Reply to thread

Message: <blockquote data-quote="chanaka89" data-source="post: 3558141" data-attributes="member: 51319">For discussion’s sake, let’s say the same core failed repeatedly.  This scenario is likely caused by a low Vcc (although it doesn’t have to be).  For you quad core users, cores 0/1 and cores 2/3 should be treated the same, so if you get some core 0 and core 1 failures, treat them like a single core failure as you consider this analysis.So, I increased the NBCore a few notches and tried a few higher Vcc settings just to see if it was enough:[code]Overclocking log, Iteration Set 2Comments: Added some NBCoreDRAM    2.100VSBCore    1.55V[b]NBCore    1.41V[/b]VTT    1.200VVcc/Prime95 success or failure1.16250V    Failed on core 2 ~2 min1.17500V     Failed on core 1 ~3 min[/code]Again, I got two quick failures across the entire chip.  Ideally, you might want to collect more data points, but I took a hunch that 1.45V should be plenty for 8.5x400, and next added some VTT keeping the newer, higher NBCore constant – remember to only change one of them per iteration set![code]Overclocking log, Iteration Set 3Comments: Added some VTT and kept the higher NBCoreDRAM    2.100VSBCore    1.55VNBCore    1.41V[b]VTT    1.310V[/b]Vcc/Prime95 success or failure1.17500V    STABLE 15 min1.16250V    STABLE 15 min1.15000V    STABLE 15 min [/code]Now, with the higher VTT, I didn’t get a single failure for at least 15 min at the three Vcc values I ran.  I concluded that the VTT gave me the stability.  To test this hypothesis, I kept the higher VTT, but lowered the NBCore back to 1.37 and repeated in the 4th iteration:[code]Overclocking log, Iteration Set 4Comments: Kept the VTT, lowered the NBCoreDRAM    2.100VSBCore    1.55V[b]NBCore    1.37V[/b]VTT    1.310VVcc/Prime95 success or failure1.15000V    STABLE 2 h1.13750V    STABLE 30 min1.12500V    STABLE 1 h1.07500V    crashed p95 (n=2)1.09375V    crashed p95 (n=1)1.10625V    BSoD after 1+h1.11875V    STABLE 11 h1.11250V    Failed on core 0 ~ 1 h 8 min[/code]Now I got some stable runs.  After evaluating the data, I was able to nail down both my NB and VTT in only 3 iteration sets, arriving at what I thought was the stable Vcc in the 4th (I was later wrong).It’s a little easier to visualize if you sort the Vcc from low to high.  If you keep your log in a spreadsheet, you can easily sort them, here are the same data sorted by Vcc:[code]Overclocking log, Iteration Set 4Comments: Kept the VTT, lowered the NBCoreDRAM    2.100VSBCore    1.55V[b]NBCore    1.37V[/b]VTT    1.310VVcc/Prime95 success or failure1.07500V    crashed p95-program exited (n=2)1.09375V    crashed p95-program exited (n=1)1.10625V    BSoD after 1 h1.11250V    Failed on core 0 ~ 1 h 8 min1.11875V    STABLE 11 h1.12500V    STABLE 1 h1.13750V    STABLE 30 min1.15000V    STABLE 2 h[/code]It would seem as though 1.11875V was the winner.  I could have stopped right here and repeated extending the time out to 24+ h with these settings, but I elected to further optimize and targeted the VTT since I thought I could do better having jumped from 1.20 to 1.31 and skipping 5 sub levels in the process.  This time through, I held the Vcc constant and varied, VTT:[code]Overclocking log, Iteration Set 5Comments: 1.11875V seemed stable, minimizing VTTDRAM    2.100VSBCore    1.55VNBCore    1.37V[b]Vcc    1.11875V[/b]VTT/Prime95 success or failure1.250V    Failed on core 0 ~ 2 h1.260V    Failed on core 2 ~ 1 h 20 min1.280V    Failed on core 0 ~ 18 h 22 min1.310V    Failed on core 1 ~ 1 h 20 min[/code]This one is a little puzzling since the 3rd run (VTT=1.280V) lasted for over 18 h, yet the 4th run with a higher VTT died in under 1-1/2 h.  My thinking was that VTT wasn’t the problem, and that I had been mislead on the Vcc.  I was also getting a little anxious for this to be finished and I broke my own cardinal rule for the next iteration set by upping two variables at once: Vcc to 1.12500V and VTT to 1.310V.  [code]Overclocking log, Iteration Set 6Comments: 1.11875V seemed flaky, so upped the Vcc and kept the higher VTT.DRAM    2.100VSBCore    1.55VNBCore    1.37V[b]VTT    1.310V[/b]Vcc/Prime95 success or failure[b]1.125000V[/b]    STABLE 21 h 34 min[/code]Okay! So maybe it was the Vcc after all since it ran for over 21-1/2 h before I stopped it.  You could argue that there’s no difference between 18-1/2 h and 21-1/2 h and you would have a valid argument.  