Core i7/X58 Overclocking Thread

[bfghornet]

Hello all,

I have 2 questions.

1. My bios says my qpi speed is 4220. Is it safe to run it that high? Its stable.

2. Even though it says 4220 in the bios CPUZ says its 3798. Which one am I to believe?


Thank you guys for any help on this matter.

EDIT:Forgot to add that my qpi/vtt is at 1.335 and dram is at 1.66

[icecold316]

All it does is increase the vdroop you are seeing when you disable it. If you were on the edge of stability with it on you will definitely crash with it off unless you raise vCore.

I finally decided to endup with 3.8 Ghz for 24x7 with CPU-z load voltage for 1.2V with max Linx load temp of 70C.

I was able to drop my CPU voltage by 0.5V, just with a BIOS update to F5E.

4 Ghz is possible with high voltage and temperature were upto 83C, so I decided to stay at 3.8Ghz, which is pretty good, I believe so.

[286SX]

What is ur QPI Voltage. I guess for Stability of I7, both CPU Voltage and QPI Voltage plays a big factor.

Temp 90 seems too high for 1.325V though.

Updated info in my previousl post:
http://www.xtremesystems.org/forums/...postcount=2464

QPIv was set at 1.225. PLLv = auto.

Getting 4.2 stable on air cooling is a real chore. But it seems like my chip could handle 4.2 (and maybe more) under water cooling...

-td

[Lukee]

Is it me or does the Asus P6T Deluxe leave you wanting more control for OCing?

The Gigabyte board has Qpi/pll and a few more voltages and I don't see them on the P6T.

/sigh

[ChaosAD]

Im planning to run my new 920 on water. Ill try to go for 20x200. IF i manage to run stable at this speed, do i have the option to run the memory at 1600, 1800 and 2000Mhz or am i limited in any way? I ask cause i just cant decide with what kind of memory kit i should go for.

[Lukee]

Im planning to run my new 920 on water. Ill try to go for 20x200. IF i manage to run stable at this speed, do i have the option to run the memory at 1600, 1800 and 2000Mhz or am i limited in any way? I ask cause i just cant decide with what kind of memory kit i should go for.

If you go 200 bclk, then you can run memory at 1600 or 2000. There is no option for 1800 at that BCLK.

I've run memory at 1600, but not 2000 with that bclk, I may be wrong but I don't know if you'll be able to get it stable.

[MasterYoda34]

Memory at 2000MHz would require uncore at 4000MHz. Not likely to happen.

[ChaosAD]

Can u explain why it is not posible plz? Do you have a link maybe? So the only option with 200 bclk is to run memory at 1600?

[rge]

you will have 6x multi, 8x multi, 10x multi, etc available. There is no 9x multi, so no 9x200 or 1800. So for 200 bclk, you can run 1600 and 2000. For example on my cpu i7 940, I can run 1600 mem (8x mem x 200 bclk) with 1.65v DRAM and 1.33qpi/vtt. Both of those are near max intel recommended volts for memory and qpi/vtt, but fine for 24/7 use. Uncore is 17x or 18x, for 3400 mhz or 3600 mhz uncore.

Uncore must be 2x memory or no boot. So now try 2000 mem. To run 2000 mem (I do for benching), I must use 10x mem multi 10x200 bclk, and thus 20x uncore at minimum. I need 1.64 qpi/vtt and 1.96 DRAMv to do so, way outside intel recommended limits, and that is for benching stability (may need more v for 24/7) albeit arena chess, which is like stress testing for 5 minutes....and I use 22x uncore or 4400 mhz, though could use 20x which would be little easier. So if you are willing to run high volts for 24/7, then you can run 2000. Even if unlike me, you buy ram rated for 2000 which will run at 1.65 v DRAM, it still requires high VTT/qpi and probably still high DRAM to drive uncore at 4000 min. But you must assume the risk of high volts, and the benefit is next to nothing in real world apps between running 1600 and 2000, as that has been shown time and again. Even benching you are talking 1% improvement in only some benches, not all, if that...1% makes diff in benching, but not so much on 24/7.

[ChaosAD]

Your post is very clear. You explained it very well.

[bfghornet]

you will have 6x multi, 8x multi, 10x multi, etc available. There is no 9x multi, so no 9x200 or 1800. So for 200 bclk, you can run 1600 and 2000. For example on my cpu i7 940, I can run 1600 mem (8x mem x 200 bclk) with 1.65v DRAM and 1.33qpi/vtt. Both of those are near max intel recommended volts for memory and qpi/vtt, but fine for 24/7 use. Uncore is 17x or 18x, for 3400 mhz or 3600 mhz uncore.

