Swiftech
Page 1 of 2 12 LastLast
Results 1 to 25 of 34

Thread: Phenom 9500 @ 2.8 in M2n-SLI deluxe.

  1. #1
    Registered User
    Join Date
    Jan 2008
    Posts
    36

    Talking Phenom 9500 @ 2.8 in M2n-SLI deluxe.

    First of all let me say hi to all of you guys. I love this forum I been using it for years and I finally have a good reason sign up and post.

    Well here is the story of my way to 2.8 with a phenom 9500 its abit long but I assure you its worth it..

    The First Phenom I got from newegg was nightmare I was totally frustrated I could not get the thing Stable at all, even at its native speed of 2.2Ghz. And I was blaming it all on the M2N-sli Deluxe and its buggy Bios. After Looking at every forums on the web I could find and 4 sleepless nights this are the essential things I found as common issues with phenoms.
    -ASUS AM2 mobos have to be set with Cool and Quiet On because if not you will always run at 5.5x or 5.0x Multi
    -A lot of people were hitting Brick walls that didn’t make much sense and I was one of them. For example. I could get 230Mhz in my FSB and boot everything fine with at 5.5 Multi. but if I went to 231Mhz or higher I could not ever post.(I’ll explain later more about this the reason why it happens)
    .
    So after trying every setting I could I was unable to run my processor stable at 2.2 much less 2.3 or 2.4. so I stopped and went back to the basics, I wanted to know if the problem really was the Mobo, Bios or Processor, since I got all of them at the time I needed to know which one to return or exchange. all I saw was random reboots and weird behaviors. After overclocking PCs for the past 8 years you can tell by the kind of blue screen or by the crashing behavior what is component that is unstable in your system, all you Veterans out there know what im talking about . I know when you hit a limit but at 230Mhz 5.5Multi being able to run full benchmarks and going up to 1Mhz to 231 and not being able to post makes no sense because ass many of you know 1mhz on the FSB from full stable you can go to unstable or even windows fail to boot but never a no post. I know the RAM is good... the processor is its100% stable at 5.5 x 230 but not when running at 11x 200… I wanted to Test each core but I couldn’t use AMD OverDrive, Im running Nvidia Chipset and Im using Vista 64bit, so AMD OverDrive wont work. It works with Nvidia when it 2.09 or less but Vista x64 compatibility was put in with 2.0.10 and higher but it doesnt with nvidia chipsets, sooo... I decided to Hack the damn thing to work for me, and I was able to, hehe.

    note: I Posted all this in the AMD official forums and how to hack AMD overdrive but they Deleted my post as soon as they saw the hack. haha.
    Ok here is how you hack it.

    1. Install AOD 2.0.14 Beta then go to "C:\Program Files (x86)\AMD\AMD OverDrive\amd64" and Copy AODDrive.sys to your desktop or some other place.
    2. uninstall AOD 2.0.14 and intall AOD 2.0.9 and replace the AODDrive.sys with the one on your desktop.

    Now you can use AMD OverDrive in Vista x64 But nothing Works except for the Muliti Changer, but who cares... that’s all I needed.

    So I went head and started the experimentation. I started moving the Multi from 5.5 to 11 in all cores. Core 0 passed. no problems, Core 1 passed, no problems, Core 2 reached 11Multi and WHAM! BSOD. oh s**t... I... I got it!. rebooted the PC moved core 0,1 and 3 to 11Multi and everything passed and worked great I slowly moved core 2 to as high as I could before BSOD and I could only get it to 2.05Ghz stable. so I got a bad core. I returned the processor and got a new one… and guess what? after 5 minutes of OCing I have 255Mhz FSB 11 Mulit at 2.8 Ghz. The thing in M2N-SLi Deluxes mobos CPU-Z does not agree with the Bios with what’s the actual Voltage given to the Processor. I think CPU-z is right because I am using the same cooler I had with my old 65nm 4300+ X2 @ 2.85Ghz 1.5Vcore and I had 52C full load, but with the quad core at 2.8Ghz full load on all cores I get 54C and CPU-Z says 1.39 and the Bios says 1.525v so there is no chance in hell that I would get those temps with 1.525Vcore. so CPU-z is right.




    now here is what I learned from the brief life of this same post at the AMD official forums.

