ASUS A8R32-MVP Deluxe BIOS Overclocking Help

Discussion in 'Processors and motherboards AMD' started by Psychlone, Apr 10, 2007.

  1. Psychlone

    Psychlone Ancient Guru

    Messages:
    3,688
    Likes Received:
    1
    GPU:
    Radeon HD5970 Engineering
    This was posted by request of a member that just got this board.
    This guide is to be used as a template *only* - DO NOT just copy down these settings and try them as your own.
    Overclocking demands a great amount of responsibility, and I will not accept responsibility for ruining anyone's hardware because of just using the settings below and not doing the footwork yourself.

    Overclock Responsibly!

    ADVANCED
    _Jumper Free Configuration
    _AI Overclocking__________________Manual
    _CPU Frequency___________________356
    _PCIE Frequency__________________100

    _FID/VID Change__________________Manual
    _Processor Frequency Multiple_______X9
    _Processor Voltage________________1.375V
    _DDR Voltage____________________2.90V
    _VCORE Voltage__________________AUTO
    _Northbridge Overvolt_____________Enabled
    ___Core Voltage__________________1.5V
    ___HT Bus Voltage________________1.5V
    ___PCIE Voltage__________________1.2V

    _PEG Link Mode___________________Auto
    _PEG Buffer Length________________Auto

    _AI Clock SkewA__________________Auto
    _AI Clock SkewB__________________Auto

    _CPU Configuration
    _GART Error Reporting______________Disabled
    _MTR Mapping____________________Continuous
    _Cool N Quiet_____________________Disabled
    _HT Link Speed___________________800 MHz

    _Memory Configuration
    _Memory Timing Mode_____________Limit
    ___Memclock Value_______________166
    _MCT Timing Mode________________Manual
    ___CAS (CL)_____________________3
    ___TRAS________________________11
    ___TRP_________________________4
    ___TRCD________________________4
    ___TRRD________________________4T
    ___TRC_________________________10T
    ___TRFC________________________13T
    ___TRWT________________________5 clk
    _MCT Extra Timing Mode____________Manual
    ___TREF________________________15.6us
    ___DDR Input Strobe Skew_________Disabled
    ___Burst2Opt____________________Disabled
    _User Configuration Mode___________Manual
    ___Read Preamble________________9.0ns
    ___Asyc Latency_________________10.0ns
    ___CMD-ADDR Timing_____________1T

    _Bank Interleaving_________________Auto
    _Burst Length____________________4 Beats
    _Software Memory Hole____________Disabled

    _ECC Configuration
    ___DRAM ECC Enable______________Disabled
    ___L2 Cache BG Scrub_____________Disabled
    ___Data Cache BG Scrub___________Disabled

    _Chipset
    ___VGA Priority___________________1st PCIE ->2nd PCIE

    _HT Config
    ___HT Link Speed_________________800 MHz
    ___CPU:RD580-NB HT Link Width____16 Bit

    ___RD580 HT Link Three State_______Disabled
    ___RD580 HT Drive Strength_________AUTO
    ___HT Receiver Comp. Ctrl__________Optimal

    ___RD580 HT PLL Control___________High Speed

    _PCI-E Configuration
    ___PCIE GFX1 Link Width___________X16
    ___PCIE GFX2 Link Width___________X16

    ___P2P Writes Between GFX Ports___Disabled
    ___10% Extra Current Between GFX__Enabled
    ___10% Extra Current Between GPP__Enabled
    ___10% Extra Current Between SB___Enabled
    ___10% Extra Current Between GFX1_Enabled
    ___10% Extra Current Between GFX2_Disabled


    (CPU Multiplier * HTT Frequency) / (CPU Multiplier/(Memory Multiplier)) = Memory Frequency

    (9 X 356 = 3200) / (9 / .83333) = 290.90MHz (582MHz DDR)

    This nets me 3200MHz (3.2GHz) at 30*C idle and 48*C full load (using the Thermaltake Big-Typhoon)

    I've increased all my latencies until I was beyond the RAM's specs (+ 3-4 steps EACH) then reduced one at a time until I ended up here, and netted a KILLER increase in bandwidth in RAM....it's NOT all about the tightest timings you can get...it's about the tightest timings you can achieve AFTER you've overclocked everything else!

    This is all backed up by a 24 hour burn-in using Orthos Stressprime at priority 10, and an 8 hour burn-in with Memtest86.

    Using Everest Ultimate Edition RAM Benchmarks, I've increased my read, write, copy and latency by DRAMATIC amounts!!!

    READ: 8110MB/s
    WRITE: 8246MB/s
    COPY: 6510MB/s
    LATENCY: 48.3ns
    (these are on par or better than DDR2 800 modules! - Check them yourself!!!)

    Oh, and yep...my HT is running at 1425MHz... I think I got a little lucky with my mobo, as it's been 24 hour Orthos Stressprime stable at a little higher than this!

    So, all in all, these settings have already helped loads of people with this board...I did alot of footwork here - countless hours of rebooting/resetting CMOS/taking notes/etc.
     
  2. tnet

    tnet Master Guru

    Messages:
    278
    Likes Received:
    0
    GPU:
    HIS X1900XT ICEQ3
    okay noob here with this motherboard.