This underscores the need to collect multiple data point per level as I mentioned in the beginning of this section (I told you it was quick ‘n dirty)!Finally, I set out to essentially repeat my Iteration Set 5 minimizing the VTT with the slightly higher Vcc.[code]Overclocking log, Iteration Set 7Comments: 1.12500V seemed stable, minimizing VTTDRAM    2.100VSBCore    1.55VNBCore    1.37VVcc    1.12500VVTT/Prime95 success or failure1.250V    Failed on core 0 ~ 1 h 3 min1.280V    Failed on core 1 ~ 1 h 0 min1.310V    STABLE 34 h 41 min[/code]Apparently VTT needs to be 1.310V on this system.  For example 2, I just wanted a REALLY quick n’ dirty slower setting for my PC that I could use for non-CPU intensive computing since it’s summer now and I wanted a setting that would minimize the heat output for general computing.  I’ll reboot into 8.5x400 if I want the faster speed.  As such, I didn’t follow my own rule about changing only one variable at a time in the final iteration as you’ll see.Example 2: 8.5x333Hardware specs for your reference: Initially, I kept the Vcc from my 8.5x400 run knowing it should be stable for less clock (stock in this case) and wanted to see if I could run a lower VTT.  I also chose the lowest NB core.[code]Overclocking log, Iteration Set 1Comments: Initial tryDRAM    2.100VSBCore    1.55VNBCore    1.30VVcc    1.1250VVTT/Prime95 success or failure1.170V    Stable 15 min1.100V    Stable 15 min[/code]As you can see, the lowest VTT was stable for 15 min.  I went right to the Vcc using these values.[code]Overclocking log, Iteration Set 2Comments: minimizing vccDRAM    2.100VSBCore    1.55VNBCore    1.30VVTT    1.10VVcc/Prime95 success or failure1.00000V    Stable 15 min0.96250V    Stable 15 min0.93125V    BSoD (wouldn’t get into Windows)0.93750V    “0.94375V    “0.97500V    Stable 15 min[/code]So here you can see that 0.97500V ran just fine for 15 min.  At this point I figured I would also see if I could drop my memory vcore since 2.1V is required for >1,000 MHz and that 667 MHz should require less.  Honestly, I should have let this run for 8-12 h before changing another variable.[code]Overclocking log, Iteration Set 3Comments: minimizing dram coreSBCore    1.55VNBCore    1.30VVTT    1.10VVcc    0.97500DRAM Vcore/Prime95 success or failure1.91V    Stable 15 min1.80V Failed on core1 after 35 min[/code]As you can see, 1.80V wasn't enough (or it could be that my previous settings weren't enough since I only ran them for 15 min)!  I decided again to break my own rule about change more than one variable since I don’t really care to find the absolute minimum set of vcores for the stock setting.  I upped the NBCore from 1.30V to 1.33V, and also added back some Vcc from 0.97500V to 1.00000V and kept the 1.91V on the DRAM for the final set:[code]Overclocking log, Iteration Set 4Comments: changed several things![b]DRAM    1.91V[/b]SBCore    1.55VNBCore    1.30VVTT    1.10VVcc/Prime95 success or failure1.00000V    Stable 18 h[/code]I stopped it after 18 h since I won’t be using this setting for CPU-intensive stuff.  This stock level is just to have a lower heat/lower energy mode for web browsing/general computing.In any case, those examples should serve to illustrate the method you need to use to attack the task.To summarize, using a stepwise approach and documenting your runs, you should be able to arrive at a stable system (assuming your hardware can operate at the level you choice).  It probably goes without saying that you will need to repeat this process if change your operating conditions (multiplier and FSB).Temperature ManagementAn overclocked quad system is often limited by the amount of heat it’s producing, and the ability of the heat sink and fans to dissipate it.  If you’re getting high temps, there are a number of things you can do to help.  Most of them are hardware related but the first is the single most important non-hardware change you can make:•    Minimize your vcores first (described in the guide above)!•    Ensure good contact between the CPU and Heat sink is a must for efficient heat transfer.  A major bang-for-the-buck modification in this regard is lapping the surfaces that transfer heat (the base of your heat sink and the top of your CPU).  This involves gently moving the surface along wet/dry sand paper in increasing grits on a flat surface such as a piece of glass.  I did both the base of my Ultra-120 Extreme and the IHS (Internal Heat Spreader) on my Q6600 and saw some pretty dramatic decreases in load temps.  It should be noted that lapping your HS and/or CPU will void the warranty.  