    1.Everyone that could not OC past 2.4 or 2.5Ghz or had BSOD problems reported that their 3rd core AKA core 2 was the problem (im guessing production problem thus the reason for the Phenom x3, I bet that when the x3 shows up you’ll see core 0,1 and 3 but no core 2 ). People who could not get past 2.5Ghz started running at 2.7 on all cores and 2.5 on core 2.

    2. the 230Mhz Brick wall was created by the Bad 3rd core, people also reported that after getting the a new processor they could OC their FSB higher even if they could not get much higher on the Processor Speed. it makes sense since Memory controller is in the Processor and since they all share even the L3 Cache if on core has problem of fails it takes all them down with it.

    I hope you all got some ideas on how to fix or better tweak your processor.
    all this problems with BSOD, FSB and Processor OC are common in both AM2 and AM2+ socket. I hope the B3 stepping helps this issue since I love my Phenom at 2.8ghz it kick@$$.

    Final Settings in Bios
    Cool and Quiet OFF
    255Mhz FSB
    Mem 2.2Volts
    Chipset 1.5V
    CPU 1.525V
    5x HT Multi
    1.4V On HT/CPU connection


    I been waiting to post this For a long time now but I had to wait for the admins to enable my account to post or something like that.

    If you guys want to see some Specific Bench Marks to see how it compare to your AM2+ system let me know what to run cuz Im curious too

  2. #2
    Registered User
    Join Date
    Jul 2006
    Posts
    16
    Thanks for this post. I have the same motherboard and I was wondering if it could do such an overclock. What stepping is your Phenom?

  3. #3
    Xtreme Cruncher informal's Avatar
    Join Date
    Jun 2006
    Posts
    6,170
    Thanks dcfan60!Your post is very informative and appreciated.
    And great results with your second 9500

  4. #4
    Xtreme Enthusiast LIKMARK's Avatar
    Join Date
    Feb 2007
    Location
    Norway
    Posts
    704
    That's what I call a proper first post
    I'm gonna test your AMD Overdrive "hack" for sure.

  5. #5
    Fused tictac's Avatar
    Join Date
    Dec 2003
    Location
    Malaysia
    Posts
    2,789
    So how do you solved core2 issue? RMA?

  6. #6
    Xtreme Mentor
    Join Date
    May 2007
    Posts
    2,799
    Quote Originally Posted by dcfan60 View Post
    If you guys want to see some Specific Bench Marks to see how it compare to your AM2+ system let me know what to run cuz Im curious too
    Nice. Have you ran any SuperPi 1M and wprime 32M runs at 2.7G and 2.8G? Can you do that please.
    Have you validated any max CPUZ for all cores?
    What's max you've benched something like Cinbench10 at and maximum 8hr P95 stable?

    IDK about your voltage, but what happens if you check with EVEREST Ultimate. What Vcore does it give?

    1.392V is usually 1.4VID if actual.

  7. #7
    Xtreme Enthusiast
    Join Date
    Aug 2002
    Location
    London,Uk
    Posts
    966
    can you run bandwidth shots, pref with sisoft though i can install something else to compare. AOD says memory is single channel, only getting around 5000mb/s bandwidth and the comparitive 9700 listed in sisoft gets almost 10k. So i'm not sure if i'm really running in dual channel, or something else is wrong. memory came out of another system that managed to kill 3 sticks of mem so possible something is wrong with the remaining two.

    but i'm not sure what kind of numbers i should be getting, the 5k is at 200Mhz htt, and 800Mhz mem , 4,4,4,15 for now, according to sisoft the 9700 is running pc6400 aswell so i get the feeling i am running single channel, though i can't find a single option that would switch it to dual channel in the bios, gigabyte ds5 board.

