Announcement

Collapse
No announcement yet.

Intel Core i7 and X58 experience?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Jupiter:

    You mentioned the "W83667HG sensor module". What is that for? What exactly did you do to set it up?

    I was able to get the coretemp module to display temperatures with lm-sensors after I saw your mention that it works. It does indeed display (slightly different) temperatures for 8 virtual cores...weird, but the temperatures seem about right. I'm getting 29-32C now under low load. Earlier, I ran 8 processes that maxed out the virtual cores and did a lot of memory access and the temperatures rose to 63-66C.

    I'm running a Gigabyte GA-EX58-UD5 motherboard (F4K BIOS), Core i7 965 at 3.2GHz with Noctua NH-U12P cooler, and 6x2GB G.Skill 1600 8-8-8-21 memory.

    Memory latency is important to a lot of the work I do, so I was curious to measure it. It doesn't look like any of the Phoronix benchmarks deal with memory latency. Anyway, I used lmbench:

    lat_mem_rd -t 32 256
    "stride=256
    0.00049 1.188
    0.00098 1.186
    0.00195 1.186
    ...
    0.03125 1.186
    0.03516 2.965
    0.03906 2.970
    0.04297 2.972
    ...
    0.15625 3.009
    0.17188 4.154
    0.18750 6.297
    0.20312 6.096
    0.21875 5.261
    0.23438 9.464
    0.25000 7.283
    0.28125 10.943
    0.31250 13.614
    0.34375 13.851
    0.37500 14.054
    0.40625 14.171
    0.43750 14.257
    ...
    28.00000 61.797
    30.00000 61.655
    32.00000 61.565

    It looks like L1 is about 1ns, L2 is 3-7ns (not sure why it rises before 256K), L3 is about 14ns, and RAM is 61ns.

    The only issue I've had is that I cannot get the first LAN port to work with Ubuntu 8.10. I was originally using the r8169 module, but Realtek's website (and some posts on the web) indicate that r8168 is the correct driver. So I compiled and install r8168, and I still cannot access the first LAN port. The second LAN port works fine. I'm not sure if this is a BIOS issue or a linux driver issue.

    Comment


    • #32
      Does anyone else have an I7 920/940 with a Gibagyte GA-EX58-UD5? I have one and I've been using Linux 64bit in it for some days now. There seems to be some problems with this hardware configuration, because I get an MCE logged every time I boot, but never more after the boot(at least for what I've seen so far). I've posted about this on the lkml and I received emails from other 2 people who have the same problem. The MCE gets logged only on kernels >= 2.6.27(64 bit).

      Comment


      • #33
        Originally posted by jwilliams View Post
        Jupiter:

        You mentioned the "W83667HG sensor module". What is that for? What exactly did you do to set it up?

        I was able to get the coretemp module to display temperatures with lm-sensors after I saw your mention that it works. It does indeed display (slightly different) temperatures for 8 virtual cores...weird, but the temperatures seem about right. I'm getting 29-32C now under low load. Earlier, I ran 8 processes that maxed out the virtual cores and did a lot of memory access and the temperatures rose to 63-66C.

        I'm running a Gigabyte GA-EX58-UD5 motherboard (F4K BIOS), Core i7 965 at 3.2GHz with Noctua NH-U12P cooler, and 6x2GB G.Skill 1600 8-8-8-21 memory.
        The 'W83667HG' is the sensor chip the Asus board has. Your mb probably
        has a different sensor chip. I am basically using a 'W83627EHF' module
        for now because it is very similar. 'modprobe w83627ehf force_id=0x8860'.
        A patch has been added to kernel-2.6.28-rc8 to add support for the
        w83667hg module. The lm_sensors repo now has a beta version that
        allows coretemp to be recognized as the Core i7 temp module.
        Sensors-detect has also been patched to detect w83667hg.

        Comment


        • #34
          Originally posted by jwilliams View Post
          Jupiter:

          You mentioned the "W83667HG sensor module". What is that for? What exactly did you do to set it up?

