Overclocking i7 965 XE C0 stepping

Toggle sidebar Toggle sidebar
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540
Hey everybody,

I'm looking to make a good overclocking of my current config. I'm still a beginner so feel free to correct me.

CPU: intel i7 965 XE C0
MB: P6T deluxe v1
GPU: MSI 770 GTX Lightning
HDD: SanDisk 256GB Ultra plus
RAM: 12GB DDR3-1666MHz ANEON

I have been looking on that multiplier focused guide: http://www.tomshardware.com/reviews/i7-c0-d0,2386-5.html

And this BCLK one: http://www.overclock.net/t/538439/guide-to-overclocking-the-core-i7-920-or-930-to-4-0ghz

The BCLK guide is really detailled, but I still have many question when I'm reading it.

1 - Because of the multiplier limitation of the 920 CPU he recommends to start from 19 or 21 odd multiplier, my 965 has a default 24 one, should I lower it to match that guide ? Or should I take first advantage of the multiplier then adjust the BCLK ?

2 - Correct me if I'm wrong, but o/c the BCLK is o/cing the memory and the uncore which is x2 the memory. Isn't it better to go for the multiplier to preserve all the stuff that are uncore dependant ? Following this statement 145 * 27 would be better than a 163 * 24 ?

3 - I have 1666Mhz 9-9-9-28 AENEON RAM, currently configured at 133*12 ~1600MHz.
In the fist case, pushing the BCLK up to 145 will o/c them to at 145*12~1740MHz.
In the second case, pushing the BCLK up to 163 will give me the choice to make 163*10 (test on my motherboard which doesn't give me RAM multiplier but direct values). What's your viewpoint about this.

I'm pretty sure that I misunderstand something, because many people seems to avoid multiplier for the BCLK...

Thanks for your leads and insight.
 
Solution
h0007


Because I suspected that the memory was creating problems into the evaluation of the overclocking, I tried to digg a stable o/c with this time the multiplier.

29*133~3,857GHz

But I encountered the same problem, everything is fine for more than a hour, but during the 10 first minutes I have one worker that fail during the PRIME95 blend test.

In the first case I did suspect that the memory was creating noise into the evaluation of the problem.
But in this second case, not doubt that it's somewhere else.

I'll post my results as soon as I can.

Thanks again for helping me, I learn things everyday through...
Sort by date Sort by votes
Hard Line

Hard Line

Distinguished
Apr 15, 2008
931
0
19,160
i use a combination of the 2 ( i7 930) I used 21 multi then used bclock. on your cpu, i would start with 23 multi. even muti's are A. less stable
when overclocking, and B. need more voltage for some reason.
 
Upvote 0 Downvote
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540


Can you develop the reason for passing by multipliers first and then tuning with the BCLK ? Is it for not too much o/c the uncore that you are doing this ?



Ok that could be my starting point, ok for the even multi's, what about the odd ones, what is the main reason for advicing to start at 23 and not at 26 for example ?
 
Upvote 0 Downvote
Hard Line

Hard Line

Distinguished
Apr 15, 2008
931
0
19,160
the reason for the odd multipliers are stability over even multipliers. 26 is even. also I chose 21 because that is my max odd multiplier. if your cpu has a higher odd multiplier then I would use it as a starting point.
 
Upvote 0 Downvote
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540


Ok my bad I was thinking 25 or even 27. As an extreme edition the i7 965 is unlocked that's why I wanted to ask you why specificly 23 taking the fact that I could go higher.
 
Upvote 0 Downvote
damric

damric

Illustrious
Sep 17, 2010
8,017
72
37,990


Because I overclock to 5GHz :D

 
Upvote 0 Downvote
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540


So there is no good practice?

I mean taking back my example : 145 * 27 OR 187 * 23 would result into approximately the same ?

About your 200 BCLK, it is creating a 3.8GHz Uncore right ?
How do you know that you're not breaking a roof with that ?
What is the pros of switching on high base clock like yours ?
Uncore o/c perform better, Memory perform better ?

Sorry for asking too many questions but I try to understand in order to go into it without making random decisions :)

EDIT 1 : One of my goal could be to upgrade the cpu and ram performance to perform better into high end games for example

EDIT 2 :

I'm currently stressing the config with this :
BKL 133MHz
MULTI 27x
UNCORE 3.2GHz
MEM 1.6GHz
VCORE 1.16875V
QPI 1.275V
 
Upvote 0 Downvote
Hard Line

Hard Line

Distinguished
Apr 15, 2008
931
0
19,160
anything over 4.3ghz i think is pointless in games due to diminishing returns. though I have managed to get 4.5ghz out of my cpu. I have had my uncore up to 4ghz and my ddr3 1600 ram up to 2000 mhz

200x21 is 4.2ghz you will need more that that vcore if you want a good oc.

you will need higher qpi voltage as you increase bclock
 
Upvote 0 Downvote
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540


This is definitely not the same playground. Considering my previous messages have you got any suggestions / advices about my goal ?

I planned this :
CPU => 145*27~3.915GHz
CPU T => 90°C MAX (when torturing on prime95)
Memory O/C => 145*8~1.740GHz
Uncore => 3.480GHz

Thanks for both of you guys for helping me!
 
