Cannot OC C2D E6300 @ ABIT AB9Pro above 310 MHz - help!

[fregimus]

Everybody,

Asking for your help 'cause I am stuck trying to overclock a new system specially part-listed and built to be overclocked

I have no overclocking experience, but I am an experienced EE, and I used to build many computers by myself.

Here's what I have: ABIT AB9-Pro board (intel P965); BIOS v 1(?).17 (their Web site has .16, but their flasher program directly pulled up one higher). C2D E6300 6/F/L2, 4x1MB G.Skill DDR2-800, sold as 5-5-5-15. Koolance Exos (original) water cooler (used on my older P4/2.4 box that failed after 6 years of 24hr work). There's also a branded ATI X1800 board in the PCIx16 slot, but I have not installed a driver for it yet, and XP uses a framebuffer driver by default. No other boards are installed.

I followed all steps in the C2D OC guide in the stickie. Here's what the essential, reproducible (but pityful) result is.

The machine boots and works @default 272MHz, 24hr+ under the load test of 4 forever-looping .vbs scripts that pull up 100% load on both CPUs. Ambient t=19..23C. Tcores=38C, Twater=29C.

Test1: Goal - to stop one step short of raising Fclk. Disabled "Thermal control", "Limit CPUID", "CIE", "Exec disable bit", "Virualization", "EIST". Set RAM timing to 5-5-5-15, Command Rate=Auto. Up VCPU 1.325 to 1.375V, VDDR 2.0 to 2.1V, VMCH 1.25 to 1.40V, VICHIO 1.50 to 1.60V. Lowered RAM clock divisor to 5:4, so that RAM won't be overclocked. In this configuration it runs (still 272 MHz, 10 min. test: Ambient = 20C; Tcores=39C; TWater=31C).

Test2: Minor overclock. Raised Fclk to 320MHz. System booted, froze up before SpeedFan could be started. Second attempt, froze on log-on. Duh!

Test3: Lowered Fclk to 310MHz. It's buzzing for 3+ hours now: Tambient = 21C, TCores=42C, TWater=33C).

These are the first results that I can call consistent, but for the last few days, it did not boot up stable @320MHz, and refused to boot at higher clock frequencies (three long green whistles up in the air; :!:; CMOS reset)

Peculiar: 1. CPUZ shows Vcore=1.213V with either 1.325 or 1.375V BIOS setting; 2. SpeedFan 4.32 does not show Tjunction, only 2 Tcore -s, contrary to the guide 3. Recommended ibid. Intel TAT does not start up on this system. I tried uninstalling ABIT's driver for their "µGuru processor", whatever it is. The program showed both Vcpu and (apparently) Tjunction, but messed with BIOS settings (yuck!), and leaked 500 handles a second (YUCK!!!); still removing it did not improve the working of any of others.

Any advice would be appreciated tha-a-at highly!

 

[dr_kuli]

First of all - I've never had an idea of OCing with not updated, non-optimized OS with default drivers. Second - TAT doesn't run on machines with disabled Thermal Control (TM control). Third - I had AB9 (non-pro) and it had no Vdroop at all (well, -0.02 max). I cannot remember all BIOS items, but there's something wrong with your Vcore or the board in general. It has nothing to do with 1. & 2. point. Fourth - this board doesn't run properly without uGuru driver.

Make clean install, latest drivers, optimize OS.
http://www.tweakhound.com/xp/installxp/installXP1.htm
http://www.tweakhound.com/xp/xptweaks/supertweaks1.htm

 

[fregimus]

Thanks for the reply dr_kuli,

First of all - I've never had an idea of OCing with not updated, non-optimized OS with default drivers.

Why is that? I am just trying to understand how the default VGA driver could prevent the system from booting up when set to 340 MHz in the BIOS...

Second - TAT does't run on machines with disabled Thermal Control (TM control).

Thanks — that's the great news! I'll try that at once.

Third - I had AB9 (non-pro) and it had no Vdroop at all (well, -0.02 max). I cannot remember all BIOS items, but there's something wrong with your Vcore or the board in general. It has nothing to do with 1. & 2. point.

Is it not too premature to conclude that there is something wrong with the board? It is also possible that there is something wrong with CPUZ, or something right with the board, when it does not do what it is not specified to? Take your own post as an example: you happen to run the same CPU and the same software as I do, and in your sample CPUZ shows 1.296V; I am sure you did not dial anything within ±0.02V of 1.296V in BIOS settings, did you?
On this topic — are there known test points on the board itself to connect a Vmeter to measure V on the output of the regulator? So far no software has shown me a figure I can trust. I did not try to get to the understanding of the Vreg (and hope I won't have to.)

Fourth - this board doesn't run properly without uGuru driver.

Not on the point of questioning that, but it is supposed to boot up when overclocked without the driver, right? I cannot get past that starting point...

 

[dr_kuli]

Why is that? I am just trying to understand how the default VGA driver could prevent the system from booting up when set to 340 MHz in the BIOS...

