*HP HDX 9000 DRAGON Owners Lounge, Part 2*

Discussion in 'HP' started by 2.0, Sep 3, 2010.

  1. 2.0

    2.0 Former NBR Macro-Mod®

    Reputations:
    13,368
    Messages:
    7,741
    Likes Received:
    1,022
    Trophy Points:
    331
    @lordgoblin,

    Yes, the one you circled is the one I was referring to (top right/upper right). That is the BIOS Chip. The big chip in the center of the pic is the PLL. See 1st Post on page 1 of this thread. (Pic there has red square around BIOS chip)

    @ Ricki,
    So far the overclock is working. Stable to 3.366Ghz (250mhz FSB) @ 1.212V. Just waiting on Abo(setFSB programmer) to make a custom entry for the PLL so I can Jump from 200Mhz FSB to 250mhz FSB without hard freeze. Otherwise, using PLL ICS9LPR501SGLF, you have to overclock in steps: 200, 209, 218, 227, 236, 245, 250mhz.


    Also, LOL @ 3500 Core, 7500 shader, 4500 mem.. Imagine if you hit apply? BOOM!
     
  2. 2.0

    2.0 Former NBR Macro-Mod®

    Reputations:
    13,368
    Messages:
    7,741
    Likes Received:
    1,022
    Trophy Points:
    331
    LOL. Sorry I missed this.

    Leaving it plugged in and cycling every month or so it the ideal compromise. You will still lose capacity every year, but nowhere near as much as 20%. More like 8-9%.

    If you don't cycle it periodically, you will lose close to 20%.
     
  3. User Retired 2

    User Retired 2 Notebook Nobel Laureate NBR Reviewer

    Reputations:
    4,127
    Messages:
    7,860
    Likes Received:
    10
    Trophy Points:
    0
    This is a big development for HDX9000 owners to extend their CPU performance. Congratulations on the success. Could accompany this with a gpu upgrade using a GTX280M/FX3700M (+modded bios containing vbios) or x2/x2E DIY ViDock :D

    I'm not sure if Abo can do anything to allow you to do a single 200->250Mhz jump instead. A ramped overclock is often required to allow components to climatize to the overclock. Could disable the PLL's spread-spectrum (byte 4, bit 1:0=00) by changing setpll's ics9lprs501pglf.lut file [INIT] section as follows to see if this helps things:

    Code:
    [INIT]
    set bytecount=0x16
    set local0=0x4505FC77[COLOR="Red"][B]FC[/B][/COLOR]F09091
    set local1=0xF2D1051C007D2710
    set local1=0x10277D001C05D1F2
    set local2=0x23A8F4F22303[BCLK_B1][BCLK_B2]
    Spread spectrum allows some PLL deviation, up to 0.5% so as to spread the 'noise' the PLL generates though it may increase instability. I've disabled spread spectrum with my 2530P overclock. Otherwise will just need to revert to using a ramped overclock using either:

    Code:
    :: Overclock 200Mhz to 250Mhz BCLK using a 9Mhz increment
    setfsb -i9 -cg[ics9lprs501pglf] -s250
    setpll ics9lprs501pglf 200.209.218.227.236.245.250
    NOTE: You'll need to transcribe the above frequency details from setfsb into ics9lprs501pglf.lut if using setpll.

    Feel free to also copy any of my OC related posts into some HDX9000 definitive CPU/BCLK overclocking guide to concentrate the info into the most simplistic yet potent form. If wanting to ack just do so as a footnote.

    @ nando,

    setpll doesn't work. It just hard crashes the system.

    I even tried it with this line removed: set local1=0xF2D1051C007D2710

    Can't figure it out.

    Suggest do a new setpll PLL definition file:

    1. Boot without any overclocking.

    2. Go to setfsb, click "set" to set default 200Mhz FSB. Some PLL bytes will change.

    3. Run 'setpll -read 0x16' to create new mypll.lut with some info pre-filled for you.

    4. Rename mypll.lut to hdx9000.lut

    5. Edit hdx9000.lut. In the [INIT] section substitute in [BCLK_B1]..[BCLK_B5] for the bytes that change when do a BCLK change in setfsb. In [PCIE] section, add table entries containing values for [BCLK], [BCLK_B1],..,[BCLK_B5] that represent the data for that BCLK. Further details are in hdx9000.lut.

    6. Now can do a test run with 'setpll hdx9000 [BCLK] -preview'. Review the setpll.rw file to see what data would be sent to the PLL. This should match what setfsb's Diagnosis window has for that BCLK.

    7. Now can do 'setpll hdx9000 [BCLK] to set your BCLK.

    Finally, change local0 in hdx9000.lut's [INIT] section, byte 4 from 0xFF to 0xFC to disable spread spectrum on the PLL and test to see if that provides more overclockability. The resultant hdx9000.lut from your efforts can be shared with the community.

