I finally figured out why my overclocks on my current main rig is unstable... at DVinfo.net

Randall Leong · February 22nd, 2011, 08:41 PM

...I had the Vcore at only 1.225V at the 3.82GHz speed. Occasionally, the system locks up at the "Starting Windows" screen, and also get an occasional BSOD (Black Screen Of Death). And sometimes, when I reboot the PC, the BIOS in my Gigabyte X58 motherboard sometimes resets to auto (in this case, the speeds are as set, but the core and DIMM voltages get pushed up to the maximum Intel-advised settings of 1.375V (for a D0-stepping i7-950) and 1.65V, respectively). I reset all voltages to Auto, and indeed those same voltages are set automatically. The system ends up stable, but the CPU temperatures went way up.

So I did another experiment: Back off my overclock in increments. Also, I reset the Load Line Calibration setting from "Level 2" to "Standard". On Auto voltages, setting the memory to DDR3-1333 to DDR3-1600 sets the DIMM voltage to 1.648V. And as it turned out, setting the BCLK to even 150 MHz (3.6 GHz at the i7-950's Turbo multiplier of 24) resulted in the Vcore being increased above 1.3V - too close for comfort. And setting the BCLK to even 140 MHz resulted in a Vcore jump from its stock 1.5V to 1.275V.

I am re-testing my stability tests and temperture monitoring as I play around with the settings.

Rod Sandage · February 22nd, 2011, 10:45 PM

I am rather new at the overclocking but with I7-980 I work the voltages down to1.256v cpu--- 1.275v v-core and my memory was rated at 1.55 volts bumped it to 1.57dram. With a good air cooler all cores run below 57*C at 100% load at 4 ghz. Dropping voltage lowered temps a good 5* Just had to keep playing with it and retest till stable. It would work fine at the higer auto settings but ran warmer and would probably reduce long term relibality.

Rod

Randall Leong · February 23rd, 2011, 12:09 AM

It might also be the motherboard: It ran unstably at lower-than-auto settings. Shame on me for picking a $200-ish motherboard with only 8-phase power instead of a $370 motherboard with 24-phase power: It turned out that with that $200-ish motherboard, I needed a voltage bump even with a slight overclock.

Randall Leong · February 23rd, 2011, 12:18 PM

I also noticed that at stock Auto settings, even enabling Turbo Boost (and therefore the 24x multiplier) at the stock 133MHz BCLK jacked up the Vcore from 1.136V to 1.216V.

Panagiotis Raris · February 24th, 2011, 01:25 PM

im at 1.24 at 4.0 on an i7 950, and i run it as high as 1.31 for 4.3.

supposedly its safe to run it up to 1.4, but i wont go over 1.32 myself.

Randall Leong · February 25th, 2011, 12:57 AM

Quote:

Originally Posted by Panagiotis Raris

im at 1.24 at 4.0 on an i7 950, and i run it as high as 1.31 for 4.3.

supposedly its safe to run it up to 1.4, but i wont go over 1.32 myself.

At least with this Gigabyte X58A-UD3R mobo, my 950 crashes (BSODs) on Windows boot at 1.25V at 4.0 GHz.

Rod Sandage · February 25th, 2011, 06:00 AM

I am also running an asus motherboard also, Rampage 3 Formula, may be a difference in motherboards - bios - or even the processors themselves. I have read there can be a difference as to where the chip falls in regards to specs, some at the low end of the scale some at the higher end. This is probably why several of the combinations I ran that other people used with very similar hardware didn't work for me. Your mileage may vary blah blah.

Randall Leong · February 25th, 2011, 02:06 PM

Quote:

Originally Posted by Rod Sandage

I am also running an asus motherboard also, Rampage 3 Formula, may be a difference in motherboards - bios - or even the processors themselves. I have read there can be a difference as to where the chip falls in regards to specs, some at the low end of the scale some at the higher end. This is probably why several of the combinations I ran that other people used with very similar hardware didn't work for me. Your mileage may vary blah blah.