Comparing my stock HS/CPU to my lapped HS/CPU, on average lapping lowered the coolest core by 7 °C and the hottest core by 10 °C.  To read more about lapping your heat sink and CPU see these two threads; I have results and pictures of the process:<a href="http://forums.extremeoverclocking.com/showthread.php?t=261085" target="_blank">Lapping Q6600 IHS</a><a href="http://forums.extremeoverclocking.com/showthread.php?t=261084" target="_blank">Lapping the Ultra-120 Extreme</a>That said my X3360 did not need to be lapped.  I’m not sure if Intel is doing this with all their 45nm chips or just the Xeons, but it came from the factory very flat.  When I run prime95, the heat spread between cores is 2-3 °C.•    If your NB chipset runs too hot, consider adding a small fan.  I put a silent 40x40x10mm fan on my NB HS via a zip tie which lowered my NB temps by ~7 °C on load.  Pretty amazing effect for $3 fan and free zip tie <img src="/styles/default/xenforo/smilies/default/happy.gif"  class="smilie" loading="lazy" alt=":)" title="Happy    :)" data-shortname=":)" /><img src="http://img443.imageshack.us/img443/3114/nbcoolerib1.jpg" alt="" class="fr-fic fr-dii fr-draggable " style="" />•    Consider an upgrade to a more efficient heat sink (like the few mentioned in the beginning of the guide).  Remember that a quad core chip will produce about 2x the heat compared to a dual core chip.  You really do need to consider using an extreme HS if you plan to overclock a quad.•    Consider an upgrade to the cooling fan on the heat sink to something that has more flow.  Most of the larger HS’s will use a 120mm fan.  Some have the option for two fans.  I think the fastest 120mm fan you can use is around 1600 RPMs.  If you have a slower one, you might consider upgrading.•    Reseat your heat sink and make sure you’re using a quality TIM (thermal interface material) such as AS5.  Consider rotating the HS 90 degrees if it is designed to do so.  I seem to get better contact with my Ultra-120 Extreme when it’s orientated “North/South” than when it’s orientated “East/West.”•    Re-evaluate the way you’re applying the TIM/don’t use too much or make sure you’re using enough.  Thermal pastes aren’t all created equally.  Some are reported to be better than others.  I have always used Arctic Silver 5 on my CPUs (and AS3 and AS1 before that).  You can find all sorts of posts out there showing one to be better than another.  I’ll leave it up to you to pick one.  Again, I like AS5.  Here is a shot of my q6600 installed in the MB with AS5 right before I added the HS.  It shows the right amount in my opinion given a lapped HS and CPU (which is a thicker line than I used before); the red triangle I drew shows where that tag is on the CPU, remember that on quad core chips, the dies are placed in a different located relative to a dual core, see the <a href="http://www.arcticsilver.com/ins_route_step2intelas5.html" target="_blank">instructions</a> on AS5's website for more on this.<img src="http://img466.imageshack.us/img466/9365/line800qe9.jpg" alt="" class="fr-fic fr-dii fr-draggable " style="" />•    Use good cable management inside your case.  Use twist ties or tie downs to bunch cables and keep them out of the way of airflow.•    Make sure you have adequate airflow inside the case and make sure you’re using a well ventilated case.  People often overlook this, but it’s important.  Not all cases are designed for good airflow.  I have an Antec P182 which is a great design.  Make sure you have several exhaust fans and at least one intake fan.  120mm fans move more air than smaller 80mm fans do and also run much more quietly.You can see that my CPU load temps will increase/decrease as the ambient temperature fluctuates.  Have a look at the following thread for details:<a href="http://forums.extremeoverclocking.com/showthread.php?t=261087" target="_blank">Effect of room temp on CPU load temps</a>Controlling vdroopRemember the vdroop you saw earlier?  If you have a P5B-Del, you can use a pencil to modify your board to minimize or fully remove this idle-to-load vdroop.  Read the following thread if you want to do that:<a href="http://forums.extremeoverclocking.com/showthread.php?t=261143" target="_blank">Get more vcore under load: vdroop pencil mod (pics)</a>That's it for the guide.  I hope you got some good info out of it and are able to successfully o/c your system as a result!by graysky</blockquote>

Verification: Awruddata maasa keeyada?

Top Bottom