          I was able to get the coretemp module to display temperatures with lm-sensors after I saw your mention that it works. It does indeed display (slightly different) temperatures for 8 virtual cores...weird, but the temperatures seem about right. I'm getting 29-32C now under low load. Earlier, I ran 8 processes that maxed out the virtual cores and did a lot of memory access and the temperatures rose to 63-66C.

          I'm running a Gigabyte GA-EX58-UD5 motherboard (F4K BIOS), Core i7 965 at 3.2GHz with Noctua NH-U12P cooler, and 6x2GB G.Skill 1600 8-8-8-21 memory.

          Memory latency is important to a lot of the work I do, so I was curious to measure it. It doesn't look like any of the Phoronix benchmarks deal with memory latency. Anyway, I used lmbench:

          lat_mem_rd -t 32 256
          "stride=256
          0.00049 1.188
          0.00098 1.186
          0.00195 1.186
          ...
          0.03125 1.186
          0.03516 2.965
          0.03906 2.970
          0.04297 2.972
          ...
          0.15625 3.009
          0.17188 4.154
          0.18750 6.297
          0.20312 6.096
          0.21875 5.261
          0.23438 9.464
          0.25000 7.283
          0.28125 10.943
          0.31250 13.614
          0.34375 13.851
          0.37500 14.054
          0.40625 14.171
          0.43750 14.257
          ...
          28.00000 61.797
          30.00000 61.655
          32.00000 61.565

          It looks like L1 is about 1ns, L2 is 3-7ns (not sure why it rises before 256K), L3 is about 14ns, and RAM is 61ns.

          The only issue I've had is that I cannot get the first LAN port to work with Ubuntu 8.10. I was originally using the r8169 module, but Realtek's website (and some posts on the web) indicate that r8168 is the correct driver. So I compiled and install r8168, and I still cannot access the first LAN port. The second LAN port works fine. I'm not sure if this is a BIOS issue or a linux driver issue.
          I've got the same motherboard, but my bios version is F3, where did you find version F4? Did you update it or did you find it already like that? I have some bios problems so I'm trying to update it, but I couldn't find any new bios version till now.
          Last edited by MisterIO; 14 December 2008, 09:19 PM.

          Comment


          • #35
            Originally posted by MisterIO View Post
            I've got the same motherboard, but my bios version is F3, where did you find version F4? Did you update it or did you find it already like that? I have some bios problems so I'm trying to update it, but I couldn't find any new bios version till now.
            I downloaded a file to a USB stick and used the Q-flash BIOS utility to flash the BIOS.

            This is my first Gigabyte board in a while. I've mostly used Asus boards in the past. It seems that Gigabyte does not release official BIOS revisions as frequently as Asus. But there are beta BIOS versions, which I think are from Gigabyte but are not official, that are floating around. I got mine from Tweaktown...there is a guy that posts them regularly that I guess is a Gigabyte employee:



            I see that he has posted an F4L BIOS now for the UD5. I'm sticking with the F4K for now (I saw it mentioned on anandtech that F4K fixed some memory problems, or I would not have dared try it at all).

            Comment


            • #36
              Originally posted by MisterIO View Post
              Does anyone else have an I7 920/940 with a Gibagyte GA-EX58-UD5? I have one and I've been using Linux 64bit in it for some days now. There seems to be some problems with this hardware configuration, because I get an MCE logged every time I boot, but never more after the boot(at least for what I've seen so far). I've posted about this on the lkml and I received emails from other 2 people who have the same problem. The MCE gets logged only on kernels >= 2.6.27(64 bit).
              I have a UD5, but my CPU is a 965. My kernel is 2.6.27-9 x86_64. If you tell me what an MCE is and what your log message looks like, I will check my logs to see if I have it.

              Comment


              • #37
                Originally posted by Jupiter View Post
                The 'W83667HG' is the sensor chip the Asus board has. Your mb probably
                has a different sensor chip. I am basically using a 'W83627EHF' module
                for now because it is very similar. 'modprobe w83627ehf force_id=0x8860'.
                A patch has been added to kernel-2.6.28-rc8 to add support for the
                w83667hg module. The lm_sensors repo now has a beta version that
                allows coretemp to be recognized as the Core i7 temp module.
                Sensors-detect has also been patched to detect w83667hg.
                http://dl.lm-sensors.org/lm-sensors-3/snapshots/
                Thanks for the explanation. Does anyone know what sensor chip the Gigagbyte EX58-UD5 motherboard uses? I've looked at lspci and lshw and either it is not there or I am not recognizing it.