I decided to perform yet another experiment: Setting both the Vcore and the QPI/Uncore voltage to around 1.25V (reverting the Load Line Calibration setting to Level 2 for the least amount of Vdroop) with the CPU Bclk set at 160MHz and the maximum Turbo multiplier to 24x. (This results in a 3.84GHz Turbo clock speed.) I had set the voltages on the CPU too low for my selected overclock. I also boosted the IOH voltage to 1.24V. (The DIMM voltage remains manually set at 1.5V.) The reason why I chose a BCLK of 160MHz is because I have 1600-speed-rated memory, and setting the BCLK above that would have forced my particular RAM to run at the slower 1300-ish speed class.

I will be giving the system 10 runs of Intel Burn Test (LinPack), and see if it works out. (Apparently, I did not give the system enough runs of IBT when I had it set at 1.225V and 1.195V for the Vcore and VTT, respectively.)

UPDATE #1: At these settings, the system passed 10 iterations of IBT but crashed during the PPBM5 benchmark tests (the system simply rebooted). I am now giving the CPU a stated Vcore of a labeled 1.275V (1.248V according to the VRM, compared to the 1.232V at the previous setting).

Randall Leong · February 28th, 2011, 07:58 PM

Another update:

After playing around with some of the voltage settings and replacing the CPU cooler (the new cooler I got is a COGAGE TRUE Spirit, which is a cheaper version of the Thermalright Ultra 120 Extreme), I finally got a stable overclock to 4.0GHz. I had certain voltages set too low all along. And further reading about the BSOD code 0x116 reveals either a video card driver issue or an IOH voltage that's set too low. And the uncore/VTT was also set too low for my selected overclock. So, I set the BCLK to 191MHz, the multiplier to 21x, the RAM divider ratio to 8x (this results in a memory speed of DDR3-1528), the Vcore to 1.275V, the QPI VTT set to 1.355V, the QPI PLL to 1.880V, the IOH and ICH core voltages to 1.2V each and the DIMM voltage to 1.66V (everything else remains on AUTO, which on this motherboard automatically readjusts the voltage to better suit the operating conditions).

I am going to leave it on this setting until something goes a bit unstable again.

Rod Sandage · February 28th, 2011, 09:10 PM

Sounds like yor on the right track. How are the core temps when fully loaded?

Randall Leong · February 28th, 2011, 09:30 PM

Quote:

Originally Posted by Rod Sandage

Sounds like yor on the right track. How are the core temps when fully loaded?

Lower 80s C with Intel Burn Test 2.50 set to eight threads and maximum memory (as measured by RealTemp 3.40). That is as high as I will go, given the overclocking guides recommending against stress-test temps above about 84°C.

Still, that is much better than my previous attempts at 4.0GHz, which (at settings deemed "stable") resulted in the CPU core temps exceeding 90°C.

Randall Leong · March 4th, 2011, 07:47 PM

I spoke too soon. My 4.0GHz attempt resulted in a BSOD code 0x124 in the middle of transcoding in CS5. And my VTT is already at Intel's maximum of 1.35V while the core voltage can't be raised above 1.275V without seriously overheating the CPU (90°C+ temps).

Rod Sandage · March 4th, 2011, 08:40 PM

I had been using prime 95 and tried intel burn for testing. My temps max at 60* all fans on low and 57* fans on high. my antec 900-2 case has 5 fans an I,m using a Noctua NH-D14 cooler it is large and just fits in case. I would think if you could get temps down a little more it would probably be stable. Is there a way to feed more air to the cpu cooler or add another fan on the other side of the cooler its self?
Rod

Randall Leong · March 4th, 2011, 09:33 PM

Quote:

Originally Posted by Rod Sandage

I had been using prime 95 and tried intel burn for testing. My temps max at 60* all fans on low and 57* fans on high. my antec 900-2 case has 5 fans an I,m using a Noctua NH-D14 cooler it is large and just fits in case. I would think if you could get temps down a little more it would probably be stable. Is there a way to feed more air to the cpu cooler or add another fan on the other side of the cooler its self?