    -- specs --
    cpu: amd x2 4400+ toledo // stepping: JH-E6?
    memory: 2gb twinx2048-3200c2

    .. i've been trying to achieve 2.80GHz with this cpu .. question i got is: should i be overclocking the cpu first, changing the HTT and all that and having it stably boot up (leaving all my memory settings at default) first? then afterwords tweaking the memory settings?
     
    Last edited: Apr 11, 2007
  3. Psychlone

    Psychlone Ancient Guru

    Messages:
    3,688
    Likes Received:
    1
    GPU:
    Radeon HD5970 Engineering
    First of all, here's AMD's overclocking equation: (You're going to want to print all this out!)

    (CPU Multiplier * HTT) = CPU Frequency
    (CPU Multiplier / (memory divider) ) = Divisor Ratio (ALWAYS ROUND UP TO THE NEXT INTEGER!!!)
    (CPU Frequency / Divisor Ratio) = RAM MHz (* 2 = DDR MHz)
    (HTT * HT Driving Strength) = HT MHz

    Memory Dividers shown as decimals for the equation:

    200MHz (400) = 1/1 (1 / 1 = 1)
    183MHz (366) = 9/10 (9 / 10 = .9)
    166MHz (333) = 5/6 (5 / 6 = .83333)
    133MHz (266) = 2/3 (2 / 3 = .66666)
    100MHz (200) = 1/2 (1 / 2 = .5)

    HT Driving Strength info
    1000MHz = 5X
    800MHz = 4X
    600MHz = 3X
    400MHz = 2X

    Now then, here is what I do when beginning a new overclock:

    Drop the CPU Multiplier to it's lowest setting
    Drop the HT Link Speed to it's lowest setting
    Drop the RAM divider to it's lowest setting

    Begin by raising the HTT (CPU Frequency) in 10MHz increments - rebooting to POST in between each change - until it wont POST (Power On Self Test) - note the number as you've just found your motherboard's maximum HTT. Drop it down to a level decided on from the equation.

    Raise your CPU Multiplier (using the equation) to probably 10X, reboot and see if it will POST with the HTT from above...if not, reduce the HTT *OR* raise the voltage in .1V increments until it will, keeping in mind to keep an eye on the CPU temperature...DO NOT GO HIGHER THAN 1.55V!!!

    You've just found a *comfortable* level from which to work...now the hard part begins!

    Raise ALL your RAM latencies to 3-4 notches above stock - in your case, straight down the list from my BIOS above:
    3,13,6,6,6,12,13,5,15.6,9.5,12

    Begin by raising your DDR Voltage to 2.85 or 2.9 (don't go higher - the BIOS text goes red for a reason!)
    Now, starting with the SECOND one (tRAS), drop it one notch and reboot to POST, repeating until it will not pass the POST, clear the CMOS if you have to, and raise that one value one notch.
    Do the same for ALL the latencies (rebooting to POST in between EACH AND EVERY CHANGE) - leaving tREF at 15.6, Read Preamble at 9 and ASYC Latency at 10 (just don't go lower on those for now...)
    This is a *very* long and arduous process, but the benefit will be more than worth the effort you put into it!

    After you've found all the lowest latencies for the *comfortable* overclock, you can begin to see if there's any more...
    By raising the CPU voltage, (NEVER HIGHER THAN 1.55) you can increase stability in your overclock - but it comes at a great price - HEAT. If your CPU gets above 50*C under full load (should be in the 30-35*C range for idle - but load temps are the most important!), you need better cooling, or need to settle on an overclock that utilizes less voltage (which means less HTT and/or higher multiplier and maybe even a different divider)
    If your temps are good (~50*C or below at FULL LOAD) then you can start experimenting with the Overvolt option. By enabling the +100mV option, you're effectively giving your CPU .078V more (ASUS has a bug in their overvolt - it's not really .1V as measured with a multimeter, it's more like .078V and the +200mV is more like .174V instead of .2V) - So, you can see that if the mobo tops out at 1.5V for your CPU (ASUS voltage regulator - it's a good thing for people that don't know what they're doing!) - by enabling the +100mV Overvolt, you're effectively giving the CPU ~1.58V, and with the +200mV, it would be roughly 1.67V - Again, WATCH YOUR TEMPS!!!

    Now, another thing you'll have to test is your HT link speed...most motherboards aren't capable of much over 1000MHz (1000MHz in EACH lane - 1000 up and 1000 down) - BUT, I have proven the A8R32-MVP Deluxe stable at over 1500MHz (1500up AND 1500down) - there isn't a motherboard out there aside from ours that is capable of that! - But, they all differ...I've seen some that struggle to reach 1100MHz, and another that is barely unstable at 1500MHz - and everything in between.

    You'll want to use the equation above to find out where you *should* be, and then use the equation to see if you can go the next step higher (i.e. 3X to 4X to 5X) - you'll know when it's not stable at that speed when it BSODs. - For now, try to keep it at around 1000MHz and after you're comfortable with testing further, try it.
    The secret for getting a higher MHz throughput out of the HT is because of the chipset and the options for it.
    I enable the Overvolt to 1.5V for ALL the Northbridge options, and enable the Overvolt option for the Southbridge as well...it helps in the stability for higher bandwidth - But you've GOT to pay attention to your motherboard temp now! Mine's sitting at 35*C right now, and doesn't really get much higher - this is due to adequate case cooling, which I suggest you definitely follow up on!!!

