesn't work on the
    > Abit AB9 Pro), but I don't see them on any of my Athlon64 systems or my
    > Xeon system.


    Interesting. I checked on both a dual Xeon (E7505 motherboard chipset,
    w83627hf sensor chip) and an old dual P-III (Via 694 motherboard
    chipset, as99127f sensor chip) before posting. I guess that wasn't enough!

  13. Re: New release of sys_basher, Linux hardware stress test and benchmarking tool


    General Schvantzkoph wrote:
    > On Wed, 10 Jan 2007 18:32:13 -0500, John-Paul Stewart wrote:
    >
    > > General Schvantzkoph wrote:
    > >>
    > >> How do I read the temps from a program? Recording the highest temp that's
    > >> reached during the test would be useful.

    > >
    > > Under 2.6.x kernels, raw temperature sensor information can be found in
    > > the /sys/class/hwmon/hwmon*/device/temp*_input files. You should be
    > > able to programmtically find and read those. (Note that there may be
    > > duplicates. On one of my systems the sensors can be read over either
    > > the I2C bus or the ISA bus, so hwmon0 and hwmon1 both refer to the same
    > > hardware. I suspect that it may be sufficient for your purposes to hard
    > > code in hwmon0, though.)
    > >
    > > AFAICT, dividing those values by 1000 will give you the temperature in
    > > degrees C. (At least that's how it looks on my hardware. No idea if
    > > that fact is device-specific, though.)

    >
    >
    > It doesn't look like that's universal. I see those files on my Core2
    > system (although they are useless because LM sensors doesn't work on the
    > Abit AB9 Pro), but I don't see them on any of my Athlon64 systems or my
    > Xeon system.


    Looks like Intel has screwed up and did not give programming info the
    the
    lm sensors programmers for g965/x975 :-[
    I don't have the temp_* files in /proc and /sys for PDSBA either.
    Learning to like Windows :-)


+ Reply to Thread