AMD x3/4 voltages and headaches

[James of the Mountains]

Hello,

Much of the info I've found regarding overclocking is specific to intel processors or locked multiplier AMD processors, so if I ask redundant questions please bear with me - I want to make sure I understand..

I picked up a 720 x3 BE. This is my first computer build in many years. I unlocked the fourth core no problem. It seems as stable as it did when it was 3 core.

I've seen some conflicting info on what voltages are a good idea to run, so I've stayed as close to stock as possible. I HAD it stable (OCCT's one hour test 3 times, prime95 for 2 hours) at 3.1 GHZ by four cores. Everything worked swell, I rebooted the computer a few times, that was normal, then I went to bed. That was about 5 this morning. (pacific time)

I crank things back up this afternoon... and the computer gets a real brief BSOD and reboots right after the windows login screen. WTF? it KEEPS doing it. Loads windows, then crashes. Linux seems unaffected, but I don't know what's causing this. I backed off the processor speed to 3.0, then to the stock 2.8 (but still at 4 cores) and it still crashed after loading windows. I reset all the stock voltages except I left the "CPU NB VID" at +.050v. I'm now in windows at 2.8x4 with that voltage.

What I cannot understand at all is why everything was totally fine and stable last night, and all of the sudden this morning it crashes loading windows? And why is it OCCT and Prime stable and twitching out on me after loading? If windows doesn't crash immediately after loading and showing me the desktop, it won't crash at all - and I don't get that either.

My system -
2.8Ghz AMD 720 x3/ 20 x4
4GB DDR3 G.SKILL ram
Gigabite ATX 770 motherboard
Windows XP 32bit
Linux mint 7 64 bit

I am seriously hoping that I just need to adjust a setting in windows or something simple, or that I'm overlooking a voltage setting somewhere.

When I was "stable" (sure as hell seemed stable according to the torture tests) last night at 3.1 Ghz my voltage settings were as follows:

CPU NB VID control +0.025v
CPU voltage control +0.025v
DRAM voltage control 1.600v (was 1.5 default)
DDR VTT voltage control .900v (that's default - can someone please tell me what this setting is? I don't wanna jack with it unless I know)
NB voltage control 1.220v (was 1.1 default)

and my temps (cpu - core temps won't read with an unlocked core) haven't strayed much above 60c during the tests. I don't think it's a heat problem.

So... what do you think?

 

[Yukmouth]

If you have the fourth core unlocked, you're going to need around 1.350v or more to stay stable. Your problem should be nothing more than voltage with such a modest OC, though I would say that enabling that fourth core can be a BSOD issue in itself.

If it was cooler in the computers room before morning hit, temps could very well be an issue if you're telling me you've seen your CPU hit 60c. You say you cant read your temps due to the unlock so I'm a little confused on how you're coming up with a temp. Do you have ACC set to AUTO? There are a few articles spread around the net where people have unlocked a fourth core, only to see it fail to unlock at a later time. Mine didn't unlock, at all.

Your north bridge should be at 1.225v default.

VTT is lowest allowed voltage, I believe. (could be wrong on this one).

Leave the VIDs alone, leave VTT auto if you can, and make up for that fourth core with some CPU Voltage. Hopefully you have CPU NB voltage option.

What cooler are you using? You shouldn't mess with any other voltage than CPU/CPU NB.

 

[James of the Mountains]

I have 2 controls with NB in it, as referenced above, I've got CPU NB VID control, and just plain NB voltage control. It's not an onboard video MB, so I'd reckon that in "CPU NB VID" vid stands for Voltage-something-something?

the NB voltage control defaulted to 1.1. I've heard that in unlocking the fourth one should bump it from "the default 1.2" to 1.3, but since mine came defaulted to 1.1 I was hesitant to do so.

I'll bump the cpu voltage if i need to. Which NB voltage should a adjust, or will they both need it?