    So, now that I've taken up ALOT of room with these volumes of information, let's play math:

    (10 * 280) = 2800
    (10 / (9 / 10 = .9)) = 11.11 (ROUND UP so = 12)
    (2800 / 12) = 233.33 (* 2 = 466.66MHz DDR)
    (4X * 280) = 1120MHz HT

    The only variable that I can see using the above equation (aside from your CPU's stepping which *may* not let you reach 2.8GHz) is the RAM's settings - remember and start with VERY loose timings, and drop them one at a time until it won't pass the POST, then raise the culprit setting a notch and move on to the next (you're going to do this alot, so start getting used to the idea!)

    Failing that, here's another scenario using only a different memory divider:

    (10 * 280) = 2800
    (10 / (5 / 6 = .83333) ) = 12.000048 (You've GOTTA round up!!! - so = 13)
    (2800 / 13) = 215.38 (* 2 = 430.77MHz DDR)
    (4X * 280) = 1120MHz HT

    Now, if your computer isn't stable there, you've got to do some real fine-tweaking...raising voltages, loosening RAM, lowering your HT link speed, or ultimately, lowering your HTT (which is only going to reduce your overclock - but hey, at least it'll be stable!)

    Ok, now that I've completely overloaded you, go print this out and give it all a shot. I expect you to take notes of what works and what doesn't (imagine writing a book from start to finish without scratch paper!!!) - seriously, the only reason I even have all this information is because I wrote all the math down...what worked, what settings pushed it over the edge, literally EVERYTHING...and it's lead directly into posts just like this one!

    Remember, all hardware is different...2 machines with identical parts (even identical stepping on CPUs, etc.) will not overclock the same. There is no guarantee. You paid for parts that will perform to their advertised specifications, and what you're doing by overclocking is getting *more* than you paid for (hence the 'no guarantees' part.)
    What may work for one individual may or may not work for the next...even given the same exact components...just always remember this...sometimes you have to settle for less than what you originally expected - the way to make it 'ok' in your head is to remember the 'no guarantees' thing...you've already got 'something' for 'nothing' if you've overclocked *at all*. ;)

    Good luck!

    Psychlone
     
  4. tnet

    tnet Master Guru

    Messages:
    278
    Likes Received:
    0
    GPU:
    HIS X1900XT ICEQ3
    okay here are my results.

    ADVANCED
    _Jumper Free Configuration
    _AI Overclocking__________________Manual
    _CPU Frequency___________________260
    _PCIE Frequency__________________100

    _FID/VID Change__________________Manual
    _Processor Frequency Multiple_______X10
    _Processor Voltage________________1.400V
    _DDR Voltage____________________2.75V
    _VCORE Voltage__________________AUTO
    _Northbridge Overvolt_____________Enabled
    ___Core Voltage__________________1.4V
    ___HT Bus Voltage________________1.4V
    ___PCIE Voltage__________________1.4V

    _PEG Link Mode___________________Auto
    _PEG Buffer Length________________Auto

    _AI Clock SkewA__________________Auto
    _AI Clock SkewB__________________Auto

    _CPU Configuration
    _GART Error Reporting______________Disabled
    _MTR Mapping____________________Continuous
    _Cool N Quiet_____________________Disabled
    _HT Link Speed___________________800 MHz

    _Memory Configuration
    _Memory Timing Mode_____________Limit
    ___Memclock Value_______________166
    _MCT Timing Mode________________Manual
    ___CAS (CL)_____________________2.5
    ___TRAS________________________5
    ___TRP_________________________2
    ___TRCD________________________3
    ___TRRD________________________2T
    ___TRC_________________________7T
    ___TRFC________________________11T
    ___TRWT________________________2 clk
    _MCT Extra Timing Mode____________Manual
    ___TREF________________________15.6us
    ___DDR Input Strobe Skew_________Disabled
    ___Burst2Opt____________________Disabled
    _User Configuration Mode___________Manual
    ___Read Preamble________________9.0ns
    ___Asyc Latency_________________10.0ns
    ___CMD-ADDR Timing_____________1T

    _Bank Interleaving_________________Auto
    _Burst Length____________________4 Beats
    _Software Memory Hole____________Disabled

    _ECC Configuration
    ___DRAM ECC Enable______________Disabled
    ___L2 Cache BG Scrub_____________Disabled
    ___Data Cache BG Scrub___________Disabled

    _Chipset
    ___VGA Priority___________________1st PCIE ->2nd PCIE

    _HT Config
    ___HT Link Speed_________________800 MHz
    ___CPU:RD580-NB HT Link Width____16 Bit

    ___RD580 HT Link Three State_______Disabled
    ___RD580 HT Drive Strength_________AUTO
    ___HT Receiver Comp. Ctrl__________Optimal

    ___RD580 HT PLL Control___________High Speed

    _PCI-E Configuration
    ___PCIE GFX1 Link Width___________X16
    ___PCIE GFX2 Link Width___________X16

    ___P2P Writes Between GFX Ports___Disabled
    ___10% Extra Current Between GFX__Enabled
    ___10% Extra Current Between GPP__Enabled
    ___10% Extra Current Between SB___Enabled
    ___10% Extra Current Between GFX1_Enabled
    ___10% Extra Current Between GFX2_Disabled

    .. here is what i found out .. i could achieve HTT of 265 however it wouldn't boot into windows. also anything above 265 just wouldn't post, so i guess no (280 HTT). i'm going to run some prime95 tests as well as memtest86+ and see how stable this is. as of right now my vcore voltage is 1.44v(ish) .. cpu is hanging at 30-32*C idle, motherboard is at 24-26*C.

