Finally time to overclock i7 920[Help please]

Associate
Joined
11 Jun 2006
Posts
1,569
i7 920 Bloomfield 920 C0
H-50 corsair cooler running push/pull with stock antec/corsair fan
Intel DX58SO
6GB Patriot Viper DDR3 PC3-12800 1600mhz
Powercolor HD4870 512mb (Needs overclocking too)
Thermaltake 750w PSU
Antec 900 (First version fans on full)

Current programs:
CPU-Z
GPU-Z
Coretemp
Realtemp

What are the best programs to bench with nowadays?
Haven't overclocked in about 4 years so how do I go about this again?

Idle temps are 28-34c

Lets do this :) thanks

EDIT - Limited by mobo as wont let me overclock much.
 
Last edited:
Intel Linx for stressing (20 passes - All memory) or good old Prime95 26.4.
Real Temp to monitor cpu temps.

There are loads of guides out there to overclock an i7 920 so I won't bore you with another ;).
 
Thanks for the information, I just remember reading something about the ram voltage possibly killing the CPU? Will get these 2 programs and let you know how I get on :) I was just posting to also check if there was any new info as I recall at the time of 2008 there was a new/different way to overclock as it involved the QPI and some other multiplier which I wasnt aware of what it did so didnt know if it was safe as people may have suggested it back then but ended up killing their chip after a year or something :D
 
Holy moly! Thats a hot CPU!

Do not overclock until you sort you temps out. Most i7 9XX cpu's require 1.3-1.35v to attain 4GHz (200x20). Yours is at 3GHz and already roasting. It's not just voltages that raise temperatures, clockspeed will also have a massive effect so if you overclock in your current voltage settings, you risk damaging the cpu.

Your temps should be a lot lower than that with the H50. Are you sure it is mounted correctly, fans on full?
 
Can you not try lowering the voltage a bit? 1.38 is very high for 3.0ghz, even on a CO/C1 stepping chip, admittedly though, theese versions do need a wee bit more voltage than a DO stepping cpu.
 
Ok I accidently linked the same image for both temps.

Stock temps are http://imageshack.us/photo/my-images/830/stocki7.jpg/

The temps were that high on purpose, I increased the vcore all the way to 1.38 assuming that was the max safe I could hit after checking the forum thread posted above. It showed that I was within 4c of tjmax so clearly was too hot so I took the vcore down to 1.25 and went from there. (I figured reach max vcore and then keep increasing the clock until it was stable but as you pointed out it was obviously too hot)

I am currently on 170 x 20 at 1.25vcore 1.66vram and 1.25 ICH? (The one below vram in memory controller) and I then had to up the QPI voltage from 1.1 to 1.2 to boot into windows after increasing the FSB from 165 > 170. Now currently stress testing for 2 hours while I am away to football.

http://imageshack.us/photo/my-images/851/clock1701251661251279c.jpg/
 
Well it wasnt stable with the settings above as I came back and my had hung, think it might have restart itself as it wasnt showing real temp or core temp just my desktop background with my default startup programs. Ideas or is it time to start upping the vcore maybe? I did a blend test on primte95 so see the screenshot above for settings

Want screenshots of my voltage settings?
 
Last edited:
just leave voltages to auto and set the multi to 19 and bclk to 180 and see where your at, this will give you 3.8 GHz, then if you want more start pushing the bclk up and keep and eye on temps and test with prime to make sure your stable...ive got my 920 C0 at 19 x 220 stable which gives me 4.180 GHz, also don't use xmp profiles, my ram is at 1780 MHz with these settings I think
 
Will try your suggestion cammy! Will up the vcore Moo as I said it was on 1.25 in bios but obviously undervolting by quite a bit.

I think a massive problem has been my idle state or pc going in to power saving mode as I did a test overnight on 160 x 20 and it passed 10hours:40 mins just fine with a max temp of 75c so thats definitely stable. I however woke up with it off or hung atleast so I thought. It turned out it was just in sleep mode or something! Disabling all that stuff in bios/windows now
 
Nope, if I can figure this out it finished stress testing after 10 hours 40 minutes and then powered down into some sort of sleep mode as when I booted back into windows all my stuff was still there and open with the timings/temps etc. It also asked me for a password so it must have come from sleep - It was a fresh install of windows 7. Either way set everything to performance.
 
Well, only thing I can suggest is the prime is failing. When each thread gets an error it stops running on that particlar one, when all fail it'll techinically be idle and therefore go into sleep mode?

Clutching at straws really, seems odd.
 
I thought it might have failed also but after 6 hours as I was at the computer from 9pm > 3am it was fine and stable. I fell asleep and after the above description I loaded and it had finished prime testing with 0 errors
 
Was unable to load windows after trying your settings cammy, it doesnt like that high BCLK and default voltage :S

Currently stress testing at:

BCLK: 170mhz X 20 multiplier (Turbo mode enabled so 21)
CPU: 1.35v (1.25 according to cpu-z) [Upped votlage to 1.35 for 1.25 because I was insure if my previous test was stable]
RAM: 1.66v (Stock 1.54)
QPI/Uncore Volt: 1.25 (Stock 1.150)
IOH: 1.225v (Stock 1.1)
QPI Data rate: Auto (4.8GT? dont know what these are exactly)
Uncore multiplier 16 (What is this exactly) afaik it must always be atleast 2x mem multi
Memory multiplier 6
QPI strap is 4.8gt always
QPI data rate increases to like 5.8k? is this safe as I cant keep it to 4.8 :'(

My bios says with turbo enabled it should be 22 multiplier but even when I have stock the 22 multiplier never kicks in it only goes on 21. Any ideas? Have checked bios and enabled eist speed step control which allows it to enable turbo settings.

Stress testing 3.5ghz at my max safe vcore for just under 2 hours now http://imageshack.us/photo/my-images/69/clock17013516612591c30m.jpg/

Edit - I had just cleaned my case completely and added a push/pull combination to my H-50 think its worth resetting as it seems extremely hot at that vcore compared to other peoples results. Its in an antec 900 case with everything on full!
 
Last edited:
mmmmmm, your board and ram may be holding you back a bit..... I would use the 19 multi I have read somewhere that this is the sweet spot for a 920 and I have found that to be the case, struggled to get mine over 4 GHz using any other multi now im nearly at 4.2, what I would do if I where you is clear your cmos so it puts it back to default, after choosing your primary hard drive and boot order etc. etc. and your happy the machine starts under stock conditions change the multi to 19 and the bclk to 160 and leave the rest to auto, keep notching up the bclk by 5 each time to see were this takes you.....when it freezes and or blue screens you try notching up the vcore then to say 1.35v and again see what happens...if it starts try adding another 5 to the bclk if not try upping the vcore to 1.36v and so on and so forth...... keep an eye on temps though very important......my vcore for 4.18 GHz is currently 1.43v but I have excellent cooling so can get rid of the heat.....you may want to change those stock fans for something with high static pressure x 2.....i use noctuas but there are many good fans :) let me know how you get on?
 
Back
Top Bottom