    pain in the ass board so far, maybe a duff core, who knows. i can get up to 2.5Ghz pretty stable, i can briefly run at 2.7Ghz, had a few 3dmark runs at 2.75Ghz on all cores but zero stability there in real gaming.
    Mail Me | 3500+ , dfi sli-dr, g-skill la, 2x6800gt, 600w pcz, stacker case, air cooled

  8. #8
    Registered User
    Join Date
    Jan 2008
    Posts
    36
    CPU-Z Validation
    http://valid.x86-secret.com/show_oc.php?id=308022


    Prime 95


    I'll run as many Benchmarks as I can. it just takes some time.

    One things. do you guys see the Vcore reported by CPU-Z? its say 1.328... yeah as soon as I started runnig a Prime it drops Straight to 1.34 from 1.39 and the it fluctuated from 1.34 to 1.318 and all setting are the same as before.



    My memory is actualy runnig at 678Mhz (I know it crappy but its 667Mhz Ram and it hapes beeing OCed more than 750)
    HT is at 1275Mhz
    5-5-4-15 for the Ram and its @ 2T
    4x 1GB for a total of 4GB

    Its a OK score.... I think(hope)
    Ones I buy some new RAM 800Mhz 2 Sticks of 2GB for a total of 4GBs it will be way better specialy since Ill be runnig at 1T
    Last edited by dcfan60; 02-04-2008 at 09:29 AM. Reason: added Bandwith Benchmark

  9. #9
    Registered User
    Join Date
    Jan 2008
    Posts
    36
    Quote Originally Posted by tictac View Post
    So how do you solved core2 issue? RMA?
    RMA was the only way. I was able to get 2.4Ghz on Core 0,1,3 and 2.05 in core 2.

  10. #10
    Registered User
    Join Date
    Sep 2007
    Posts
    58
    what NB speed do you have now?

  11. #11
    Xtreme Cruncher
    Join Date
    Jan 2007
    Location
    Massachusetts
    Posts
    715
    DCfan60, would you please post all the codes from the IHS (or just a picture of the IHS, if that's easier for you)?

  12. #12
    Xtreme Enthusiast
    Join Date
    Aug 2002
    Location
    London,Uk
    Posts
    966
    don't suppose you could redo bandwidth test at stock fsb, and stock fsb.

    i'm still only getting 5.3gb/s ish, with mem at 800mhz, 4,4,4,15. not used sisoft in a good few years but i seem to recall it scaling with speed, though when i bump just cpu speed by 300Mhz i get same results, i get the feeling it is just stuck in single channel mode but can't seem to get it to run dual channel.

    does AOD or cpu-z actually show your mem is in dual channel at all?

    i seem to be having issues getting an htt over stock, possibly just some settings i need to mess with, not used a amd board for quite a long time, or maybe this particular board/chip/chipset needs a big bump in voltage to pass a fsb barrier, who knows. possible need to drop northbridge and HT speed hugely and remove that as a limit, seem to be able to run 2.6/2.7Ghz on stock voltage pretty much fine with HT at 200Mhz just upping the multiplier. might just be this bios doesn't agree with this board.

    EDIT:-

    just ran prime for ya to check voltages, running stock which seems to be 1.25v in bios gives me 1.28v in reality, shown in cpu-z and other apps. under load on prime, superpi it actually goes up to 1.296v, its basically VERY solid with voltage.
    i have no idea what ASUS's problems are but they always seem to have vdroop issues. everyone es about it on every revision of every board they make. my 965, p35, old amd and older intel boards all seem to have it. my asus 965 died and dropped in a gigabyte p35 and with that vdroop dissappeared, it had crappy bios settings for voltage, IE certain settings gave way less than listed, 1.75v would only give me 1.65v , but that was from boot to windows to under highest load, mislabeled bios settings and vdroop are very different things.
    Last edited by drunkenmaster; 02-04-2008 at 03:00 PM.
    Mail Me | 3500+ , dfi sli-dr, g-skill la, 2x6800gt, 600w pcz, stacker case, air cooled