    *update*
    Prime95: [Wed Apr 11 01:39:15 2007] (Ran for 0 Minutes)
    FATAL ERROR: Rounding was 0.5, expected less than 0.4
    Hardware failure detected, consult stress.txt file.

    okay currently as of right now how things are .. the computer isn't stable .. i'm getting random lockups and reboots =X
     
    Last edited: Apr 11, 2007

  5. Psychlone

    Psychlone Ancient Guru

    Messages:
    3,688
    Likes Received:
    1
    GPU:
    Radeon HD5970 Engineering
    Dude...you're not going to get anywhere with those RAM timings - you've got to loosen them up and start tweaking from there.
    Orthos (which is what you should be using) and Prime95 will error like that if it's a memory corruption (it effectively looses it's place) - which says to me that you've got alot of work to do on your memory overclock (you haven't done much of anything for the RAM yet, and without that, you're not going to be stable at where you want to be!)

    If you want to run the timings where they are right now, change your memory divider to 133 (266) instead, and your HT Link Speed to 600MHz - give that a try and post back.

    You're half done, but the hardest part is tuning your RAM to your system...this is where *most* people don't know what they're doing and/or just quit - leaving an unfinished overclock. You can see that I'm getting 582MHz out of my Corsair TwinX XMS3202 v1.3 RAM...it's not because I'm a God, or got really lucky, it's because I did the footwork myself and tweaked every latency [1 at a time] to my system.

    Giving your RAM more voltage (to 2.9V) can be beneficial to these sticks...they seem to run best at around 2.8-2.85V but I've had a couple that really liked 2.9V. That along with the properly tweaked latencies will get you where you want to be (providing the stepping of your CPU will allow it in the first place!)

    Good luck!

    Psychlone
     
    Last edited: Apr 11, 2007
  6. tnet

    tnet Master Guru

    Messages:
    278
    Likes Received:
    0
    GPU:
    HIS X1900XT ICEQ3
    okay i'm confused then. you stated to start out with 3,13,6,6,6,12,13,5,15.6,9.5,12.6,9.5,12. that is what i did. i then started with tRas, dropped it one notch at a time (12, then 11, then 10, etc.) until it could not POST. when it didn't post, I then raised the value one notch (ie: if it was 4, didn't post, i rebooted and raised it to 5). i did this for all the latencies leaving those that you said to leave, and what i posted was what i got.

    i don't exactly understand what this "loosen" up my timing means.

    btw .. when i am doing this procedure, what is my MEMCLOCK VALUE suppose to be?
     
    Last edited: Apr 11, 2007
  7. Psychlone

    Psychlone Ancient Guru

    Messages:
    3,688
    Likes Received:
    1
    GPU:
    Radeon HD5970 Engineering
    You're on the right track, but looking at the BIOS you posted above, it looks like you're running 2.5,5,2,3,2,7,11,2 and the rest look fine.

    I think that you're unstable because you haven't relaxed those specific timings with your processor already where you wanted it...that or you just dropped them all too far for stability at that speed.
    Raise them all back up to where mine are (from my *first* post here and see if that helps your stability. The CAS is definitley the first one to raise to 3...then the rest.

    Psychlone
     
  8. tnet

    tnet Master Guru

    Messages:
    278
    Likes Received:
    0
    GPU:
    HIS X1900XT ICEQ3
    those timings would be correct .. however those are the timings that i found would let me post and was the lowest, that is why i left it as is .. what memory frequency? 166MHz? and by "stability" .. after changing each item, should i be running memtest or going into windows and running orthos?
     
    Last edited: Apr 12, 2007
  9. Psychlone

    Psychlone Ancient Guru

    Messages:
    3,688
    Likes Received:
    1
    GPU:
    Radeon HD5970 Engineering
    I would run Orthos after changing a few, but not between every one...just keep track of your changes and if it shows unstable with Orthos after a few minutes (10-12) then go back and change the ones up a notch.
    Memtest86 should be run for a complete cycle before you get too far into it - I suspect a faulty stick because of the error you were experiencing, but could be wrong.

    Psychlone
     
  10. tnet

    tnet Master Guru

    Messages:
    278
    Likes Received:
    0
    GPU:
    HIS X1900XT ICEQ3
    well i finally am back to testing out some stuff. i've restarted my overclocking and what not, and figured to just find the ideal cpu HTT and Multiplier would be. so far i've found that 260 HTT with a 10x Multiplier is the highest this computer will go. with that said, in order to achieve that however here is what were the settings.