I'm running the stock HSF. I haven't been able to get one locally that will clear my ram and I don't have time to mail order one before I go back to work for a few weeks. The temp I can read is the overall "cpu temp" which I can get from the HW monitor in EasyTune6. (came with the board) I cannot read individual core temps in OCCT or AMD overdrive. I have read this is normal when you unlock the 4th core, that all your core temps will go away. Processor is at 34c as I type this, running prime 95 it will run around 60c at its hottest. It doesn't seem to matter what voltage or clock speed I set it to, that's all the hotter it really gets.

Any ideas why things would be so stable last night, and not this morning? I rebooted windows probably fifteen times last night for various reasons (tinkering) and I never had it hang once at 4x 3.1 ghz. I AM showing all four cores working in the XP hardware monitor.

*edit* it doesn't seem to be any more or less stable running on 3 cores than 4, fyi

*edit2* Came across elsewhere that VID was for voltage ID and that this setting and the processor setting are inter-dependant somehow. Can anyone clear this up for me? Thanks so much

 

[Yukmouth]

Put everything back to auto and narrow your adjustments to CPU/CPU NB, these are the two greatest contributing factors to stability and you'll probably find your answer in their manipulation upwards.

CPU NB VID is the *minumum* amount of voltage fed to the CPU, much like VTT voltage for the ram should be. There's no reason to mess with either. BSOD's mean NB or CPU voltage for me and I wouldn't imagine your chip should be all that different from mine (outside of that fourth core, which should require nothing more than more voltage so long as its not crippled in some way).

1.2v is too low for the NB, even at default clockspeed with only 3 cores in use. Also, be sure to report back your actual CPU voltage via the Hardware Monitor of your BIOS. Some motherboards can fall short .020v to .025v and require a bump up. The phenom 965BE has a default vcore of 1.40v, so you know your chip is safe up to that voltage. Same goes for the north bridge, 1.310 should be your max.

Here are my numbers:

At full load I never usually pass 44c at 3700mhz. Mind you, this is at close to 1.5v cpu voltage and a steady 1.287 NB voltage. On a hot day though, I'll see temps max out at 46c full load.

 

[James of the Mountains]

Thank you so much! Then my 1.1 NB voltage is WAY low. No wonder...

I'll do what you suggested and report back after a couple hours of testing. I'm stable currently @ 2.8x4, with the cpu vid @ +0.050 and everything else at the defaults (including my 1.1v NB). I'll switch that around so the VID is at default and feed the power to the cpu voltage, and raise the NB from 1.1 to somewhere more like where you have it and try some higher clock speeds again. Thanks for all the help

 

[alyarb]

of course. if your fourth core survived an aggressive break-in period (regardless of frequency), you can pretty much eliminate it from your "instability equation" and go on to investigate voltages, NB, memory etc as the remaining possible culprits. thankfully this was no big deal. glad you got a good unlock. once you've got your set of voltages in their harmonious sweet spot, you will likely be eligible for the same 3.8 ghz that the majority of X3 specimens achieve.

 

[Yukmouth]

Congrats on the X3 to X4 :beer:. I was hoping for that unlock also but I'll have to wait for a four core CPU as no amount of tinkering will unlock mine.

While that Northbridge is low, you never know what might just work from motherboard to motherboard. The lower your stable voltages the lower your temps, so you're smart to test it at default as you're doing. If you run into more issues, just adjust your voltages. It seems as though your CPU likes the vcore you added. If you get a blue screen or a failed core in prime95, up your northbridge as a next step, slowly of course.

 

[James of the Mountains]

That's what has been confusing me is that it loves prime95 and OCCT, and I'm passing all their tests, but when I leave it shut off for a long time and come back to it, that seems to be when windows is most likely to hang up on me, and only right after it starts and loads windows and the stuff that runs in the system tray. The cpu stress tests all come out fine.

The advice has been invaluable. When I get it all sorted out I'll post what I find so the next guy will have the info too

*edit* oh and I got the stress test temps down by moving the whole box to my living room where it's a tile floor and the biggest room of the house. I'm not hitting above 55c degrees in OCCT anymore. Turn out that the office space I had set up in the back bedroom was 15f hotter when I was running OCCT than the rest of the house. duh. *smacks forehead lol.