dennyb

Shado--when any mobo is @ stock settings there is an order (ratio if you will) as to how the voltages relate to each other. I am pointing to V-Core ,MCH-VTT. I have found that loosely maintaining that ratio leads to a fairly decent overclock. So in that respect I am not understanding your view on high V-Core and way low MCH and VTT. Not digging at you just not agreeing

edit:1.4+MCH and VTT would scare me as well-I am looking at it from a lower level (1.26/1.3)

Shado

I'll give that a shot. I've pushed PLL up a notch to 1.57 and things "feel" better so far (no stutter in TF2).

Shado

pushing PLL to 1.57 didn't help, it still caused BSOD when folding. I tried 3.8 at a previously stable setting and it BSOD'd too. Is there possibly something about overclocking a CPU that makes the GPU become unstable?

I downclocked it back to stock and will let it fold while I get lunch. I have a feeling it's just an unstable OC even though it passed OCCT 10 hour (on high priority and large data set)

Shado

So today I downclocked everything back to stock for the lan party, partially to test if the stock settings work well.

Apparently, it does. I played for hours and let it fold while I went out for dinner and it worked out well.

To reiterate, I got my E8500 stable at 4.2 Ghz with the following bios setting, but for some reason was not stable when the GPU comes into play (gpu folding@home client, games like L4D or TF2 gets microstutter that doesn't exist in stock clock):

Robust Graphics Booster.........................[ Auto ]
CPU Clock Ratio.................................[ 9 ]
Fine CPU Clock Ratio............................[ 5 ]
CPU Frequency 4.2GHz...........................[ 444 x 9.5 ]

******Clock Chip Control******

CPU Host Clock Control..........................[ Enabled ]
CPU Host Frequency (Mhz)........................[ 444 ]
PCI Express Frequency (Mhz).....................[ 100 ]
C.I.A 2 [Disabled]

******Advanced Clock Control******

CPU Clock Drive.................................[ 800mv ]
PCI Express Clock Drive.........................[ 900mv ]
CPU Clock Skew..................................[ 0ps ]
MCH Clock Skew..................................[ 0ps ]

******DRAM Performance Control******

Performance Enhance [Turbo]
Extreme Memory Profile (XMP)....................[ Auto ]
(G) MCH Frequency Latch.........................[ Auto ]
System Memory Multipler.........................[ 2.40B ]
Memory Frequency 1066...........................[ 1066 ]
DRAM Timing Selectable..........................[ Manual ]

CAS Latency Time................................[ 5 ]
tRCD............................................[ 5 ]
tRP.............................................[ 5 ]
tRAS............................................[ 15 ]



******Motherboard Voltage Control******

Load-Line Calibration [Enabled]
CPU Vcore [ 1.250 ].............................[ 1.36250 ]
CPU Termination [ 1.20 ]........................[ 1.26 ]
CPU PLL [ 1.50 ]..................................[ 1.50 ]
CPU Reference [ 0.766 ].......................[ 0.798 ]

MCH Core [1.1 ]....................................[ 1.2 ]
MCH Reference [0.760 ]........................[ 0.798 ]
MCH/DRAM Refernce [0.900 ]..................[ .900]
ICH I/O [1.500 ]...................................[ 1.50 ]
ICH Core [1.100 ]...................................[ 1.1 ]

DRAM Voltage [1.800 ]...................[ 2.10 ]
DRAM Termination [0.900 ]...................[ .9]
Channel A Ref [0.900 ]...................[ .9]
Channel B Ref [0.900 ]...................[ .9]

So for some reason, while my CPU OC is stable for 10 hours OCCT (high priority, large data set), when my GPU comes into play it becomes unstable.

Does the southbridge ICH Core has anything to do when it comes to the GPU interface stability? I was suggested that I don't need 1.2V on ICH Core, but I have a feeling it could help.

Or does anyone have experience with OC'ing a CPU that somehow the GPU becoming unstable?

System spec:
E8500
Corsair XMS2 DDR2 2x2 gb 1066
Gigabyte EP45-UD3P
EVGA GTX 260 Core 216 SC 55nm
Corsair 650W PSU
Seagate 1TB HDD

Shado

I may have found the problem ... knowing that the GPU becomes unstable after overclocking, I tried to push my MCH Core up to 1.26 from the previously stable 1.20V, it didn't help, so I pushed it up a little more to 1.30, and it wouldn't even boot.

Against conventional wisdom (more volts!) I tried to down volt MCH back to stock 1.1V, and the system miraculously booted up without problem. I pushed PLL up a notch to 1.57 just for the sakes of it.

I am now folding for 2 hours without error so far. I'll post some update after I fold for a day. I just find it interesting that more voltage to MCH doesn't equal stability in my case.

str8wick3d

If your system can handle 3+ hours of gta4 then your stable. try it.

__________________

Shado

Don't have GTA4 But it looks like a fun game when I played it on my friend's PS3.

Mattb2e

What are the options available for your MCH frequency strap. Your NB has a set of timings just like your memory does, when you increase your FSB you need to increase the voltage supplied to the NB (MCH) in order to be able to regain stability at those timings, or loosening the timings.

The frequency strap does this for you, by selecting a different strap you tighten or loosen the timings in order to regain stability at higher FSB frequencys. My strap lets me choose from specific frequency settings however I believe that yours is in multipliers.

In order to achieve a 1066mhz ram speed I needed to run a tighter strap (333mhz) and thusly more voltage through my northbridge, the next strap selection for me would have been 400mhz but that would mean I would have to run my ram at 890mhz and loose some speed, I opted to run a higher NB voltage in order to maintain a higher ram frequency. However from the sounds of it too much MCH voltage for you makes you lose stability, I would think that ~1.2 should be fine but some experimenting with that will tell you for sure.

Here is a photo showing multipliers for different strap settings.

Attached Thumbnails

 

__________________
"Being wrong is erroneously associated with failure, when in fact to be proven wrong should be celebrated. For it is elevating someone to a new level of understanding, furthering awareness."

"Always question the recieved reality, the consensus reality is often intentionally misleading"

Thermal Compound Application Instructions:
http://www.arcticsilver.com/arctic_s...structions.htm

Intel E8400 Club

Shado

I have mine set to 2.4B, and my memory is running at the stock rated 1066 5-5-5-15 at 2.1V. I don't know what this has to do with NB, but I intentionally set it so my memory runs at its rated speed.

I lowered MCH Core to 1.1V and PLL back to 1.5V, and have ran OCCT (large data set, high priority) for 8 hours, and folding for 10 hours, and everything seems to be fine now.

For some reason, a high MCH core does indeed seem to be causing in stability in my case. I am now running 4.22 (444x9.5) at 1.33VCore.

Mattb2e

If you have a stability problems with NB or mem the best setting to use is medium dataset in occt or Blend in Prime95, it stresses Memory and MCH more than a large data set does.

As far as running 2.4b goes that means your running your strap at 333mhz with a 2.4x memory multiplier, the 2.0D or 2.6D is a 400mhz strap and will put less stress on your MCH as it will loosen your MCH timings. The only problem with running that is your memory will run below or above its rated frequency. You might be able to clock your ram higher than rated frequency but I would run it at or lower until you can ensure the rest of your system is stable.

Try running a medium data set test for 8 hours and see if you get any errors or bsod's.

__________________
"Being wrong is erroneously associated with failure, when in fact to be proven wrong should be celebrated. For it is elevating someone to a new level of understanding, furthering awareness."

"Always question the recieved reality, the consensus reality is often intentionally misleading"

Thermal Compound Application Instructions:
http://www.arcticsilver.com/arctic_s...structions.htm

Intel E8400 Club