    1.40V Processor Voltage
    +200mV VCORE Voltage

    i ran orthos just running the CPU Stress test to see if these settings would error. i found that they would not however temperatures were a little tad bit to high (at least i think so) to have ran it even longer to see if it would of crapped out. i ran orthos for about 5 minutes each.

    my results were this

    @1000MHz (5x HT) - 45-50*C/IDLE & 70*C/LOAD
    @800MHz (4x HT) - 35-50*C/IDLE & 65*C/LOAD

    .. i'm guessing running at those temperatures is to risky so i'm going to have to guess that i'll have to lower the HTT :(

    my heatsink is also a zalman CNPS9500 LED.

    i've tried multiple times as well to lower the Processor Voltage with +200mV as well as 1.40V with +100mV .. they all would fail within 1 minute of orthos.

    other info, Northbridge Overvolt and Southbridge Overvolt were Enabled with Northbridge Core Voltage 1.50V, HT Bus Voltage 1.50V, and PCIE Voltage 1.20V.

    also i think by default the Processor Voltage was at 1.375V(ish) .. i noticed with 11x multiplier & 200 HTT (default) .. orthos failed .. it was only until i bumped the voltage to 1.40V that it ran good.

    default temperatures (1.40V):
    @1000MHz (5x HT) - 30-35*C/IDLE & 50-52*C/LOAD

    let me know what you think Psychlone
     
    Last edited: Apr 17, 2007

  11. Satan

    Satan Ancient Guru

    Messages:
    6,684
    Likes Received:
    0
    GPU:
    some
    If I do get this mobo, I will DEFINITELY check here!
     
  12. Psychlone

    Psychlone Ancient Guru

    Messages:
    3,688
    Likes Received:
    1
    GPU:
    Radeon HD5970 Engineering
    OK...it's obviously either your CPU limiting your overclock at this point. (if you removed the other variables correctly by dropping your RAM divider to it's lowest, and the HT Link Speed to it's lowest) OR your cooling (as a whole) limiting the overclock (due to heat-soak)

    Our board is proven stable at 400MHz HTT as well as 1500MHz HT. Our RAM is running at 582MHz in my rig as we speak, so that's not it. We're both running 550W Antec PSUs - both with the same amperes on the 12V rails, so that's not it. - THE only thing that's left is the CPU itself.

    I have a question about your case cooling. - heh...what IS your case cooling like? ;)

    Your HT MHz (and overvoltage on the NB) is going to be the biggest contributor to your idle/load case temps, probably even raising the temps on your CPU hs/f at the same time considering there *may* not be enough air throughput inside your case.

    I have an Aspire X-Navigator with 4 X 80mm fans and 1 X 120mm side panel fan. And, I've tried EVERY possible combination with them and concluded that the 2 80's in the rear are PUSHING air INTO the case, the 1 80 in the front PUSHING air INTO the case, the 1 80 on top BLOWING air OUT of the case and the 1 120 on the side BLOWING air out of the case. It creates a slight positive pressure to which all the fans are capable of their best efficiency. Negative pressure is a HUGE no-no with fans, as is sitting one on top or too close to another...they'll cancel each other out, or cause a negative pressure to where the fans cannot work efficiently, or they'll ultimately cause turbulence.
    I actually downloaded several softwares and spent days calculating the airspeed velocity of leading and trailing edges of blades with different CFMs, RPMs, etc. - and ended up with the PERFECT combination for my system.

    Not to degrade from what I just said, because case cooling is *just* as paramount to cooling as a good CPU hs/f assembly, but I'd definitely have to suggest a better hs/f as well as better case cooling. You can usually pick up a Tt Big-Typhoon at a local retailer for close to $50, they're almost silent, and do the job REAL well.

    Now that we know what the problem is, you've got something new to work on...case fans are really cheap, and a great place for you to start. This would be the first thing I'd do...as well as acquire a Tt Big-Typhoon (you're not going to find better cooling for cheaper IMO!) - AND, they're not going to be trash in time for your next major upgrade...unless you decide to get a new case...

    Anyway, that's my input for now. If you do indeed decide to get better case/CPU cooling, and play around some more, I'm *SURE* you're going to get better results in your temps, which means that you'll be able to achieve higher overclocks. - If you do get messing around like this, post your BIOS again (just copy/paste my format and change the values to your own) so that I can see where you're at.

    At some point, your CPU is going to be the major limiting factor in your overclock. Not to worry, remember...if you've overclocked at all, you've gotten something that you effectively didn't pay for...and you're already ahead of the game!

    Good luck!

    Psychlone
     
  13. tnet

    tnet Master Guru

    Messages:
    278
    Likes Received:
    0
    GPU:
    HIS X1900XT ICEQ3
    1) my case is a Asus Vento 3600. i usually leave the side left open just kuz i thought more air would enter (perhaps that actually causes problems). however the case has

    Front: 80mm
    Rear: 120mm
    Side: Venting hole with air duct

    2) my HSF is a Zalman CNPS9500. do you think the TT typhoon will be that big of an improvement?

    3) should i keep dropping my HT Driving Strength? i've only been sticking to 1000MHz and 800MHz.

    if indeed cooling is the problem, i think i'm just going to not overclock this system and wait until i upgrade come summer time for warhammer online ^_^

    and/or i may just buy a opteron 165 with a steeping like yours off ebay and go watercooling. i think with ur current setup there isn't really a need to upgrade until a year from now or maybe more.

    i have already learned much more then i did about overclocking with your help. really wish this would work out lol .. let me know what you think.
     
    Last edited: Apr 17, 2007
  14. Psychlone

    Psychlone Ancient Guru

    Messages:
    3,688
    Likes Received:
    1
    GPU:
    Radeon HD5970 Engineering
    Well, the CNPS95 does a *little* better in all the reviews I've seen, but real-world performance, I've not seen them perform as well as the Typhoon - although I haven't owned the Zalman.