                I'd like to check any motherboard temperature sensors I can, since I am using a CPU cooler that blows air parallel to the motherboard, and I am wondering whether the motherboard chipsets are receiving adequate cooling.

                Comment


                • #38
                  Originally posted by jwilliams View Post
                  I have a UD5, but my CPU is a 965. My kernel is 2.6.27-9 x86_64. If you tell me what an MCE is and what your log message looks like, I will check my logs to see if I have it.
                  MCE = Machine Check Exception(very very bad, it usually means an hardware problem)

                  In this particular case though, it happens only at boot, so it may be related to a bug in the bios, which causes the MCE at boot POST time.

                  To search for it, first you need to have a kernel with MCE enabled. If you use a debian kernel, that's good. If you don't know, just do a "grep MCE config-file", where config-file is the config file used to build your kernel. You should see this :
                  CONFIG_X86_MCE=y
                  CONFIG_X86_MCE_INTEL=y
                  Then just boot your system with that kernel, wait ~10 minutes(even a lot less should suffice, but just to be sure), then on root terminal type "dmesg | grep Machine". If you get something like "Machine check events logged", then an MCE got logged. depending on the distro(and the log daemon), you may also try "cat /var/log/messages | grep Machine", but just using dmesg should suffice.

                  Another thing, have you ever heard the cpu fan in Linux? I haven't, not even when I touched more than 70?C. I guess that's because this motherboard and cpu is still not supported. I can get the readings of the cpu sensors, but the system doesn't seem to react to to the temperature readings.
                  Last edited by MisterIO; 15 December 2008, 12:37 AM.

                  Comment


                  • #39
                    Originally posted by MisterIO View Post
                    MCE = Machine Check Exception(very very bad, it usually means an hardware problem)

                    In this particular case though, it happens only at boot, so it may be related to a bug in the bios, which causes the MCE at boot POST time.

                    To search for it, first you need to have a kernel with MCE enabled. If you use a debian kernel, that's good. If you don't know, just do a "grep MCE config-file", where config-file is the config file used to build your kernel. You should see this :
                    CONFIG_X86_MCE=y
                    CONFIG_X86_MCE_INTEL=y
                    Then just boot your system with that kernel, wait ~10 minutes(even a lot less should suffice, but just to be sure), then on root terminal type "dmesg | grep Machine". If you get something like "Machine check events logged", then an MCE got logged. depending on the distro(and the log daemon), you may also try "cat /var/log/messages | grep Machine", but just using dmesg should suffice.

                    Another thing, have you ever heard the cpu fan in Linux? I haven't, not even when I touched more than 70?C. I guess that's because this motherboard and cpu is still not supported. I can get the readings of the cpu sensors, but the system doesn't seem to react to to the temperature readings.
                    I am running a stock Ubuntu kernel, 2.6.27-9. I'm not sure if MCE is enabled, but dmesg | grep -i machine does not find anything (also nothing in /var/log/messages), so I guess not.

                    I have been running the computer quite a bit (I am using it now) so if an MCE is very bad, probably I'm not getting any?

                    Also, I replied in another message that I got my F4K BIOS update from a tweaktown board, but I think my post got sent for moderation because of the URL. Hopefully it shows up soon.

                    Comment


                    • #40
                      Originally posted by jwilliams View Post
                      I am running a stock Ubuntu kernel, 2.6.27-9. I'm not sure if MCE is enabled, but dmesg | grep -i machine does not find anything (also nothing in /var/log/messages), so I guess not.

                      I have been running the computer quite a bit (I am using it now) so if an MCE is very bad, probably I'm not getting any?

                      Also, I replied in another message that I got my F4K BIOS update from a tweaktown board, but I think my post got sent for moderation because of the URL. Hopefully it shows up soon.
                      To see if you have MCE enabled, you can go to the boot dir and do "grep MCE config*"(in that dir you should have a file called config-"something") and look for the things I posted on the previous message.

                      Comment

                      Working...
                      X