Question about vcore with 955 be on asus m4a79xtd

[jgree32]

I've searched high and low for an answer to this issue and only found one other person that even mentioned this but there was never a response to his question. I'm just recently coming back to the amd scene. I haven't overclocked an AMD since socket 754 days. I just recently built a Phenom 2 with these specs
955 black edition
Asus m4a79xtd (bios 0605)
ati 5770
gskill ripjaws 4gb 1600 cas 7

I started off the overclocking by simply raising the multiplier from stock of 16x to 18x. I had left the vcore at auto which i know is a typical no no but i just wanted to see what it would do at stock voltage. Well, everything was running seemingly fine but i checked the core voltage with cpu-z and the amd cool n quiet software and under load they were reporting vcore at nearly 1.5V!!! So my question is why on earth would the vcore increase by simply raising the multiplier? I have also verified this behavior in the bios vcore reading. Also if I set the vcore to the stock value of 1.35V i believe it is would I still retain the cool n quite functionality that automatically lowers multiplier and vcore? I appreciate any help on this. If anyone has any links to decent guides for overclocking phenom x2 i'd appreciate as I'm new to the amd scene again.

[AlleyViper]

I also came from the 754 boat, so you're not alone : p.
Not every 955be has the same stock vcore, if you leave everything at auto including the 16 multiplier you might check it under cpu-z. On some motherboards as my dfi AOD will report 1.35V no matter what is set. In my case it's a 1.325V part, but most times it's 1.350V or even 1.375V.
Cool n quiet functionality when overclocking will depend completely on your mobo implementation, but at first if you intend to overclock, completely ditch it so you can control all parameters. For a 18X multiplier on a c3 955be you'll probably only need 1.35V or 1.375V to be prime stable, and 19X should be possible with good temps and still a low vcore as ~1.400V. Above that, that's when very good cooling is needed on air.
On these CPU's even before you can sustain a dangerous vcore, you'll be failing with errors due to heat, so you should keep vcore and cpu-nb the lower you can. You should only need 1.5V from 3.85 to 4.0GHz. In these PII, you'll get big gains in overclocking your NB, so setting it at 2.6 should be possible without a big voltage increase. Probably in the range of 1.25-1.30V from the default 1.125V, 2.4 might be .025 to .050V less.
Your sticks aren't quite famous around here for 1600 cl7, but for starters they should run ok at 1333 7-7-7-24 1.65V, and then you can try to build your way to CL6. Test every variable on it's own, and look for some guides chew as put around.

[mäggäri]

I must say that Ripjaws CL9 sticks are the most 5h1ttiest choice for AMD mobo. Mby they've fancy looks, but that's all what they have to offer.

1) Upgrade to bios 0704. (gives PCIE spread disabled option, mby some other improvement too - hey, it's a newer version anyway!) :p

2) VCore rises because you've left to it "AUTO" option. I'm not sure if that feature is only on ASUS boards. So set it manually to value you like.

3) I prefer CNQ set to disabled when overclocking.

Depending on your cooler, u can set VCore actually up to 1.6V on air.
OT: I'm running Phenom II 550 BE @ X4 and 4020MHz. On water ofc. But it went around 3920-3950 on air too, cheap ~30-35euro coolers (like Mugen2, EKL Brocken), 1.5VCore was needed.



SETTINGS FOR 4GHz mby works mby not, try!

Leave FSB 200MHz
set 1.6VCore (if Mugen2 or equivalent air cooler)
CPU/NB Frequency 2000MHz (1:1 divider, in bios those are actually dividers, it does NOT show correct NB or HT speed, Asus engineers have f00ked up it, can be seen in the picture what I mean)
CPU ratio 20x
CPU/NB Voltage 1.1V
CPU VDDA Voltage 2.8
DRAM frequency to 1600
VMEM 1.65 for example (or higher, if the specs of your sticks say so)
Timings can be manually adjusted to 7-7-7-24 or what ever your sticks were binned for.
Set HT also to 2000MHz
Those other unmentioned values can be left to auto.
---------------------------------------------
with my 550 BE I also use
NB 1.28V
NB 1.8V = 1.9V
SB 1.35V
---------------------------------------------

IF works (boots up just like that), have fun! Time to start lowering the VCore from that 1.6V until your setup becomes unstable. Then add the minimum amount to it.

ps. There's also 8 slots for saving different BIOS settings.

After all testing it's time to start tightening MEM CL, and getting some speed for NB. Usually NB goes 2600-3000MHz. And when we now know quite much like the max of CPU is capable of (in 50MHz range, after playing with multiplier), and all mem settings and how the NB goes. Comes time to start finding the max FSB.