    Keeping your HT at 1000 or 800 is just fine, but, if it will actually lower your temp another 5*C by dropping to 600 (3X) then by ALL means, do it! (you're not really loosing that much performance anyway - the PCI-E lanes we're talking about aren't even *close* to saturated at full load.

    If you're wanting to upgrade to an Opty, let me talk you out of it...chances are that there aren't any of these older 'known good' steppings out there anymore. - I have heard some good things about 1 of the new steppings, but there's no 3.2GHz monsters out there anymore (unless you get *really* lucky!) - seriously, with the new ATI R600, the new Barcelona as well as the RD790 all just around the corner, upgrading right now is the *worst* possible time.

    Seriously, if I were you, I'd spend a little $ on case cooling and get that squared away for the future...get a couple (few?) good-quality fans that are quiet but get the job done...this is the kind of upgrade that you get to take with you on your next upgrade, and the one after that, and the one after that...it's $ wisely spent.
    A *little* positive pressure inside the case has shown to be more effective in dispersing heat than leaving the side panel off (not to mention the dust! - you must have to clean that thing out all the time!) - with positive pressure, dust doesn't really have much chance to enter the system, and if it does, it's not there for long enough to stick to anything because of the pressure.

    Anyway, that's my feelings on all that. I wouldn't upgrade anything major right now, save your money and wait for the next iteration of AMDs powerhouse (along with ATIs newest chipset and graphics cards)

    You're welcome for the help in this post, I'm sorry that it didn't go as easily as it *could* or maybe *should* have, but such is the beast with overclocking. - At least you learned how to overclock an AMD properly. ;)

    Psychlone
     
  15. tnet

    tnet Master Guru

    Messages:
    278
    Likes Received:
    0
    GPU:
    HIS X1900XT ICEQ3
    true. thanks for all your time spent helping out. i really really really appreciate it :) i got one last question. by changing the HT to 600 (3x) .. what exactly would i be doing? if i were to use 260 HTT with a 10X multiplier then my HT Link would be around 880MHz? not really sure what that number does. could you explain :)
     

  16. Psychlone

    Psychlone Ancient Guru

    Messages:
    3,688
    Likes Received:
    1
    GPU:
    Radeon HD5970 Engineering
    From the AMD website about HyperTransport:

    AMD HyperTransport™ Technology

    * HyperTransport technology is a high-speed, low latency, point-to-point link designed to increase the communication speed between integrated circuits in computers, servers, embedded systems, and networking and telecommunications equipment up to 48 times faster than some existing technologies.
    * HyperTransport technology helps reduce the number of buses in a system, which can reduce system bottlenecks and enable today's faster microprocessors to use system memory more efficiently in high-end multiprocessor systems.
    * HyperTransport technology is designed to:
    o Provide significantly more bandwidth than current technologies
    o Use low-latency responses and low pin counts
    o Maintain compatibility with legacy PC buses while being extensible to new SNA (Systems Network Architecture) buses.
    o Appear transparent to operating systems and offer little impact on peripheral drivers.
    * HyperTransport technology was invented at AMD with contributions from industry partners and is managed and licensed by the HyperTransport Technology Consortium, a Texas non-profit corporation. The full specification and more information about HyperTransport technology can be found at HyperTransport.org.

    And, an excerpt from the almighty Wiki:

    HyperTransport comes in three versions — 1.0, 2.0, and 3.0 — which run from 200 MHz to 2.6 GHz (compared to PCI at either 33 or 66 MHz). It is also a DDR or "Double Data Rate" connection, meaning it sends data on both the rising and falling edges of the clock signal. This allows for a maximum data rate of 5200 MTransfers/s when running at 2.6GHz; this frequency is auto-negotiated.

    HyperTransport supports an auto-negotiated bit width, based on two 2-bit lines to 32-bit lines. The full-sized, full-speed, 32-bit interconnect in each direction has a transfer rate of 20,800 MByte/s (5200 MT/s * (32 bits / 8)), making it much faster than many existing standards. Buses of various widths can be mixed together into a single application (for example, 2x8 instead of 1x16), which allows for higher speed interconnects between main memory and the CPU, and lower speed interconnects among peripherals as appropriate. The technology also has much lower latency than other solutions.

    HyperTransport is packet-based, with each packet always consisting of a set of 32-bit words, regardless of the physical width of the connection. The first word in a packet is always a command word. If a packet contains an address, then the last 8 bits of the command word are chained with the next 32-bit word in order to make a 40-bit address. An additional 32-bit control packet is allowed to be prepended when 64-bit addressing is required. The remaining 32-bit words in a packet are the data payload. Transfers are always padded to a multiple of 32 bits, regardless of their actual length.

    HyperTransport packets enter the interconnect in segments known as bit times. The number of bit times that it necessitates depends on the width of the interconnect. HyperTransport can be used for generating system management messages, signaling interrupts, issuing probes to adjacent devices or processors, and general I/O and data transactions. There are usually two different kinds of write commands that can be used - posted and non-posted. Posted writes are ones that do not require a response from the target. This is usually used for high bandwidth devices such as Uniform Memory Access traffic or Direct memory access transfers. Non-posted writes require a response from the receiver in the form of a "target done". Reads also cause the receiver to generate a read response.