Uncore must be 2x memory or no boot. So now try 2000 mem. To run 2000 mem (I do for benching), I must use 10x mem multi 10x200 bclk, and thus 20x uncore at minimum. I need 1.64 qpi/vtt and 1.96 DRAMv to do so, way outside intel recommended limits, and that is for benching stability (may need more v for 24/7) albeit arena chess, which is like stress testing for 5 minutes....and I use 22x uncore or 4400 mhz, though could use 20x which would be little easier. So if you are willing to run high volts for 24/7, then you can run 2000. Even if unlike me, you buy ram rated for 2000 which will run at 1.65 v DRAM, it still requires high VTT/qpi and probably still high DRAM to drive uncore at 4000 min. But you must assume the risk of high volts, and the benefit is next to nothing in real world apps between running 1600 and 2000, as that has been shown time and again. Even benching you are talking 1% improvement in only some benches, not all, if that...1% makes diff in benching, but not so much on 24/7.

RGE would running qpi at 4220 and 1.335 volts be ok? Will the speed cause any damage? or just the volts. My ram is set at 1.66 aswell.

Thanks for any help.

[rge]

Some reviewer did testing between running qpi at 36, 44, 48 on gigabyte (listed 18, 22, 24 ie different units on asus but same), cant remember where, maybe someone else will. There is no benefit to running higher qpi than lowest 36x gb (18x on asus) so all you are doing is making it more difficult to run stable. I benched myself using a bclk I could run all 3 at, got same scores, not faster at all. There is a difference in slow mode, but i think that is just so people can test bclk issues.

I dont think it is an issue of hurting anything, just that it has no benefit, it detracts from stability, so no point in doing it. Guessing you are 21x191?? I would just put on 36x gb/18x asus qpi, for 18x191 or whatever your bclk is.

If you meant uncore, that is different, as higher uncore does help performance, but again detracts from stability.

[bfghornet]

Some reviewer did testing between running qpi at 36, 44, 48 on gigabyte (listed 18, 22, 24 ie different units on asus but same), cant remember where, maybe someone else will. There is no benefit to running higher qpi than lowest 36x gb (18x on asus) so all you are doing is making it more difficult to run stable. I benched myself using a bclk I could run all 3 at, got same scores, not faster at all. There is a difference in slow mode, but i think that is just so people can test bclk issues.

I dont think it is an issue of hurting anything, just that it has no benefit, it detracts from stability, so no point in doing it. Guessing you are 21x191?? I would just put on 36x gb/18x asus qpi, for 18x191 or whatever your bclk is.

If you meant uncore, that is different, as higher uncore does help performance, but again detracts from stability.

Thankyou RGE for the great reply. Im sorry man, I should have given you better information about my setup.
Im running 211X19 with I920 and GigaUD5. My qpi and uncore settings are set on auto and ram is running at 8x.

I got mixed up... CPUZ says my QPI speed is 3798(in bios is says 7.59GHz), and my Uncore in bios is 4220.
Do you think it looks like I need to change anything?

Thanks again

[liftedcj7on44s]

It goes on, and on, my friend. I think that's part of the fun, though. Unless, of course, you're a certain individual with an MSI board who's about to throw his i7 rig out the window.

I will give you 100 dollars however you like it if you think you can get my i7 stable at 4ghz. in all seriousness.
You are spreading my words around this forum like i am an or something. i got frustrated, when you see all these people getting 4ghz stable and cant do it for yourself then it gets frustrating.
100 dollars however you like it.

[rge]

Thankyou RGE for the great reply. Im sorry man, I should have given you better information about my setup.
Im running 211X19 with I920 and GigaUD5. My qpi and uncore settings are set on auto and ram is running at 8x.

I got mixed up... CPUZ says my QPI speed is 3798(in bios is says 7.59GHz), and my Uncore in bios is 4220.
Do you think it looks like I need to change anything?

Thanks again

Your qpi speed is 7596 GT/sec and 3798 in cpuz...same speed different units...just like GB bios just uses 36 qpi multi versus 18 on asus. You are running same lowest qpi multi as everyone else, you cant go any lower anyways.

Your uncore, you are running much higher (20x) than most, surprised it is stable with that low dram/vtt. I run mine at 17x or 18x. I use 20x and higher for short benching as I need much higher volts than you are using for that. Not going to hurt anything other than making stability more difficult, so if you run into stability problems, I would take uncore off auto and put at 18x. Gaming, apps and stuff you wont see much diff between 18 and 20, but a few benching programs it makes a small difference...at least on my testing, and again at least one review site did that ? anandtech..cant remember.