Takes time, I know, been there before =D But this way you will get the max perfomance of your system.

[jgree32]

I put everything back to stock settings and decided to run memtest just to check my RAM out at stock settings. Memtest crashes after only a few seconds of running whenever I use both sticks of RAM. I've tried the RAM in all slots and get same result. If I use either stick of RAM by itself in any slot it runs just fine. So what setting could I be missing that would prevent me from running two sticks of RAM with everything stock? Thanks for help guys.

[crazydiamond]

seems like same problem blitos is having with his ram....

http://www.xtremesystems.org/forums/...d.php?t=240496

[jgree32]

Phew....sanity check there. Decided to check for latest memtest update. I was running v3.5 and there is a 4.0. That seemed to fix the issue with the immediate crashing with 2 sticks. Still fooling with overclocking this 955 but not having much luck. It will 3400mhz at stock voltage but even with vcore up to 1.4125 i cannot get it to run at 3600mhz. I have no problems posting but get crashes very quickly in P95. Anything else I should know about that will hold back the oc potential? Also, just want to make sure. Am I correct that the ACC option largely has no effect on phenom 2 as that was something for phenom 1?

Thanks for the help

[AlleyViper]

In prime95 when I have bcc code 124 crashes it's usually not enough vcore, if that's what you're getting. Double check your temps if you are using the stock cooler. Over 1.4V for 3.6GHz it's quite a bit unusual on a 955.
But you should double check if prime crashes too at lower settings as your sticks seem not to play very well. If you're still at stock CPU-NB (IMC, not the external NB), and stock 10x multiplier, up it's voltage to 1.2V just to check if it's the IMC crapping at 1.125V, but latter you'll want to run it very least at 12x or 13x with more voltage (1.25-1.30V).
For now, disable CnQ, force ht at 2.0GHz, disable ACC if you're not fine tuning each core.

[jgree32]

Thanks for suggestions. I'll try some of the things you noted. My sticks seem to run fine at their rated CL7 1600 speeds. I have even lowered div to 1333 once just to ensure ram wasn't causing error. So I'm fairly confident my ram isn't causing the errors. Also i forgot to note. When I crash prime 95 it's a total lockup with blue screen. Not sure if that is indicative of a certain problem or not. I get "windows has shutdown to prevent damage to hardware" which seems to be a generic error.

[AlleyViper]

On most bsods there will be an error code, my usual culprit for low vcore is 00000124, which will be referred as bccode 124 after windows reboots with the error message.

Just because ram doesn't error out immediately in memtest, it doesn't mean it's prime stable. And even for a stock multiplier and voltage for both cpu and cpu-nb plus ram at 1600 stock, cpu-nb (imc) voltage might not be enough due to the added workload.

Then, there's another matter, you should check Tony's testing on imc speed vs ram speed sweet spots. Up until nb 2.4, 1333 cl6 will be faster than 1600 cl7, while at 2.6 they perform almost the same, trading blows depending on tests. Higher NB speeds will favor higher ram speeds, but the needed cpu-nb voltage will start to ramp up too.
Remember that cpu-nb speed is as important as cpu mhz for performance overall.

[jgree32]

What do suggest for completely isolating the cpu then? Is putting the divider at 1333 and timing on auto which makes it like 9-9-9-9 not enough for 1600 cas 7 RAM?


Thanks,

[AlleyViper]

What do suggest for completely isolating the cpu then? Is putting the divider at 1333 and timing on auto which makes it like 9-9-9-9 not enough for 1600 cas 7 RAM?


Thanks,

If you really want to start by the cpu try 1333 1T Unganged CL7 instead, your sticks should do it with no sweat, and give a slight bump to your cpu-nb leaving it at stock 10x, 1.2V maybe. Then you can double check if 1600cl7 or 1333cl6 is as stable.
Then after you find your mhz and vcore, work your way on the NB at very least to 2.4GHz,2.6 or more will be better if you can keep good temps.

[jgree32]

Starting to conclude that I may just have a dog of a processor. Looking around everyone else's 955 don't seem to need as much vcore as mine is requiring for similar clocks. For 3.6ghz i'm needing ~1.425V to be remotely stable. Thinking of just giving up on 3.6ghz and going with 3.4ghz since I can run that at 1.3625V. Still seems really odd to me I need that big of a jump in vcore to go to 3.6. I've tried increasing NB/cpu voltage to 1.2V+ with no help really. Still can't help but think there is some funky setting or something holding me back, but I just haven't figured it out yet.