    It's crash city over here:

    Here's what I did for 209mhz:

    Code:
    [INIT] 
    set bytecount=0x10 
    set Local0=0x4505FC77FFF09091 
    set Local1=0x10277D001C05D1F2 
    set Local2=0x23A8F4F22303[BCLK_B1][BCLK_B2]
    set Local3=0x8475750DFFFFFFFF
    :: If USING [PCIE] then set a default PCIE clock if no commandline pcie   
    :: parameter is given. Default to 100Mhz                                  
    :: if "%pcie%"=="" set pcie=100                                         
                                                                              
    ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: 
    :: [BCLK] Move setfsb's FSB slider noting the BCLK and the data bytes     
    :: setFSB when move the slider noting the BCLK they correspond to.        
    ::[BCLK], [BCLK_B1], [BCLK_B2], [BCLK_B3], [BCLK_B4], [BCLK_B5]           
    ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::: 
    [BCLK] 
    200, B6, B0
    209, B6, C4
    218, B6, D7
    227, B6, EA
    236, B6, FE
    245, 76, 11
    250, 76, 1C 
    Based off this:

    [​IMG]

    Preview file output results:

    Code:
    :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::    
    :: Only commands starting with ">" are applied by r-w everything    
    ::                                                                   
    :: Eg: setFSB diagnosis window OR r-w "Clock Generator" appearance  
    :: 00 01 02 03 04 05 06 07   08 09 0A 0B 0C 0D 0E 0F                
    :: 10 11 12 13 14 15 15 17   [      local3         ]                
    :: [      local4         ]   [      local5         ]                
    ::                                                                  
    :: Converts to be in the following format below                     
    ::                                                                  
    :: set local0=0x0001020304050607                                    
    :: set local1=0x08090A0B0C0D0E0F                                    
    :: set local2=0x1011121314151617                                    
    :: set local3=..................                                    
    :: set local4=..................                                    
    :: set local5=..................                                    
    ::                                                                  
    :: NOTE: it is possible to read the PLL, perform bitwise changes    
    :: then write to the PLL but I found it freezes my system. Instead  
    :: I do a single hardcoded write which is very fast and reliable.   
    ::                                                                  
    :: If this freezes your box, then confirm that rw can read your PLL 
    :: by running 'rw\rw', command, 'smbus read block 0xd2 0 [bytecount]' 
    ::                                                                  
    :: To run this script type "rw\Rw /Command=setpll.rw", adding       
    :: "/Logfile=Log.txt" if want to see what rw does in detail         
    ::                                                                 
    ::**************************************************************   
    ::* mypll :  209Mhz **** 
    ::**************************************************************   
    >local0=0x4505FC77FFF09091   
    >local1=0x10277D001C05D1F2  
    >local2=0x23A8F4F22303B6C4 
    >local3=0x8475750DFFFFFFFF 
    :: Reverse bytes so they are in rw's big endian format  
    >local0=bswap(local0) 
    >local1=bswap(local1) 
    >local2=bswap(local2) 
    >local3=bswap(local3) 
    >smbus write block 0xd2 0 0x10        
    >rwexit                                       
    
    Also tried it with local 3 with "0"s instead of "F"s.

    Change "set bytecount=0x10" in myPLL.lut [INIT] section to be "set bytecount=0x16" and try again which I believe should work. The terminating 0xFF in local3 might need to be changed to 0x00 - not 100% sure on that one. Recommend you rename myPLL.lut to hdx9000.lut as otherwise would be re-written if you do a 'setpll -read 0x16'.

    Suggest filling the table with entries from 200 -> point of instability in 1Mhz increments AND add the pci-e clocks for a complete HDX9000.lut definition file. The pci-e clocks being a way to get more peformance if doing a DIY ViDock. It's a little bit of effort but very useful to the NBR HDX9000 community.
     
  4. slot

    slot Notebook Enthusiast

    Reputations:
    32
    Messages:
    47
    Likes Received:
    0
    Trophy Points:
    15
    Thanks Ricki.

    500 is not stable I worried
    Higher temperature than normal ....
    I have already changed the resolution of the previous image
    How to stop fixed configuration??
    Start Menu??
    After some minutes the values ​​change





    Thank you very much
     
  5. 2.0

    2.0 Former NBR Macro-Mod®

    Reputations:
    13,368
    Messages:
    7,741
    Likes Received:
    1,022
    Trophy Points:
    331
    @Nando4

    Using SetFSB
    This works like a charm to ramp up to 250mhz: setfsb -i9 -cg[ics9lpr501sglf] -s250
    And this works to ramp back down to 200mhz: setfsb -i9 -cg[ics9lpr501sglf] -s200

    I set this by creating a .bat file.

    Going to try using setpll next.

    Haven't tried PLL deviation yet.
     
