Individual core testing, stablity, and ACC...
This mobo or cpu will not run 3000MHz NB speed.... If I go 3000mhz in the bios it safe modes itself im thinking cus it will boot, but will only be running 2100 MAX.... 2999MHz works and boots fine as well as shows running at its set speed. But then weird shit happens lol.... Windows Defender is popping up asking to update.... the side bar closes and pops up asking to restart the gadgets lol....
Yep, whenever you're using a setting that's too far out of range/incompatible on the AM3 systems it seems that the system will automatically do that. It's the samething when trying to run 1800Mhz ram for my system, no matter what settings used it will not work it seems, and then just safe modes itself.
The other bit about the weird stuff is what happens (I've seen that LOADS of times testing undervoltages) just before the system is about to freeze/reset on you. It always happens when there is not enough voltage so the component is unstable causing things to shutdown automatically, virus programs to try to update, system clocks to be incorrect, windows to say it's not valid, etc.
Any tips on how to find the strongest core on my 955BE?
I meant to answer this right after you posted it. It's not a quick process finding the strongest/weakest cores, but it's a very important thing to find out and worth the time. What I do is load uo K10stat (or AOD if you wish, but K10stat is much faster to use), task manager, and LinX or some other serious torture program. Select the stock cpu voltage in the bios (1.35v) since it's the only voltage that we know is absolutely 100% safe. That part will save you a lot of time as well instead of chasing settings to start testing at.
Once you have those 3 things loaded up, set LinX to have only 1 thread running (settings) then go back to K10stat. Using K10 choose whichever core you want to start with first (I recommend Core 2 or 3), and up the speed from the stock 3.2Ghz/16 FID to say 3.5Ghz/19 FID and hit the little apply button next to that core. Using the task manager go to the processes tab and select whatever torture program you have up, then right click and change it's affinity to only use the core you're testing.
After that it's as simple as hitting start on the torture program, I usually do a quick 5 minute run of LinX so I don't have to wait all day and I know it gets worked like mad. Make 100% certain that you are not doing anything else on the computer, you're are only stressing a single cpu core at a time, and that EVERYTHING else in the system is 100% stock/safe settings so that you don't get an error from something other than the cpu core.
If the core passes that test, let the usage go back to 0% and then up the core speed again another notch...don't make jumps even though it takes time to do each one. I personally play it smart during this testing and cap the max core speed to 3.7Ghz/ 21 FID. By doing that you can just set the core back to it's stock speed, and move right on to testing the next core without worrying about having to reboot from an error. Once you've done all the cores at 3.7Ghz (if the system makes it that far) then go back to the first core you started on and try 3.8Ghz...it may or may not make it, but that's what we're looking for. If it has an error then reboot, and go to the next cpu core to test. Once you've finished testing all the cores at 3.8Ghz or whatever the top speed you're able to get to ON STOCK 1.35V still then you can work on whatever the weakest core is.
========================================================
ACC testing with specific cores
Now that you've found what the weakest cpu core is you can work on helping it out to achieve a higher clock. Go back to your bios and still leaving stock voltage and overall cpu speed change the ACC option to enabled PER CORE and set all the cores to 0% except for the weaker core that you're testing. That core should be set to either +2% or -2% to begin your testing. Again, do not try to test more than one core at a time as it will defeat the purpose of what we're doing.
Go back to windows and load up the 3 programs again, set the core speed to whatever failed for you last time again. This time for the torture test however you will be doing a 10 minute test not just 5 minutes so that we can be sure that stabilty has been achieved or at least ~80-90% (it better make it 7-8 minutes). If it doesn't make it then go back and change the ACC level up one more notch and see what happens. Make notes of how long it makes it during each attempt if it doesn't pass. If it fails super quick and you were using +2% then go to -2% and vice versa then move on. If it fails after a few minutes then proceed to the next notch from wherever you were with the ACC level.
If you have 2 cores that didn't make it during the stress test then check them individually with ACC, if it's 3 cores or all four then don't bother with the ACC test and just up the cpu voltage 1 click to see if they pass. Your goal is to get it so that not all of the cores fail...just one or two...and then testing again until there is only 1 that fails.
=========================================================
Once you've finished you now have an idea of how to really attack the system when going for an overclock. You'll know which core will need the most help and how to work with it. You'll know how much voltage you need to compensate for the worst core compared to the others, and you'll know how much ACC is the starting point to achieve greater stablity for that core.
In my case testing all cores gave me the following results...