Because running non-optimized OS with default MICROSOFT drivers is like driving the cab from passengers' seat, when MICROSOFT is your cab driver. All you can do, is tell him where to go and HOPE he'll get you there ALIVE and through fastest, safe route. But it's only HOPE, we OCers need certainty. The list of possible issues is longer than this forum. And you DO use OS for testing, don't you?

I am sure you did not dial anything within ±0.02V of 1.296V in BIOS settings, did you?

Don't be sure. Of course, that I've set 1.375V in BIOS. That's Vdroop exactly.

On this topic — are there known test points on the board itself to connect a Vmeter to measure V on the output of the regulator? So far no software has shown me a figure I can trust.

Surely they're, but I don't know where exactly. And I remeber, that Abit uGuru app, used for monitoring, was trustful.

it is supposed to boot up when overclocked without the driver, right? I cannot get past that starting point...

So you know, why I suggested that something is wrong with the mobo in general. I've never heard of AB9 not booting with FSB 320 - with correct setings in BIOS, of course.

As I think of this, I'm pretty sure you didn't set Vcore manually. That's the point. And CPU/NB Strap, and PCIe speed, and PCI speed, and I have no idea what more. Am I right?

EDIT1: I'm sorry, you did set 1.375V, but what's with the rest?

EDIT2: CPU-Z 1.39 shows Vcore=1.213V constantly when it's higher than 1.45V. So it looks like BIOS is setting Vcore=Auto and upping it above 1.45V, no matter what you set. This is what I wanted to say, that something is wrong with Vcore.

 

[fregimus]

Bingo! The culprit was found. The Northbridge.

The original cooler on it was a brass plate, connected with a Perkins pipe to a large heatsink, sitting on CPU VREG MOSFETS. Since I am using water cooling, I removed the plate and the tube, and installed a water heat exchanges onto the chip. There was a layer of silicon(?) film (of the color that I'd describe as the darkest beige) in the original assembly, that I kept. That shown to have been the wrong move made; it is also possible that I tightened my assembly up to a pressure that was too low for this type of medium.

All in all, after I replaced the liner with some metal-laden paste, the system easily booted up at 395 MHz with all the same voltages except for Vcore = 1.400 V, showing Tcore=49C and Twater=33C at ambient 22C, although I had to back up to 380 MHz for stability reasons. This by far transcend my initial goal of a completely stable system at 350 MHz, so I am there.

CPU-Z 1.39 shows Vcore=1.213V constantly when it's higher than 1.45V.

Apparently, the limit depends on make of motherboard. I experimented lowering Vcore at low Fclk down to system instability, but never registered anything but 1.213V displayed. Although I did not directly meter the voltage, I noted a rising of Tcore when upping Vcore in the setup, which should serve as an indirect evidence of real Vcore growth. Nothing of that suggests that Vcore was less than 1.45V, of course, but if it was this high at the BIOS setting of 1.200V, then my top voltage test of 1.5V setting would perhaps evaporate the CPU, unless we assume that the dependence of "real" vs. "set" Vcore is highly non-linear, and that I hit its "sweet spot" when noted Tcore following the BIOS setting. An assumption that CPU-Z does not fetch the voltage correctly on this board sounds much more reasonably to me.

[…]running non-optimized OS with default MICROSOFT drivers is like driving the cab from passengers' seat, when MICROSOFT is your cab driver. All you can do, is tell him where to go and HOPE he'll get you there ALIVE

You are right, to a degree; consider, however, that there were billions of Windows installation in the world, and how complex it is to write software that works in such a huge variety of situations this number alone prompts, hang-up upon installation being a rare exception. Not to prove anything, but consider my software experience during this system installation: ABIT own control program leaked 1,500,000 handles already by the time when I found out "what was wrong" with the system; Intel Matrix Storage manager complained that the system was below minimum requirements for it (BS), and I was fortunate to recover only the driver from the package; Intel thermal test did not start up (unfortunately, your suggestion did not work, as did not any one of sensible BIOS settings affect it); ZCPU and SpeedFan installed and kind of worked, with reservations for the wrong Vcore indication in ZCPU, and the fact that SpeedFan displayed only 2×Tcore and 2×HD temperatures, no voltages); Intel chipset INFs installed, but killed all USB root hub drivers, requiring their manual reinstallation. Microsoft software looks sweet on this background, for it required no significant intervention and did its job.

All this being a matter of personal preference, if I were to choose whom should I put into the driver seat of your cab example among Microsoft, Intel, ABIT and the two individual authors of respective ZCPU and SpeedFan, then certainly I would have chosen Microsoft. I am fully agree with you that Microsoft makes imperfect software, but alternatives, when exist, are usually still even further from the perfection unattainable…

And thank you for your replies: in the area such "unofficial" as overclocking, any piece of information is self-valuable, and I do appreciate your time and effort on sharing it.

Edit: s/ZCPU/CPU-Z/

 

[dr_kuli]

I can't remember my settings on AB9, but it seems that uGuru chip driver is grabbing the whole info - that's why CPU-Z cannot read anything. As for Vcore - I observe it on Asus P5B-E, if it's higher than 1.45V CPU-Z reads 1.213V. All in all - it's great you've found the failing point, NB.