Upvote 0 Downvote
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540


BKL 133MHz
MULTI 27x
UNCORE 3.2GHz
MEM 1.6GHz
VCORE 1.16875V
QPI 1.275V

My main problem is that I have a watercooling without maintenance on my CPU (I mean you cannot change the liquid inside of it, and I do think that it performs not that really good.

With the configuration I'm torturing at 100% I'm currently 77°C 77°C 75°C 72°C is it still too much for 3.608GHz ?
 
Upvote 0 Downvote
Hard Line

Hard Line

Distinguished
Apr 15, 2008
931
0
19,160
under prime 95 you should be safe to 80-85C ( during normal operations it will be about 10C cooler ) which all in one liquid setup do you have? increasing multi will result in higher cpu temps than increasing bclock to overclock. so look for a middle ground
 
Upvote 0 Downvote
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540


This is the main problem I encounter, it's a computer originaly made by alienware, ALX X58, and the water cooling is hidden behing black metal plate.

Here is a bunch of photos I made of it: https://www.dropbox.com/sc/hq9s9jfj6n95n9l/AACPbs54sW0Llnd_mo155MR6a

I'm searching spec of it but since alienware was bought by Dell, this is kinda hard to find usefull things on the website.

EDIT: I find a slideshow where the watercooling system is visible (mine is the same except that it's located at the bottom front of my rigg) http://www.imageslides.com/Technology/gallery/10553-Photos-Cracking-open-Alienwares-Area-51-ALX#3
 
Upvote 0 Downvote
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540


1.8 * Bclk + 1, I'm sorry I didn't get that one.

to get the uncore I'm calculating that ways (mabe I'm wrong)
uncore = mem * 2
mem = bclk * (6 or 8 or 10 or 12 or 14 or 16)

your case 1 gives 173 * 10 * 2~3460MHz OR 173 * 12 * 2~4152MHz (Which sound not a correct value)
your case 2 gives 160 * 10 * 2~3200MHz OR 160 * 12 * 2~3840MHz
 
Upvote 0 Downvote
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540


Ok tomorrow I'll take the test with an other configuration (A or B presented below)

Comparing
A- 173*23 gives DRAM 1730MHz gives UNCORE 3460MHz
B- 160*25 gives DRAM 1920MHz gives UNCORE 3840MHz

Taking all you teach me A might be the less temperature solution, whereas the B might be more o/c.
The only things according to B I don't know is if it is really effective to o/c DRAM 1666 and UNCORE till that FQ and if the DRAM and the others uncore components will accept that kind of frequency. Do you have encountered problems according to that kind of uncore fq ?

As a methodology I think I'll make iteration to maximize the voltage

1st
25*143~3.575GHz
DRAM 1.716GHz
VCORE 1.16875V
QPI/DRAM 1.27500V

2nd
25*147~3.675GHz
DRAM 1.764GHz
VCORE ?
QPI/DRAM 1.3V

3rd
25*151~3.775GHz
DRAM 1.812GHz
VCORE ?
QPI/DRAM 1.3V

4th
25*155~3.875GHz
DRAM 1.860GHz
VCORE ?
QPI/DRAM 1.3V

5th
25*159~3.975GHz
DRAM 1.900GHz
VCORE ?
QPI/DRAM 1.3V

Any time crash by freeze or BSOD 101 I'll add +.025V to VCORE
Any time crash by BSOD 124 I'll add +.025V to QPI/DRAM

VDIMM 1.64V
PLL AUTO
LLC AUTO
EIST, Spreactrum, C1/0, Virtualization DISABLED
 
Upvote 0 Downvote
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540


I see, so it remains the x10 mem multiplier, that will give me the 1590MHz DRAM and 3180MHz Uncore. I have no idea of the impact of the o/c of the DRAM and the Uncore on encoding and gaming performance.

I'll search for it tonight.

 
Upvote 0 Downvote
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540


Ok I notice that there is a big difference between the VTT = DRAM BUS VOLTAGE I configure into the BIOS and the one shown by HWINFO64

Here is a screenshot: https://www.dropbox.com/s/hd170rnu7566x3z/Presse-papiers-1.jpg

What do you think about it ?

I'm currently stucked with my test :

25*143
DRAM 1740GHz
UNCORE 3441GHz
VCORE 1.16875V -> 1.19375V
QPI/DRAM 1.27500 -> 1,3 -> 1,325 -> 1.33750V

I find that weird that I'm stuck at this low fq, maybe I should downclock the memory to focus only on the cpu

Ok for 1:30 the test above was solid. I have one worker that fail in 10 minutes but the seven others remained ok.
 
Upvote 0 Downvote
h0007

h0007

Honorable
Jun 19, 2013
34
0
10,540


Because I suspected that the memory was creating problems into the evaluation of the overclocking, I tried to digg a stable o/c with this time the multiplier.

29*133~3,857GHz

But I encountered the same problem, everything is fine for more than a hour, but during the 10 first minutes I have one worker that fail during the PRIME95 blend test.

In the first case I did suspect that the memory was creating noise into the evaluation of the problem.
But in this second case, not doubt that it's somewhere else.

I'll post my results as soon as I can.

Thanks again for helping me, I learn things everyday through o/c, but this story of worker fail is really annoying because on the web the only advices are when there is a failure on small fft workers.
 
Upvote 0 Downvote
Solution
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom