FSB downclock mod on the intel GL960 and GL40 --- useful info for PLL Modders

Discussion in 'Hardware Components and Aftermarket Upgrades' started by naton, Aug 21, 2011.

Thread Status:
Not open for further replies.
  1. RickiBerlin

    RickiBerlin Notebook Deity

    Reputations:
    609
    Messages:
    979
    Likes Received:
    18
    Trophy Points:
    31
    And here again, a personal thanks to "naton". Naturally, on behalf of all the "HDX Dragon owner." :notworthy:

    After a successful test with "T4400" I immediately tried "QX9300", and "Q9200" on PM965.

    My private mail to you, you've unfortunately overlooked. Here is a link of this mail. The links within the e-mail represent the beginning of the tests and are now obsolete.

    Here is a summary of test results with quad on PM965.
    Here only the images ..........and the text here

    Maybe you have another idea. If not, then these links might help other users.

    Between "test beginning with Q9200" and "preliminary test end with" QX9300 "There are many interesting articles on this topic.
    I also used pictures of you, always putting your name.
     
  2. User Retired 2

    User Retired 2 Notebook Nobel Laureate NBR Reviewer

    Reputations:
    4,127
    Messages:
    7,860
    Likes Received:
    10
    Trophy Points:
    0
    Enabling all four cores on a Q9200/QX9300

    I've PMed Unclewebb requesting assistance in identifying if MSR bits are required to enable all four quads. Though you can see at http://software.intel.com/en-us/articles/intel-64-architecture-processor-topology-enumeration/ that each logical core needs to be assigned a APIC ID to be usable. So it appears the APIC is responsible for starting all four cores. I'd suggest that the solution here would be extracting the APIC initialization code from the 8730W bios (since it supports quads) and inserting it into the HDX9000 bios.
     
  3. apostmax

    apostmax Newbie

    Reputations:
    0
    Messages:
    9
    Likes Received:
    0
    Trophy Points:
    5
    Dear naton,

    I think you can now add my laptop to your list of successful mods! :D

    So let me go through the whole story; as I had told you I had a HP G7000 with GL960 motherboard.. as a person always fighting for the best I tried to disable TME on my PLL even though I didn't have the necessary tools (fine tip soldering iron etc) ending up in damaging the PLL and therefore the motherboard... SO, that opened up the chance to upgrade my mobo to a GM965 (and therefore supporting up to 8gigs of RAM as has been stated here in NBR numerous times).. and as you can imagine I wouldn't waste this chance.
    Well, to keep the story short I did change the motherboard to a GM965, reflashed the bios using my old T2370 to F.35 (i.e. the version I had been using in my old mobo as well), eventually did the TME unlock mod successfully (this time soldering at another point though not to mess up with the PLL) and today I received my P7350.
    I short-circuited BSEL1 to Vcc to establish 200fsb and it WORKS LIKE A CHARM. I also used Throttlestop to undervolt to 1.0V and running IntelBurnTest not only gives absolutely stable results but also made the fan hardly go off..

    For the sake of argument you can have a look at the attached screenshot.

    Now, I want your advice; should I go for a X9100, X9000 or anything else? Or would it just be a waste of money? I have already tried overclocking my P7350 without success, probably because of the memory being the bottleneck. To make things even clearer I have already ordered a 800MHz dimm to see how it goes.

    Thank you for your time and your invaluable tips.
     

    Attached Files:

  4. niffcreature

    niffcreature ex computer dyke

    Reputations:
    1,748
    Messages:
    4,094
    Likes Received:
    28
    Trophy Points:
    116
    You HDX dragon owners are crazy :eek: in a good way.

    I'll start reading the pm965 and pm45 spec and have a good read.

    My guess is we need to try quads on as many pm965/gm965 laptops possible, then maybe one will just work...
    I will also look at the ACPI and core 2 spec.
     
  5. niffcreature

    niffcreature ex computer dyke

    Reputations:
    1,748
    Messages:
    4,094
    Likes Received:
    28
    Trophy Points:
    116
    This is interesting:

    From the core 2 quad mobile datasheet. These pins are "reserved" on dual core processors.
    But I wont post more randomness before convincing myself I understand it.
     
  6. User Retired 2

    User Retired 2 Notebook Nobel Laureate NBR Reviewer

    Reputations:
    4,127
    Messages:
    7,860
    Likes Received:
    10
    Trophy Points:
    0
    I've made some prelim comments on how to get a quad to work on a 965PM-equipped HDX9000 here. Perhaps worth a look?
     
  7. RickiBerlin

    RickiBerlin Notebook Deity

    Reputations:
    609
    Messages:
    979
    Likes Received:
    18
    Trophy Points:
    31
    @niffcreature

    I am delighted with this response to my post. :) So I also see, I do not write useless.
    Every new idea is always welcome, preferably with the "Quad success." ;)
    Thank you for your work and much fun to read on further!