I figured out the 0x124 BSOD code: Vcore or VTT too high or too low. It seems to be working better after I lowered the VTT (QPI/Uncore voltage) from 1.355V to 1.315V.

Rod Sandage · March 5th, 2011, 12:00 PM

Do you have turbo, TM Load or C1E enabled these allow the processor to throttle up and down for load. I do have them enabled (processor runs slower at idle and low loads) but have read that it can cause problems. You may try disabling these they may be called something different in your bios. This locks the processor at one speed (multiplier) continouesly

February 22nd, 2011, 10:45 PM	#2
Rod Sandage New Boot Join Date: Apr 2009 Location: Bloomington,IL. Posts: 8	Re: I finally figured out why my overclocks on my current main rig is unstable... I am rather new at the overclocking but with I7-980 I work the voltages down to1.256v cpu--- 1.275v v-core and my memory was rated at 1.55 volts bumped it to 1.57dram. With a good air cooler all cores run below 57C at 100% load at 4 ghz. Dropping voltage lowered temps a good 5 Just had to keep playing with it and retest till stable. It would work fine at the higer auto settings but ran warmer and would probably reduce long term relibality. Rod

February 23rd, 2011, 12:09 AM	#3
Randall Leong Major Player Join Date: Nov 2009 Location: Melrose Park, Illinois, USA Posts: 936	Re: I finally figured out why my overclocks on my current main rig is unstable... It might also be the motherboard: It ran unstably at lower-than-auto settings. Shame on me for picking a $200-ish motherboard with only 8-phase power instead of a $370 motherboard with 24-phase power: It turned out that with that $200-ish motherboard, I needed a voltage bump even with a slight overclock.

February 23rd, 2011, 12:18 PM	#4
Randall Leong Major Player Join Date: Nov 2009 Location: Melrose Park, Illinois, USA Posts: 936	Re: I finally figured out why my overclocks on my current main rig is unstable... I also noticed that at stock Auto settings, even enabling Turbo Boost (and therefore the 24x multiplier) at the stock 133MHz BCLK jacked up the Vcore from 1.136V to 1.216V.

February 24th, 2011, 01:25 PM	#5
Panagiotis Raris Major Player Join Date: Apr 2010 Location: Reading, PA USA and Athens, Greece Posts: 269	Re: I finally figured out why my overclocks on my current main rig is unstable... im at 1.24 at 4.0 on an i7 950, and i run it as high as 1.31 for 4.3. supposedly its safe to run it up to 1.4, but i wont go over 1.32 myself.

February 25th, 2011, 06:00 AM	#7
Rod Sandage New Boot Join Date: Apr 2009 Location: Bloomington,IL. Posts: 8	Re: I finally figured out why my overclocks on my current main rig is unstable... I am also running an asus motherboard also, Rampage 3 Formula, may be a difference in motherboards - bios - or even the processors themselves. I have read there can be a difference as to where the chip falls in regards to specs, some at the low end of the scale some at the higher end. This is probably why several of the combinations I ran that other people used with very similar hardware didn't work for me. Your mileage may vary blah blah.