  13. #13
    Registered User
    Join Date
    Jan 2008
    Posts
    36
    Quote Originally Posted by drunkenmaster View Post
    don't suppose you could redo bandwidth test at stock fsb, and stock fsb.

    i'm still only getting 5.3gb/s ish, with mem at 800mhz, 4,4,4,15. not used sisoft in a good few years but i seem to recall it scaling with speed, though when i bump just cpu speed by 300Mhz i get same results, i get the feeling it is just stuck in single channel mode but can't seem to get it to run dual channel.

    does AOD or cpu-z actually show your mem is in dual channel at all?

    Sure thing. I ran it againt I got 8153 this time. I think you are stuck with single channel.

    3 things I have personaly seen create this.

    1.-Odd number 1 or 3 sticks of Memory, it must be even numbers 2 or 4 sticks
    Make sure your computer sees all your RAM, you can have a bad stick or might need to re-seat them.
    2.-Back in the DDR1 days if you Mixed Single sided with double sided Ram there was a chance that it would not see all the RAM or it would not even work, but as far as I know DDR2 is all double sided. Im mixing different Brands of DDR2 in this PC, they differ in timings, Speed and even Voltage but they all work fine.
    3.-Its your Bios. I personaly had not seen this untill I started flashing bioses with the crappy Built-in utilities that are in the Bios like ASUS EZ flash. For some weird reason with some flashes it would be stuck at 64bit no matter what... and re-flashing the bios fixed it and I got my 128bit dual channel back. Also some Bioses have a setting that you can change that will allow it ot check if it can do Dual channel, if it was disabled by accident you will always have Single-Channel.

    Good luck m8t

  14. #14
    Registered User
    Join Date
    Jan 2008
    Posts
    36
    Quote Originally Posted by mrcmtl View Post
    what NB speed do you have now?
    2296Mhz on the NB.
    I read in some forum or some website ( I cant rember if it was here or not) that there was theory that the NB was the Bottle neck in the proccessor for Overclocking and performance, since it locked at 9x in the 9500 and 9600 and 9600BE.
    You can Change it in some of the High-end AMD 790 Mobos in the Bios, its a little tricky at first since you get your NB multi by dividing some two numbers... the New-er AMD OverDrive versions have that option but it does not work as of yet.
    Unfortunately I cant experiment with this to see if It's as the theory claims "a Performance and OC bottle neck" an none seem to report on it so Far.
    I think it is a Performance Bottle-neck but from my experience with the Bad 9500 phenom. I think the OC of the NB is affected by the OC or stability of all your CPUs. With all good cores I can reach 255Mhz stable compared to the 230Mhz.

    sorry for rambling on about NB subject, but I think your question was a good excuse for me to talk about it.

  15. #15
    Registered User
    Join Date
    Jan 2008
    Posts
    36
    Quote Originally Posted by MotF Bane View Post
    DCfan60, would you please post all the codes from the IHS (or just a picture of the IHS, if that's easier for you)?
    Damn, I was really hopping no one would ask me that, I knew I should have written them down.
    Since Im using some crappy watter cooling(big watter SE.. I know, I know... its bad but it works great for me and I l'll be building my own soon) its a pain the butt for me to check them. But let me tell you this, I compared it to the Ones in the BAD Proc with the messed up 3rd core, and all of them were exatly them Same, exept for the Serial number since that one is unique to every CPU.

    Dunno if that helps but if one night im drunk or bored enough i'll pull it out clean out the AC5 off the Proc and i'll write them down.

  16. #16
    Registered User
    Join Date
    Jan 2008
    Posts
    36
    Oh yeah, If you guys Try out the little AOD hack let me know if it worked for you... and if you figured out a way to make some of the other stuff work... even better.
    Last edited by dcfan60; 02-04-2008 at 04:22 PM.

