i7-920 can not get to 4GHz

i'd prime it for about 2-3 hours, honestly i don't think it really matters much after that, and as such i don't prime for more than that amount of time before getting on and using the pc :)

to give you an idea, i went for a 4.2 clock, primed it every time i increased Vcore, if it crashed increase and/or increase qpi, maybe trying different qpi pll, not very scientific, but i got it stable at that with little faffing...

temps are fine, people moan about over 80, i honestly don't think these chips worry at 90. weather is a bit warm, so that'll add to it, but i would try and get Vcore qpi down a bit now, and test more with prime.

ram, leave it as is if it's stable, don't think you've got any problems with ram stability.

set uncore to 17 as well, found that worked better than an even number ;)
 
Thanks.

So 17 is better than auto for uncore? (was going to switch to auto)

Also I decided to leave the QPI at 1.34V for the time being and begin reducing the Vcore 2 notches at a time and see what happens.

Will I notice any speed difference between 6 and 8 ram setting? (a bit of a noob on ram tbh) (or should I switch to Auto?)
Also
DRAM Timing Selectable (SPD) .......... [Quick] -- dunno what this means... is it fine?
Performance Enhance ...................... [Standard] -- whats the difference between this and turbo? should I leave it standard?

Thanks again :)
 
Last edited:
Using x17 uncore is fine if you are having issues, but I've not had any problems with using Auto. Got two i7 920 systems at 4GHz running 24/7 with almost all settings on Auto. The few things I changed were vcore, qpi voltages, ram voltages and timings.

Standard/Turbo etc is (as far as I'm aware) how tight auto ram timings are set, so it can be ignored if you have manually set it to the rated timings (on the side of the module) using the Quick menu. I usually stick it on Standard regardless.

To stress test it I use 10 runs of LinX (all memory) first and then 1 hour of Prime95 blend if it passes LinX. If it passes both it is more than likely stable so you can tweak the settings down and go again. I've found that some errors show up quicker in LinX and some in Prime95 so it pays to do short bursts of both one after another for a quick stability test.

Of course as you get to voltages you are happy with, longer runs of both (overnight is good if possible), plus things like 3DMark Vantage (or gaming) will ensure total stablity.
 
I set memory multiplier to Auto, Uncore to Auto, increased Dram Voltage to 1.66

Qpi still at 1.34V and Vcore at 1.35V

PC froze after 9 mins of prime95

Next step: Increase the Vcore?

Edit -
Increased Vcore to 1.375V - froze after 5 seconds in desktop
Increased Vcore to 1.38V - froze after 5 seconds in desktop
Increased Vcore to 1.4V - froze after 5 seconds in desktop and then kept on restarting... by some miracle i went into bios again with a warning about voltages.

So now, running my last stable OC with some differences which is
QPI - 1.34V
Vcore - 1.35V
Uncore - Auto (as opposed to x16)
Memory Multiplier - 6.0 (on Auto which is actually 8.0 I never got it stable) The only annoying thing is using 1154 MHz when the ram is rated for 1600MHz
DRAM Voltage - 1.66V (as opposed to 1.64V)

After this if it is stable for at least 2-3 hours I will decrease the Vcore and go from there.
 
Last edited:
You seem to be getting same problems as me Nameless 87 I hope that we both start getting some better results soon - its so frustrating...

Tried decreasing QPI from 1.375 in stages all the way back to 1.295 and it is clear that the system is most stable at 1.315. I also tried increasing from 1.375 to 1.395 and 1.415 and the machine froze the instant I started Prime 95.

I am therefore setting my QPI back to 1.315 and increasing the vcore up from 1.375 to 1.395
 
didgemaster, what is your memory multiplier at?

I am just trying 8.0 again but I set the IOH and ICH Core to 1.2V and the QPI PLL to 1.18, QPI 1.34V (anything higher is red), Vcore 1.35V. Uncore and QPI Link are Auto.

4 mins in prime 95 and it pc rebooted again!

I have the base overclock, I will just tweak the vcore and qpi down from what they are and with Memory multiplier being 6 it is stable...

However I want to get the full potential of the ram so 8... but don't know what to do to keep it stable :/
 
Nameless_87 said:
didgemaster, what is your memory multiplier at?

I am just trying 8.0 again but I set the IOH and ICH Core to 1.2V and the QPI PLL to 1.18, QPI 1.34V (anything higher is red), Vcore 1.35V. Uncore and QPI Link are Auto.

4 mins in prime 95 and it pc rebooted again!

I have the base overclock, I will just tweak the vcore and qpi down from what they are and with Memory multiplier being 6 it is stable...

However I want to get the full potential of the ram so 8... but don't know what to do to keep it stable :/
Click to expand...

Well at least you are making progress. Sounds like you may need more QPI voltage for 1600MHzish. Try putting QPI voltage on Auto with ram divider on 8 and see if it is stable.

I had a similar problem with this system and it turned out that the memory controller needed higher voltage to remain stable.
 
Speed said:
Well at least you are making progress. Sounds like you may need more QPI voltage for 1600MHzish. Try putting QPI voltage on Auto with ram divider on 8 and see if it is stable.

I had a similar problem with this system and it turned out that the memory controller needed higher voltage to remain stable.
Click to expand...

I tried that 3 hours ago.... did not manage to get it stable... pc froze in 5 seconds on desktop.

I could leave the RAM at 6x... it is still running 1146MHz opposed to the 1066 I was running before the overclock. Is that good or is it better to get it to 1528 (8.0). Would I notice any difference in speed?