  6. RickiBerlin

    RickiBerlin Notebook Deity

    Reputations:
    609
    Messages:
    979
    Likes Received:
    18
    Trophy Points:
    31
    New test results for 267.76 WHQL / NVPMManager / NVInsperctor:

    Test 1:

    NVPMM Profile on MedPerf, ie 0.9V.
    At "NVInspector" GPU frequency raised to 500MHz.
    DualIdaFixed CPU at 2.7 GHz. 1.0375 V, test system Win764bit HDD.

    Start now 3DMark06 / results, see here:

    CLICK TO ENLARGE

    Yes, this is no maximum number of points. But run perfectly stable at 0.9V.
    This is super! No image distortion, no artifacts.
    This may be different for each user. Hardware differences.
    For me GPU500MHz stable at 0.9V. = less heat, less noise, less WEAR

    Test 2:

    NVPMM Profile on MaxPerf, ie 1.0V.
    Raised the same time as "NVInspector" or "NVIDIA System Tools with ESA Support GPU frequency to 650MHz. For more details see the picture!
    DualIdaFixed CPU at 2.7 GHz. 1.0375 V, test system Win764bit HDD.

    Start now 3DMark06 / results, see here:

    CLICK TO ENLARGE

    This is my highest result ever with the HDX and T9300.
    No image distortion, no artifacts.
    I also believe "HDD or SSD" 3DMarko6 make no difference.

    --------------------------------------------------------------------------------------------------------------------------------------------------------------------

    For comparison, an old test and previous best score:

    Here again for direct comparison, an older 3DMark06 score.
    Test system was on SSD VistaHPx32 257.21 or higher.
    Powermizer "off" always on "RickiBerlin.
    Otherwise the same conditions.
    A little less points, but minimal difference.

    Start now 3DMark06 / results, see here:

    [​IMG]

    Click picture and you see the old post!
     
  7. RickiBerlin

    RickiBerlin Notebook Deity

    Reputations:
    609
    Messages:
    979
    Likes Received:
    18
    Trophy Points:
    31
    How could you see in my previous post, for me it is super stable.

    I have no higher temperature. In my 3DMark06 run at 500MHz at 0.9V, I had a maximum temperature of 53 C. (fan level 2, run without HWMonitor on test system)
    Please do not use this level as idle level! This is completely unnecessary!
    I am currently using 275MHz at 0.9V as a working stage. That's enough for almost any task.
    Idle temperature in 43 C. Fan at idle.


    Thank you! :)

    You have to create shortcuts!
    Use NVInspector for it!
    Go there to "Create Shortcut Clocks"!
    Now you've got every single shortcut, create for each NVPManager stage!

    All 12 profiles were. (4 frequency levels, each selectable performance levels)
    All profiles are not necessity. Performance Level 0 and 1 are still the same. (0.9V)
    Performance stage number 2, there is not.
    And performance level number 3 is then 1.0V.
    Ie 12 Shortcats without overclock profiles.


    Use my pictures from this post as an aid.

    ??????????????

    I do not understand. If Powermizer "off", then the GPU can not perform a frequency change.
    My guess:
    1.The driver 197.45. I have never written, but this is not a real notebook driver.
    See here on the "Nvidia page! (not Microsoft!)

    Please try the driver 257.21. Makes no problems. Download 32bit / 64bit
    2. It is also possible that it's not without "NVIDIA System Tools with ESA support" work. This test I can not, I have it on every system.

    As I said, these are just guesses! I think more about it.
    If something occurs to me, then I write.

    I like to help.




    EDIT:

    I forgot my "NVIDIA System Tools with ESA Support" presumption justified.

    If you install this, then a release for "manual frequency change" are confirmed.
    Previously, no change possible.

    If you have not installed, then you might be missing the lock-out.

    Do you hate tried to NVPMM "Heat Over Slow Down Over Ride" to use?
    If not, then try to your luck!
    Put it up on "Disable Heat Over Slow Down"!
    Perhaps this corresponds to the release as "NVIDIA System Tools with ESA Support.

    I had already set this way also. For me there was no change.

    All just guesses!
     
  8. RickiBerlin

    RickiBerlin Notebook Deity

    Reputations:
    609
    Messages:
    979
    Likes Received:
    18
    Trophy Points:
    31
    That would 2x666MHz more than now with the T9300dualIDAFixed. :)
    Very nice. I hope so much that your work is rewarded.

    Text in red, I can not translate, unfortunately. Especially as a question mark at the end of a sentence. I've translated each word individually, but the point remains hidden from me.

    One more request:

    Unfortunately I did not get a response to this post of mine. Maybe you can give a statement on F.40E.
    No, not today. Now working on FSB on!

    Or can someone else confirm the same behavior for F.40E?
     
  9. RickiBerlin

    RickiBerlin Notebook Deity

    Reputations:
    609
    Messages:
    979
    Likes Received:
    18
    Trophy Points:
    31
  10. slot

    slot Notebook Enthusiast

    Reputations:
    32
    Messages:
    47
    Likes Received:
    0
    Trophy Points:
    15

    If Ricki, that's right.
    Now if we have understood.

    Thanks and best regards
     
Loading...

Share This Page