[SteveRo]

Good morning, I think I might have 4.3 HT on prime stable.
I needed to up the vcore to 2 clicks above 1.55v.
I tried to justify this was ok because of vdrop.
Looks like cpuz idle is at 1.536v - so maybe ok (certainly on the edge).
Anyway - 4.3ghz 9 hrs prime stable under water at just under 80C max T -

[SteveRo]

So I picked up some new memory from the egg.
Arrived yesterday - gskill pi black 3x1gb 2000 - http://www.newegg.com/Product/Produc...82E16820231240
Installation was a breeze and memory check out went fine.
Not much bling, don't care much for the heat spreader tops but seems functional!
Turned on xmp and yikes!- gskill has the xmp set to qpi of 1.55v!
Well - I let it run over night and all seemed fine - I'm concerned about letting it run at that for 24/7 - the intel specified max is 1.35v.
Interesting that gskill is doing this.
Screenies to follow -

[Lukee]

Good morning, I think I might have 4.3 HT on prime stable. I needed to up vcore to 2 clicks above 1.55. I tried to justify this as ok because of vdrop. Looks like cpuz idle is 1.536 so maybe ok (certainly on the edge). Anyway - 4.3ghz 9 hrs prime stable under water at just under 80C max T -

Nice OC Steve!!! Too much voltage for me though. This forum is really bad because with all of the great OCs people are doing, it makes you want to take yours even further.

I remember saying, "I'd be happy with a 4.0 C2D", then I got the E8600 and hit 4.2 with no issues, but instead of being happy, I kept pushing until I hit 4.5.

This time I am trying to find a nice balance between voltage and performance. I'm at 4.008 GHz at 1.288V and anything higher requires a lot more voltage, so I'm thinking that I may stop there and call it a day.

1.53V, wow..

[Lukee]

So I picked up some new memory from the egg.
Arrived yesterday - gskill pi black 3x1gb 2000 - http://www.newegg.com/Product/Produc...82E16820231240
Installation was a breeze and memory check out went fine.
Turned on xmp and yikes!- gskill has the xmp set to qpi of 1.55v!
Well I let it run over night and all seems fine - I am concerned about letting run at the v for 24/7 though.
Interesting that gskill is doing this.
Screenies to follow -

Steve, can you run an Everest Test with that memory? I'm curious to see how it does.

[SteveRo]

Will do - i need to buy the thing first.

[Lukee]

Will do - i need to buy the thing first.

No need, just use the trial and post your results. It won't show everything, but it shows enough.

[eva2000]

So I picked up some new memory from the egg.
Arrived yesterday - gskill pi black 3x1gb 2000 - http://www.newegg.com/Product/Produc...82E16820231240
Installation was a breeze and memory check out went fine.
Turned on xmp and yikes!- gskill has the xmp set to qpi of 1.55v!
Well - I let it run over night and all seemed fine - I'm concerned about letting it run at that for 24/7 - the intel specified max is 1.35v.
Interesting that gskill is doing this.

known procedure by nearly all manufacturers as the facts rated modules >DDR3-1866 will have CPU VTT in the 1.5+ region - same for Corsair DDR3-1866 validated at CPU VTT 1.5v IIRC

How much CPU VTT you need depends on cpu IMC too

i7 920 @4.6G 32M PI

BCLK 220
Multi: x21
QPI/VTT : 1.4v
Vdim : 1.8v
Vcore : 1.55v
cpu vtt : 350%
cpu pll 1.8v
qpi pll 1.4 v
ioh core 1.25v
ich 1/0 1.45v
ich core 1.2v

http://cfs13.tistory.com/original/20/tistory/2009/01/29/12/28/49812253b2f5c[/IMG]
http://cfs13.tistory.com/original/36/tistory/2009/01/29/12/28/49812253ceab4[/IMG]
http://cfs13.tistory.com/original/13/tistory/2009/01/29/12/28/49812253ebbae[/IMG]

i7 920 [3838B079]

nice just need to work on 32m times as you should be well under 8min for 32m at 4.6+ ghz

[SteveRo]

eva - but is 1.5 qpi safe?

[eva2000]

benching 1.5-1.6v cpu vtt seems fine but 24/7 no real idea too early to tell heh

[SteveRo]

suicide everest memory bench below - I have no clue if this is right on no -