Thanks
 
Nameless_87 said:
I tried that 3 hours ago.... did not manage to get it stable... pc froze in 5 seconds on desktop.

I could leave the RAM at 6x... it is still running 1146MHz opposed to the 1066 I was running before the overclock. Is that good or is it better to get it to 1528 (8.0). Would I notice any difference in speed?

Thanks
Click to expand...

Ah right you never said hence me mentioning it again.

You wouldn't really notice it to be honest, if you wanted it running at 1528 you need to do some more tweaking. Maybe save your settings (F11/F12) and go back to stock CPU (load optimised defaults), set memory to 1600, voltage to 1.66 and manually set the timings. Then run Memtest86+ from a USB stick etc to make sure that your memory can run at its rated speed without any errors.

If you've been running it at 1066MHz it is possible that it can't and that is where your problem lies. If you get errors, up the QPI voltage or use the memorys built in XMP profile.
 
Ran memtest as well... no errors on 1146MHz (only ran 2 loops)

But since I won't notice much difference I think I'll leave it as it is...

Am now into 1 hour prime with these settings:
QPI - 1.34V
Vcore - 1.3275V
Uncore - Auto
Memory Multiplier - 6.0
DRAM Voltage - 1.66V (as opposed to 1.64V)

So I began reducing the Vcore.

Next step is reducing the Vcore to 1.3V and DRAM to 1.64V as I don't need it at 1.66V since running at 1146MHz right?

Also what about QPI should I try reducing it or keep reducing the Vcore for now?

Thanks
 
Well I meant run it at 1600MHz, not 1146MHz as you know it is stable at that. You could try upping the divder to 8.0 and running memtest to see if you get any errors with the overclock.

Thats correct with regards to ram voltage, you could go even lower I expect.

I'd focus on vcore first, as you are running the memory at 1146MHz you can likely get away with a lower QPI anyways. Just do that afterwards.
 
Yeah!

I will leave it for 24 hours once I find the right settings... that way it will be stable for sure right?

If I just could run the memory at 8.0! But I don't think it is worth the extra voltage to the cpu just so memory runs higher... still unsure about this
 
Well the only way to ensure stability is just to use the system as you normally would for a week or so, if you get problems then it isn't stable. I tend to use LinX (All Memory) for 50 Runs and then Prime95 Blend overnight (9 hours) just to be on the safe side. Theres no huge advantage to use stability testing programs for long stretches of time (24 hours etc) as usually 2-3 hours is a good indication it is stable.

With regards to the memory the only thing I can suggest is to keep at it, make sure it can run 1600MHz at stock CPU speeds using Memtest86+ and then back to overclocked settings and keep upping the QPI. This particular CPU I had was unstable with 1528MHz until 1.460 QPI, even when using QPI PLL, although unlike yours it was stable with auto QPI Voltage and gave BSOD rather than freezing so that may suggest you have an under lying issue with the memory.
 
Last edited:
ok will memtest 1600MHz on stock cpu with everything set to auto.

The thing about raising qpi is that from 1.36V onwards it is red... so I am not supposed to increase it further than 1.34V otherwise I might damage the components is that correct?
 
The key thing to remember about overclocking is pretty much any voltage over the Intel standard/auto values carries some risk. The colours are there just to inform you, now obviously you want the lowest voltage possible but some CPU/RAM/Motherboard combinations do require much more voltage than others.

As far as I'm aware the Intel Spec is 1.35v QPI with 0.5v or less difference between RAM voltage and QPI voltage, but going over slightly is unlikely to cause any issues. I've seen people run 920s at 1.4-1.45v since release with no ill effects.

It is possible that going with higher QPI won't help, but you won't know unless you try.
 
forgot to ask one thing...

Right now I have one HDD which is a Samsung F3.

Next week after successful and stable overclock at 4GHz I will buy another F3 and an Intel SSD 320 Series 160GB.

I will format both F3s and run them in RAID 0 but the SSD will be my main boot and application drive.

I won't need to readjust the settings of the overclock when I add these right? It will just work I presume?

Thanks

On a side note... Memtest running on 1600hz on stock cpu.... so far no errors... just started 2nd loop. Although tbh since I won't see that much difference I think I'm going to leave it at 6.0, uncore on Auto, and reduce the Vcore further, cause it seems that increasing the Vcore and the QPI just for the ram to reach 1528 is not worth it.

Btw, Vcore 1.275V, QPI - 1.34V - stable for an hour under prime 95. Now I will try Vcore 1.25V and see what happens. Question: When should I try to decrease the QPI?

What do you think of the above? Is it going well if I can get the Vcore decreased and memory running at 6.0? I think performance would be better than running everything at stock like I was doing before.
 
The overclocking settings are stored on the motherboards CMOS chip, not the hard drive so you won't have to change them if you switch drives. Just make sure to save the settings using F11 when you get to something you are happy with and that is stable. That way if the BIOS is cleared by mistake then you can load your settings with F12 very easily.

Good to hear, yes you are likely right the difference won't be noticable so I wouldn't worry about it.

When you get to a lowest stable Vcore I'd start lowering QPI, sometimes QPI and Vcore can be a balancing act, you may have to add a bit more QPI to keep it stable but it is better to have lower Vcore if that is the case.

Make sure to test with LinX as well as Prime95: http://www.youwatched.com/datajay/linx(0.64).zip (They both have their merits and find certain types of errors quicker than others so I use short bursts of each, usually LinX all memory first and then Prime95 blend)
 
You must log in or register to reply here.
Back
Top Bottom