Why Voltage Increases Not Helping ?

This is an update on an older thread where I was at much higher voltages.

System:
Asus R2E
Mushkin 998692 w/ Ascent Cooling and 6-7-6-18 timings
i7-920
Prolimatech Megashadow

Working w/ my son on his new build. We have built several boxes together and I have been building for 20+ years but only boxes which had a primary duty as CAD machines, so never played much w/ aggressive OC'ing (read > 20%). Right now we are working on a stable OC about 3.7 with no "tricks", that is no shutting off BIOS features.

Have a very stable OC at 3.7 GHz, 176 BLK, 1.2000 CPU Voltage. 1.56 DRAM Voltage. Very satisfied as the goal for regular use was to hit 3.66 w/o disabling SpeedStep, HT or anything else. However, that doesn't mean we don't want to understand more about the intricacies of OC'ing and still looking to improve things.

Here's where we are puzzled. Leaving everything else as is (except voltage)....bumping the BCLK up to 177 fails....I used to have to wait 6 hours of Prime95 to discover this but now using OCCT which usually sets off the failure alarm with 22 minutes left in the 60 minute test.

Memtest passed 12+ hours BTW, at both 160 BCLK x 10 and 200 BCLK x 8 at 1.56 voltage.

So, in the jist of education .....

-Is there any value to lowering CPU Voltage below the 1.2 volts which is the low range of most guides we have read ? Core temps are 66/63/63/63

-What correlation has anyone seen between successful OCCT tests and Prime95 ?

-We are a bit stumped as to how we can be stable at BCLK 176 and those low voltages and yet at 177, no matter how far we have gone with voltage (up to 1.375 CPU / 1.64 DRAM), we can't get 1 single steenkin extra BCLK increment. This was done w/ prime 95 and a higher QPI/DRAM core voltage than we have now so I am redoing with the lower QPI voltage but using OCCT instead of prime95 (1 hours nicer than overnight) ....once I get a pass...will try P95 overnight. Haven't gotten past the 22/60 minutes left mark with 1.2125 / 1.225 / 1.2375 but am at 14 minutes left w/ 1.25 volts. Plan is if successful, to run P95 over night

-What effect if any does "power saving" features have on prime 95 tests....I often sat at the machine for 5-6 hours on web while testing w/ prime 95 w/ no failures ... go to bed and get a failure....so wondered if the sleep / wakeup was messin w/ things.

-I saw a post once and read that if you had a failure linked to [wish i remembered] adjust CPU voltage and if you had a failure linked to [again wish i remembered] adjust QPI voltage... system kaputed while I was reading and never got to save so would love a pointer or education in that regard.

Thanks in advance for the educational opportunity :)

PS, just looked up at OCCT and it completed at 177 and 1.25 volts....yeah for progress :)

But 180 is a no go ..... went to 1.25 / 1.30 and OCCT fails....1.35 BSOD's

Seems to me that the limit of this particular PSU has just been reached. That 1 BCLK increase cost me a whole 0.05 volts. And 0.15 volts to get 3 more just doesn't seem worth it.

BIOSsettings-1.jpg
 
Nice to see a dad helping a son with his overclocking an whatnot.. Not too many of those "dads" around.....

Well first of all not being able to hit a higher clock with a bit more voltage does not always mean you are doing something wrong, you might have hit the CPU's threshold on that specific board. As far as the i7 goes, I have only overclocked on an Asrock X58 Extreme and the bios values are fairly different that the RAMPAGE, overshocked can gladly help you with that question :

http://www.tomshardware.com/forum/page-253365_29_300.html
 

overshocked

Distinguished
Feb 14, 2009
1,999
0
19,960
Well, not sure if this is right, but sure as hell seems like you are reaching the threshold of the QPI.

Try raising the QPI and the VTT voltage. Tell me if that helps at all.

EDIT: OvrClkr, you have an i7 system now! [:lectrocrew:6]
 
I did notice that using the CPU level up feature bumped the QPI/DRAm voltage all the way up to 1.3875 .... while I have seen lots limits posted for CPU (1.45 typically) and DRAM (1.65) hardly ever see anything for QPI.

Ir there a relationship (read ratio) for raising QPI/CPU voltage .... proportionally ....1st one then the other ?