    HyperTransport also facilitates power management as it is ACPI compliant. This means that changes in processor sleep states (C states) can signal changes in device states (D states), e.g. powering off disks when the CPU goes to sleep.

    - Effectively, having your HT set at 600 (which is *really* 1200MHz considering both the 'up' lanes *and* the 'down' lanes) is no performance loss...let me explain it this way - for many years AGP was the standard for graphics cards...it was a 'Bus' that allowed a certain amount of data to be transferred between the GPU and the CPU and back again. Even though nothing has AGP anymore, the big companies saw a new and better way to transfer that data (but not just between the GPU and CPU anymore, more as a compliment to and shared between many components, making it's *own* 'bus' and in the meanwhile, outdating the 'old' way.) - The old and now completely outdated AGP system wasn't even *close* to done before they did away with it - what I mean to say is that the 'bus' wasn't even close to saturated with data even with the newest and top of the line AGP graphics cards available at the time of it's demise.

    Now, with HT, HTT and PCI-E (all coming to maturity and several versions higher than their respective originals) - we have the advantage of having effective '52 lane highways' to transfer all the information.

    Anyway, to make a long story short, if you run your HT at 600MHz (3X), and multiply that by an HTT of 260, you're going to be running those lanes at 780MHz. - Too bad you can't get some better case cooling. I know for a fact that the ASUS A8R32-MVP Deluxe can easily be stable at higher than 1200MHz and as high as 1500MHz...but if it's causing your entire case to heat up because of it, it's not worth it...and you're not going to see any kind of performance loss. - In fact, try this: Run SuperPI to 1Million using your HT at 800MHz (4X) - then drop your HT to 600MHz (3X) and see if it happens to get to 1 million digits any slower (+ or - 2% is normal, and within the range of something running in the background, etc.)
    Fact is, you're just not pumping enough information through the HT to saturate it enough to take a performance hit.

    Psychlone
     
    Last edited: Apr 18, 2007
  17. tnet

    tnet Master Guru

    Messages:
    278
    Likes Received:
    0
    GPU:
    HIS X1900XT ICEQ3
    gotcha .. well i've just realized that this isn't the cpu nor the case to be overclocking with .. so although this is a great motherboard .. i'm not in the best conditions to do anything with it .. however i have learned a lot from it. one last question though for you psychlone if you could help me out. with stalk settings .. can you help me out with my memory timings =)

    seeing as how the HTT is 200 and the multiplier is 11x at default .. i'm running the memory at 200MHz .. pretty much i'm running on (AUTO) .. the only one i wanted to change was the CMD-ADDR Timing to 1T. i was under the impression the memory you and i have could be ran at 2-3-3-6-1T timings on 2.75V. not exactly sure why the motherboard gave me these timings when i changed the voltage and the memclock value. anyways have a look and let me know!

    _Memory Configuration
    _Memory Timing Mode_____________Limit
    ___Memclock Value_______________200
    _MCT Timing Mode________________Manual
    ___CAS (CL)_____________________3
    ___TRAS________________________8
    ___TRP_________________________3
    ___TRCD________________________3
    ___TRRD________________________2T
    ___TRC_________________________11T
    ___TRFC________________________13T
    ___TRWT________________________4 clk
    _MCT Extra Timing Mode____________Manual
    ___TREF________________________15.6us
    ___DDR Input Strobe Skew_________Disabled
    ___Burst2Opt____________________Disabled
    _User Configuration Mode___________Manual
    ___Read Preamble________________6.0ns
    ___Asyc Latency_________________7.0ns
    ___CMD-ADDR Timing_____________1T

    _Bank Interleaving_________________Auto
    _Burst Length____________________4 Beats
    _Software Memory Hole____________Disabled
     
  18. Psychlone

    Psychlone Ancient Guru

    Messages:
    3,688
    Likes Received:
    1
    GPU:
    Radeon HD5970 Engineering
    Mine does the same exact thing...don't know why other than maybe Corsair improperly badged these sticks, OR the motherboard has some sort of problem reading the correct SPD...
    You can, however, set those advertized timings in the BIOS without any problems whatsoever. You can also crank up the voltage to them to 2.9V (if you need, but if you *do* end up giving them more juice, test with the lowest in increments) without damaging them or creating much more excess heat if they happen to be unstable at those settings.

    Psychlone
     
  19. tnet

    tnet Master Guru

    Messages:
    278
    Likes Received:
    0
    GPU:
    HIS X1900XT ICEQ3
    okay i'm back at it again .. a case with better cooling .. and a heatsink that boggles my mind :)

    got the opteron 165 now with CCBBE 0615EPMW .. good stepping so they say .. but i just must have the wrong configurations or something :p

    one thing i'd like to ask first is .. the default core voltage should be 1.350V correct? if that is so .. i don't understand why CPU-Z keeps showing the voltage at 1.376V-1.392V .. even when i set the core voltage at 1.350V or 1.325V in BIOS.

    [​IMG]

    as far as overclocking concern .. so far i've only found this to be the stable setup that boots up .. haven't done any orthos testing.