    @nando4
    Some of your contributions to remain without a response from me. That does not mean I do not notice them. On the contrary.
    Each of your posts will remain in my head. For example, post "#42". (I gave you recognition for it)
    Please copy this post but also in the "HDX Forum"! The user "AlphaHex" should be no missing information.
    Thank you for your tireless work! :) Your ideas must be implemented, but some things take time.
     
  8. RickiBerlin

    RickiBerlin Notebook Deity

    Reputations:
    609
    Messages:
    979
    Likes Received:
    18
    Trophy Points:
    31
    For details of my test with "X9100 QHBQ Rev.E0 engineering sample" and
    "X9100 SLB48 Rev.C0 OEM" on PM965 chipset go to this post.

    In this paper, are then three further links. ("1" / "2" and "here")
    These links show all the relevant information to overclock the X9100.

    The page "1" in this thread mentioned frequency of 4.2 GHz corresponds unfortunately no stable state.
    "X9000" and "X9100" is thus identical to overclocking.
    There is only one price advantage of the "X9100 SLB48" and a temperature advantage of the X9100 QHBQ Rev E0 "to the" X9000 ".

    Sorry for my links. This is easier for me. I have to translate any text because I can not English. :rolleyes:
     
  9. RickiBerlin

    RickiBerlin Notebook Deity

    Reputations:
    609
    Messages:
    979
    Likes Received:
    18
    Trophy Points:
    31
    How to read even in this post was, there is a mysterious behavior.

    During my "1066 CPU tests" on PM965 I used different SocketMod.

    For example, "266MHz to 200MHz" and "266MHz to 166MHz."
    The reason for this was the quad test.

    In each of these "SocketMod" We mapped the corresponding frequency in the BIOS and the OS correctly.

    Example: X9100 multiplier 11.5
    at 266MHz = 3059MHz (with PM965 is not possible)
    at "SocketMod 266 to 200" = 2300MHz
    at "SocketMod 266 to 166" = 1909MHz

    The correction is then made with "ThrottleStop" in the OS to 3.8 GHz, for example.
    What is "SocketMod" used does not matter. (only for boot)

    Now to the mystery:

    I swapped my CPU about 130x. (only for "quad Try")

    Now I noticed a wrong frequency display in the bios.
    example X9100
    at "SocketMod 266 to 200" = 1909MHz
    at "SocketMod 266 to 166" = 1909MHz
    without "SocketMod" = 1909MHz (yes read correctly "without SocketMod"!) :eek:
    (in the OS, always a correct display from 2300MHz)

    My numerous studies have shown my socket behaves as a 166MHz socket.

    For the correct frequency display bios I need to get a "SocketMod 166 to 200".
    This is crazy.

    This applies to each CPU. Also original 800MHz CPU. Only the good of the multiplier is not locked on "x6".

    The OS is not interested in the "SocketMod". Here, each CPU without "SocketMod" is displayed correctly. (real 200MHz frequency)

    I initially suspected the bios. Meanwhile, I think to a contact fault of my socket. Which pin is concerned, I just could not find out until now.

    This bug affects my laptop in any way. Nevertheless, it is very interesting.

    "nando4" or other users what you say?
     
  10. naton

    naton Notebook Virtuoso

    Reputations:
    806
    Messages:
    2,044
    Likes Received:
    5
    Trophy Points:
    56
    RickiBerlin
    First, I want to salute you for the intensive experimental work you described here and the HP HDX lounge. It is really amazing :)

    Regarding one of the post you posted yesterday in the HDX lounge; a CPU is at full speed and uses its maximum Vcore when the laptop is first powered ON. This can be verified by the fact that the fan spins when one stays inside the BIOS menu for a while. The CPU speed and Vcore drops after the OS starts loading. This happens because the OS takes over the power management of the CPU from the BIOS.

    Second, I find it weird that the x9100 can post with an FSB of 200 while the T9900 can't. Is it possible that there are other differences between a normal and an extreme CPU that we don't know about besides the locked an unlocked multipliers?

    I think the only way to answer the above is to test with another 1066MHz FSB normal Core 2 Duo CPU such as the P7350.

    Third, check below:

    BSel 2 1 0 | Frequency
    ---- L L L | 266
    ---- L H L | 200
    ---- L H H | 166
    ---- L L H | 133

    I'm not sure if I'm reading you correctly. Doing 166 to 200 mod means turning BSel0 to Low. That would not do anything since x9100 BSel0 is set to Low by default. Doing the 166 to 200 mod is similar to not doing the mod since this mod has no effect.

    Now, doing the 200 to 166 mod is a different story. This mod turns BSel0 from Low to High. This mean that BSel-2-1-0 becomes L L H; i.e. FSB 133MHz. With this mod the x9100 max frequency should be 1.53GHz; i.e. half its original speed.
     
Loading...
Thread Status:
Not open for further replies.

Share This Page