Core 0 = 3.8Ghz passed no mods/retests
Core 1 = 3.8Ghz passed no mods/retests
Core 2 = 3.8Ghz failed - add +2% ACC quick failed - add -2% long failure (good) - add +2 cpu volts total pass
Core 3 = 3.8Ghz failed - add +2% ACC failed - add +4% mid failed - add -2% long fail - add +2 cpu volts total pass
The results above don't show the whole picture, but Core 2 is easily the weakest out of my available cores. It failed the fastest w/o any mods (less than 1 minute), adding the first mod (less than 30 seconds), the third mod (failed roughly 6min 1sec), fourth mod total pass. I already knew that Core 2 was my weakest by far with Core 3 being next as I did this testing a long time ago, however now that I have a board with ACC and DDR3 it was time to retest all over again. On the previous board with DDR2 I was only able to get 3.6Ghz partially stable on Core 2 and 3.7Ghz stable on Core 3, the other two were the same passing 3.8Ghz each and able to run 3.9Ghz for a very short while.
If I didn't have the information then I would end up trying to run 3.8Ghz at something like say 1.50v because it would just fail at anything less. With that information I know that I need to start off with -2% ACC and then test 3.8Ghz to see if it passes. If it fails then add +2 clicks to cpu volts and see if it passes then. That core just needs a few clicks more volts to make it happen than the others.
I'm typing to you right now from 3.8Ghz @ 1.425v on all four cores right now. The best I loaded 3.8Ghz before on any board was 1.39v but it wasn't stable. Added -2% ACC to both cores 2 & 3, clicked the cpu voltage up 2 clicks and I've been running 1.425v for more than 1 hour now without so much as a hiccup. Hopefully it'll pass full stablity testing, but either way it's working great...1.40v would make it into Windows for only 30 seconds or so.
Hope this helps, have fun testing...the initial core tests will take you roughly 1 hours time with 5 minute tests starting at 3.5Ghz.
========================================================================
Northbridge approximate voltage versus clock guide...
During testing with my PII 955 on more than one board I've found the following to be true for both boards to within 1 click of the voltage. This is an APPROXIMATE guide you can use when trying for stablity with Northbridge clocks, so don't think it's law as your system may vary though I doubt by very much. Adjust the voltage with the "cpu/nb" setting in the bios to adjust the northbridge/memory controller voltage level. The other Northbridge voltages are for the chipset.
2.0Ghz @ 1.06v
2.2Ghz @ 1.07v
2.4Ghz @ 1.10v
2.6Ghz @ 1.16v
2.8Ghz @ 1.26v
3.0Ghz @ 1.35v+
Those are the voltages that I've found to get those specific Northbridge/memory controller speeds 100% stable when using the 3.2Ghz speed of the PII 955. The same has proven true so far when testing all the way up to 3.9Ghz, and has worked for 4.0-4.1Ghz in my testing though cooling problems arise at those speeds from large cpu volts on air cooling. The default voltage for the Northbridge/memory controller is 1.10v on the Phenom II processors with the 2.0Ghz default speed. Much like cpu voltage it can be lowered and stll prove stable, you can also raise it and prove stable though please remember that heat increases significantly once the voltage is raised as it's housed in the cpu casing as well.
Also note that if you're having problems getting a specific cpu clock stable, it may be because you're running the Northbridge at too slow a speed and choking it. Instead of adding cpu voltage sometimes all you need is to simply raise the Northbridge clock speed and it will cure your problem. If you're getting BSOD as certain clocks...it's probably time to take a look at your Northbridge speed. 2.0Ghz is great for a 3.2Ghz setting...but 4Ghz will need more like 2.2-2.5Ghz for best stablity.
I can't find whom said it right now, but I read recently that your CPU to Northbridge speed should be roughly the following. (CPU * 2)/3.15 = Northbridge speed
There is of course a margin of freedom in that formula, but in all my time testing with Phenoms (since the first Phenoms came out) I can say that formula goes along with what I've found. I wish I could find his/her name so I could give them credit, but although I find the same things as they did
PLEASE NOTE that is
NOT my formula but I do believe it has some truth to it.
The Northbridge voltages/speeds mentioned above were arrived at with the rest of the system at the following voltages...
CPU VDDA 2.50v (stock)
HT Voltage 1.20v (stock)
NB Voltage 1.30v (stock)
NB 1.8 Voltage 1.80v (stock)
SB Voltage 1.20v (stock)
Memory Voltage 1.50v (stock)
Kei
(if you don't have ACC you can do the samething, I've been doing this a long time, the ACC part is just an extra step/bonus)