    ADVANCED
    _Jumper Free Configuration
    _AI Overclocking__________________Manual
    _CPU Frequency___________________290
    _PCIE Frequency__________________100

    _FID/VID Change__________________Manual
    _Processor Frequency Multiple_______X9
    _Processor Voltage________________1.350V
    _DDR Voltage____________________AUTO
    _VCORE Voltage__________________AUTO
    _Northbridge Overvolt_____________Enabled
    ___Core Voltage__________________1.5V
    ___HT Bus Voltage________________1.5V
    ___PCIE Voltage__________________1.2V

    _PEG Link Mode___________________Auto
    _PEG Buffer Length________________Auto

    _AI Clock SkewA__________________Auto
    _AI Clock SkewB__________________Auto

    _CPU Configuration
    _GART Error Reporting______________Disabled
    _MTR Mapping____________________Continuous
    _Cool N Quiet_____________________Disabled
    _HT Link Speed___________________800 MHz

    as far as the memory .. memory divider of 166 and the timings are just 3,11,4,4,4,10,13,5,15.6,9.10 like yours above for now

    temperatures with this setup: 25*C idle/40*C load .. however some of my fans are off because i didn't bring enough cable connections for the PSU so some are running off the same wires which i feel is being over-used and causing problems .. so i believe the temperature can get lower with everything running properly.

    i've tried to get above 2610MHz .. but the combo's just don't seem to be working out .. perhaps i just have all the voltage wrong.

    i've seen the overclock database and it says 3.0GHz is reachable with a vcore of 1.40ish or 1.50ish .. then again having the stepping doesn't guarantee the same results .. but i just don't believe this is the top speed :(

    *update*

    i've noticed that if i don't input memory timings myself and leave it all on AUTO .. the system doesn't boot up even with the HTT & Multiplier up top, so i'm going to take a wild guess and say that if i did want to go higher i need to figure out the correct timings?

    *update 2*

    ran orthos for over 2 hours small FFT's .. cpu seems very stable .. i'll be leaving it over night to see how that goes .. i tried to run a blend however and it ran into a lot of PAGING problems so i'm guessing the memory still needs tweaking for sure
     
    Last edited: May 2, 2007
  20. Psychlone

    Psychlone Ancient Guru

    Messages:
    3,688
    Likes Received:
    1
    GPU:
    Radeon HD5970 Engineering
    Your VCORE is on AUTO...that's why CPUz is showing a different voltage than what you put in the BIOS - VCORE automagically adjusts the voltage going to the CPU to help with stability at higher MHz - if you watch it, you'll see it fluctuate quite a bit...it'll drop to 1.35 and probably go as high as 1.45V - it's normal, but you should set it manually as well as your DDR Voltage.
    You should find your max stable overclock at stock CPU volts (with VCORE DISABLED) - then raise it a bit, find it's max overclock, raise it a bit more, find it's max overclock - ALL THE WHILE WATCHING YOUR TEMPS!!! - Once you run out of adjustments, you can begin with the VCORE OVERVOLT +100mV option, then the +200mV option - you'll also need to experiment with lower CPU voltages + the VCORE OVERVOLT +100mV as well as lower CPU voltages + VCORE OVERVOLT +200mV...WHILE WATCHING YOUR TEMPS! (seriously - you're messing with giving your CPU more voltage than AMD wants you to, but you can do it safely as long as your temps never go higher than say 58*C)
    Example:
    1.30V CPU Voltage + 100mV (~.074V actually) = 1.374V total
    1.35V CPU Voltage + 100mV (~.074V actually) = 1.424V total
    ---------------------
    1.30V CPU Voltage + 200mV (~.178V actually) = 1.478V total
    1.35V CPU Voltage + 200mV (~.178V actually) = 1.528V total
    You can see that if you're giving your CPU max voltage 1.40V AND utilizing the CPU Overvolt option at +200mV, you'll be pumping +-1.578V to your CPU - and I wouldn't even think about that without some serious Peltier or dry ice!!!
    ASUS boards have a 'bug' on the overvolt...+100mV *should* be +.1V, but as measured by a reliable multimeter, it's actually roughly .074V, and the +200mV *should* be +.2V, but it's really closer to .178V

    I would also set your DDR Voltage to 2.85V and leave it there...I've been running mine right there for months.
    Your RAM, even though it's the same as mine will overclock differently. NO 2 components are exactly the same. I have the CCBBE0615DPMW 165 that hits 3.2 stable and with .-25V over stock...see any other CCBBE0615DPMW's out there that are doing that? - No, because all components aren't created equally, even given the *same exact stepping/serial/lot#, etc.*
    You've got alot of work to do on your RAM - visit the post above where I guide you through it...you can't just set my paramaters and figure you're 'there'...my BIOS in the first post is no more than a template to teach you where you should be looking, and what you're looking at...nothing more. (seriously...if someone with an AM2 CPU just copied down all the info above into their BIOS, they'd fry their stuff in one shot - I have a discalimer just for people that try that stuff...I can't be held responsible for others not doing their own legwork.

    Go back, re-read post #3, do the math, and you'll figure it out...it's all about your latencies at this point (after you fix the voltages above manually!)
    You're going to be resetting your CMOS alot in the next little while - but the end result FAR outweighs what you're about to go through - keep telling yourself that, and revisit this thread, and when you read these words again, you'll laugh a sigh of relief, for you *know* what you're getting vs. what you would have had if you didn't tweak your latencies.

    Good luck, and I'm glad you got your case figured out - remember to watch your temps and hook up those fans, keeping everything out of the way - you want good airflow through the case without restriction of wires, etc.

    ;)

    Psychlone
     
    Last edited: May 3, 2007

Share This Page