  17. #17
    Xtreme Enthusiast
    Join Date
    Aug 2002
    Location
    London,Uk
    Posts
    966
    hehe, yeah i get this feeling i'm stuck in single channel aswell, infact, sisoft i just noticed, because it now(didn't years ago) changes the results bar to show diff information every 5 seconds, when it comes up with timings it mentions it thinks its 1x2gb ddr stick, when i'm running 2x1gb ddr in slot1 and 3 which should be fine. might try reflashing bios.

    i think my overclock isn't bad speed wise at least, hell dual channel will give me a bigger bandwidth boost than HTT speed for now so gotta get that working, only on the stock cooler with the 9600BE, which frankly, sucks :p

    trying to work out how to get tubing from external cooling routed well into new case so 2.7Ghz might not be bad, happy to provide it with some volts when it gets back on water

    this gigabyte with new bios lets me increase northbridge/htt speeds but so far just setting to 2Ghz from default 1.8Ghz i had to reset cmos completely, possible it didn't like that 10 multi or voltage or something so will try 11 multi maybe for 2.2Ghz. from what other threads say higher northy speed is possible, might help but wasn't clear. but l3 runs at same speed as northy so in theory the faster it is the better, as l3 works significantly slower than the rest of cache at default 1.8Ghz.
    Mail Me | 3500+ , dfi sli-dr, g-skill la, 2x6800gt, 600w pcz, stacker case, air cooled

  18. #18
    Xtreme Cruncher
    Join Date
    Jan 2007
    Location
    Massachusetts
    Posts
    715
    Quote Originally Posted by dcfan60 View Post
    But let me tell you this, I compared it to the Ones in the BAD Proc with the messed up 3rd core, and all of them were exatly them Same, exept for the Serial number since that one is unique to every CPU.
    That is extremely unfortunate.

    Thanks though.

  19. #19
    Xtreme Mentor
    Join Date
    May 2007
    Posts
    2,799
    Quote Originally Posted by drunkenmaster View Post
    can you run bandwidth shots, pref with sisoft though i can install something else to compare. AOD says memory is single channel, only getting around 5000mb/s bandwidth and the comparitive 9700 listed in sisoft gets almost 10k.
    Latest Sandra should show around 9k for stock 800 4-4-4-12 2T Phenom and AOD will wrongly report single channel but Memset, EVEREST, CPUZ won't.

    Yes, stock Phenoms at 1066 5-5-5-15 will get plus 10k in Sandra.

    How many hours have you ran small FFT stable and your temps? That shows just starting. I'd like to know what is stable mainly, upto 2.8-3.2G ss/benching is found on most Phenoms.

    I'll run as many Benchmarks as I can. it just takes some time.
    Super Pi Mod 1.5 XS 1M is a 30s bench at max and wPrime 1.55 32M is less than 15s. Run these when you can please.

    One things. do you guys see the Vcore reported by CPU-Z? its say 1.328... yeah as soon as I started runnig a Prime it drops Straight to 1.34 from 1.39 and the it fluctuated from 1.34 to 1.318 and all setting are the same as before.
    That's how it's supposed to be, although that droop is high. You're at 1.392V but the load vdroop kicks in to put it low. Near to 1.37V would be much better.

    http://i266.photobucket.com/albums/i...0/Bandwith.jpg

    My memory is actualy runnig at 678Mhz (I know it crappy but its 667Mhz Ram and it hapes beeing OCed more than 750)
    HT is at 1275Mhz
    5-5-4-15 for the Ram and its @ 2T
    4x 1GB for a total of 4GB

    Its a OK score.... I think(hope)
    Ditto.

    Quote Originally Posted by dcfan60 View Post
    2296Mhz on the NB.
    I read in some forum or some website ( I cant rember if it was here or not) that there was theory that the NB was the Bottle neck in the proccessor for Overclocking and performance, since it locked at 9x in the 9500 and 9600 and 9600BE.
    It's not a theory, it's fact anyone with Phenom will tell you. Drop the NB multi to sub 8x and your CPU may well get to 3.0G the way it looks.