February 22nd, 2011, 08:41 PM	#1
Randall Leong Major Player Join Date: Nov 2009 Location: Melrose Park, Illinois, USA Posts: 936	I finally figured out why my overclocks on my current main rig is unstable... ...I had the Vcore at only 1.225V at the 3.82GHz speed. Occasionally, the system locks up at the "Starting Windows" screen, and also get an occasional BSOD (Black Screen Of Death). And sometimes, when I reboot the PC, the BIOS in my Gigabyte X58 motherboard sometimes resets to auto (in this case, the speeds are as set, but the core and DIMM voltages get pushed up to the maximum Intel-advised settings of 1.375V (for a D0-stepping i7-950) and 1.65V, respectively). I reset all voltages to Auto, and indeed those same voltages are set automatically. The system ends up stable, but the CPU temperatures went way up. So I did another experiment: Back off my overclock in increments. Also, I reset the Load Line Calibration setting from "Level 2" to "Standard". On Auto voltages, setting the memory to DDR3-1333 to DDR3-1600 sets the DIMM voltage to 1.648V. And as it turned out, setting the BCLK to even 150 MHz (3.6 GHz at the i7-950's Turbo multiplier of 24) resulted in the Vcore being increased above 1.3V - too close for comfort. And setting the BCLK to even 140 MHz resulted in a Vcore jump from its stock 1.5V to 1.275V. I am re-testing my stability tests and temperture monitoring as I play around with the settings.

February 28th, 2011, 07:58 PM	#9
Randall Leong Major Player Join Date: Nov 2009 Location: Melrose Park, Illinois, USA Posts: 936	Re: I finally figured out why my overclocks on my current main rig is unstable... Another update: After playing around with some of the voltage settings and replacing the CPU cooler (the new cooler I got is a COGAGE TRUE Spirit, which is a cheaper version of the Thermalright Ultra 120 Extreme), I finally got a stable overclock to 4.0GHz. I had certain voltages set too low all along. And further reading about the BSOD code 0x116 reveals either a video card driver issue or an IOH voltage that's set too low. And the uncore/VTT was also set too low for my selected overclock. So, I set the BCLK to 191MHz, the multiplier to 21x, the RAM divider ratio to 8x (this results in a memory speed of DDR3-1528), the Vcore to 1.275V, the QPI VTT set to 1.355V, the QPI PLL to 1.880V, the IOH and ICH core voltages to 1.2V each and the DIMM voltage to 1.66V (everything else remains on AUTO, which on this motherboard automatically readjusts the voltage to better suit the operating conditions). I am going to leave it on this setting until something goes a bit unstable again.

February 28th, 2011, 09:10 PM	#10
Rod Sandage New Boot Join Date: Apr 2009 Location: Bloomington,IL. Posts: 8	Re: I finally figured out why my overclocks on my current main rig is unstable... Sounds like yor on the right track. How are the core temps when fully loaded?

March 4th, 2011, 07:47 PM	#12
Randall Leong Major Player Join Date: Nov 2009 Location: Melrose Park, Illinois, USA Posts: 936	Re: I finally figured out why my overclocks on my current main rig is unstable... I spoke too soon. My 4.0GHz attempt resulted in a BSOD code 0x124 in the middle of transcoding in CS5. And my VTT is already at Intel's maximum of 1.35V while the core voltage can't be raised above 1.275V without seriously overheating the CPU (90°C+ temps).

March 4th, 2011, 08:40 PM	#13
Rod Sandage New Boot Join Date: Apr 2009 Location: Bloomington,IL. Posts: 8	Re: I finally figured out why my overclocks on my current main rig is unstable... I had been using prime 95 and tried intel burn for testing. My temps max at 60* all fans on low and 57* fans on high. my antec 900-2 case has 5 fans an I,m using a Noctua NH-D14 cooler it is large and just fits in case. I would think if you could get temps down a little more it would probably be stable. Is there a way to feed more air to the cpu cooler or add another fan on the other side of the cooler its self? Rod

March 5th, 2011, 12:00 PM	#15
Rod Sandage New Boot Join Date: Apr 2009 Location: Bloomington,IL. Posts: 8	Re: I finally figured out why my overclocks on my current main rig is unstable... Do you have turbo, TM Load or C1E enabled these allow the processor to throttle up and down for load. I do have them enabled (processor runs slower at idle and low loads) but have read that it can cause problems. You may try disabling these they may be called something different in your bios. This locks the processor at one speed (multiplier) continouesly