    And 9600BE doesn't have NB multi locked upwards, 9500/9600 do.

    Unfortunately I cant experiment with this to see if It's as the theory claims "a Performance and OC bottle neck" an none seem to report on it so Far.
    It's already well established NB increases perf. and memory performance especially. Run a Pi 32M at high MHz with your NB speed/RAM speed and I'll run one at the same MHz/settings for all but increase NB speed. Watch the gain.

    Spend many weeks on Phenom before concluding anything. It's an odd CPU and a new architecture, not K8 or C2.

  20. #20
    Registered User
    Join Date
    Sep 2007
    Posts
    58
    hey KTE, do you know any way of changing NB multi on an AM2 board?

  21. #21
    Xtreme Enthusiast
    Join Date
    Aug 2002
    Location
    London,Uk
    Posts
    966
    yup, looks like my problem was slotting of mem. did it change at some point on amd systems where slots next to each other became the norm for dual channel?

    i did what i do in every system for dual channel for the past 5 years, one in slot 1 and one in slot 3, this is normally first slot for each channel. infact to rule it out i checked the manual which also said dual channel mem config for 2 dimms is slot 1/3 or slot 2/4, i switched mem so its in slot 1/2 now and boom 9.5k bandwidth. I tried it out of frustration and also wondering if mem/slot was borked, freaking manual has it wrong.

    now to see if that bandwidth gives me anything in terms of performance that makes keeping this 9600 phenom seem worth it :p

    EDIT:- AOD is reporting dual channel aswell now, still can't see anything cept mem size in cpu-z though.
    Mail Me | 3500+ , dfi sli-dr, g-skill la, 2x6800gt, 600w pcz, stacker case, air cooled

  22. #22
    Registered User
    Join Date
    Jan 2008
    Posts
    36
    Quote Originally Posted by MotF Bane View Post
    That is extremely unfortunate.

    Thanks though.
    Why is that? were you hoping for a new Core that would OC well?.
    I thinks is a buggy deal with the current B2 ore the manufacturing process. Im not saying it cuz of the Errata things but I thinks there is alot of hit and miss involved with it, and core 2 problems is the effect of those problems.

  23. #23
    Xtreme Cruncher
    Join Date
    Jan 2007
    Location
    Massachusetts
    Posts
    715
    Quote Originally Posted by dcfan60 View Post
    Why is that? were you hoping for a new Core that would OC well?.
    I thinks is a buggy deal with the current B2 ore the manufacturing process. Im not saying it cuz of the Errata things but I thinks there is alot of hit and miss involved with it, and core 2 problems is the effect of those problems.
    More or less, or at least for it to be an unusually good batch.

  24. #24
    Xtreme Enthusiast
    Join Date
    Aug 2002
    Location
    London,Uk
    Posts
    966
    hmm, next to nothing in terms of boost in performance from dual channel and about a 90% bandwidth increase. 3dmark06 scores were almost identical though i guess if i'm cpu limited that might not have shown as much of a difference, guess i'll rerun bench at high res/aa/af and see how that compares to before.
    Mail Me | 3500+ , dfi sli-dr, g-skill la, 2x6800gt, 600w pcz, stacker case, air cooled

  25. #25
    Registered User
    Join Date
    Jan 2008
    Posts
    36
    Quote Originally Posted by drunkenmaster View Post
    hmm, next to nothing in terms of boost in performance from dual channel and about a 90% bandwidth increase. 3dmark06 scores were almost identical though i guess if i'm cpu limited that might not have shown as much of a difference, guess i'll rerun bench at high res/aa/af and see how that compares to before.

    3D Mark Score means nothing.
    Play some real games. specialy new ones and youll see and feel the difference. I if remever corrently dual-channel can yield up to 30% more FPS in some games.

Page 1 of 2 12 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •