Overclocking i7 930

[Atomic_Sheep]

Hi guys, I'm finally up to the stage of over clocking my system. I've already got the CPU up to 4Ghz last night for the first time for around 10 minutes or so. I'm not really sure whether I'm doing the over clocking right or not though.

After over clocking last night, I decided to downclock to stock speeds that I was sitting on before overclocking i.e. my memory was at 1066Mhz, 2.8Ghz (before turbo of course) for CPU. After playing around with standard Gigabyte overclocking GT6 program or whatever it's called... I suddenly noticed that when you overclock to the Gigabyte presets, sure it changes the base clock but I think it leaves all the other settings auto and for the first time today I noticed that my DRAM voltage was at 1.7v+! when I overclocked to the highest the GT6 program will go i.e. 3.36Ghz. After seeing those DRAM voltages I immediately crapped my pants because I had overclocked to 4Ghz last night although in BIOS as the Gigabyte prog as I said, doesn't go that high but I didn't change any of the other settings from auto so I'd hate to imagine the sort of voltages I was getting on the memory!!!

Are you supposed to switch all the auto options to off and set everything manually because I don't see how leaving stuff on manual will give us voltages that don't go to ridiculous levels!

P.S. My specs are:
i7 930
X58A-UD3R
G-Skill F3-12800CL9T (9-9-9-24 1.5V)

P.P.S. is there any chance that I damaged anything last night? How would I tell if I did?

UPDATE: When I reduced the memory multiplier to 8 or 6 or whatever the lowest setting that my mobo has, the uncore multiplier to the lowest and all the QPI to the lowest... the computer wouldn't turn on... kept turning on for a bit and switching off... had to reset the CMOS. (I realised that overclocking through the standard Gigabyte overclocking thingy is... um... not the best way to go as you can see above... kinda freaked out by the whole thing actually) which is why I'm trying to do it all through BIOS exclusively now.

 

[Draizien]

Are you supposed to switch all the auto options to off and set everything manually because I don't see how leaving stuff on manual will give us voltages that don't go to ridiculous levels!

It is not necessary to set every voltage manually, the main ones you should be concerned with are Vcore, QPI/DRAM (VTT), DIMM, IOH and perhaps CPU PLL. CPU PLL is somewhat touchy so if you change it make sure that is the only variable you are changing, although only changing one thing at a time is a good practice anytime, particularly for your first foray into OCing. Furthermore, I would not worry to much about the 1.7 VDIMM, I know Intel says not to go above 1.65V but I have never heard any reports of frying the memory going above it, within reason of course.

P.P.S. is there any chance that I damaged anything last night? How would I tell if I did?

Highly unlikely, if you want to be sure give the board a good look over, pull the CPU out and check the pins, and give the DIMM's a once over.

UPDATE: When I reduced the memory multiplier to 8 or 6 or whatever the lowest setting that my mobo has, the uncore multiplier to the lowest and all the QPI to the lowest... the computer wouldn't turn on... kept turning on for a bit and switching off...

Since you did not supply actual frequencies here I'm going off intuition. My suspicion is you lowered your QPI to a frequency that was less than twice the DRAM frequency. If this is the case, raise your QPI frequency so it as at least twice the DRAM frequency.

For a great beginners guide to i7 overclocking, look here.

 

[Atomic_Sheep]

Thanks, for the info... very helpful and a great relief also.

Since you did not supply actual frequencies here I'm going off intuition. My suspicion is you lowered your QPI to a frequency that was less than twice the DRAM frequency. If this is the case, raise your QPI frequency so it as at least twice the DRAM frequency.

Good point... totally forgot about this... will make sure I'm keeping that ratio in the future... I was wondering why a video on youtube didn't lower the QPI to the very lowest setting that was available in BIOS. The guy didn't mention the reason for it so I decided to be stubborn and try something else just to show him what's what! But there you go... learned something new in the process of being humbled.

For a great beginners guide to i7 overclocking, look here.

Yep, I've been reading that for my reference. When I initially started reading about overclocking, I basically understood the process thanks to that very article. But there were some months between reading about it and actually doing it a couple of days ago and I just remembered that it was a rather simple back and forth process but I forgot the details and decided to jump in the deep end before refreshing my memory on a few things.

 

[Atomic_Sheep]

After rereading the overclocking guide... I'm confused about one thing... it tells you that the first step is to lower the all the multipliers in order to be able to establish the highest bclock that is possible and then bump up each of the other multipliers individually etc etc until you're done. The thing I'm confused about is that in the guide... voltages appear to be set after the maximum bclock is established i.e. they are talked about last in the guide and it seems a bit confusing to me because if it's all set to auto, won't the voltages increase with the increases in the multipliers/bclock up to the levels that I almost crapped my pants at? Or do we manipulate the voltages at the same time that we are playing around with the multipliers and the bclock?

Also... when it says to stop the bclock overclocking at:

You reach your desired bclock and successfully pass your stability test.
You reach your maximum safe CPU VTT voltage.
Raising the CPU VTT does not allow for additional stability.

I'm a bit confused about what bclock is exactly and why it is that we are increasing it seperately from everything else? Like to me, overclocking a cpu means setting the multipliers that you will be using and then increasing the bclock until just below unstable or until you're happy with your overclock. I don't understand why bumping up the base clock has to be done individually? Won't Prime95 pop up with the same errors if we simply overclock at the correct cpu multipliers? I hope that last bit is clear... I'm just a bit too confused at the moment in this respect to say it any clearer. So I guess to sum up, why don't we find the maximum stable bclock with the CPU multiplier at it's end state? I sort of just came up with a possible answer to that... being that not everyone will be able to thermally push their CPU's to levels where bclock is max? Is that the only reason?

Last but not least. I would like a bit more info about this whole voltages issue. The way I see it, if you increase the bclock... you're essentially increasing current? Thereby you're increasing the temps of the CPU? By increasing voltages, you are reducing the current? I'm no electronics buff and I'm sure what I said is wrong but basically, from what I've seen, if you increase voltage, you basically reduce the tempereature of the component?

Also how is increasing voltage for any given level of bclock/frequency going to make the CPU or the memory more stable? Why do the errors occur in the first place when we overclock?

 

[Draizien]

After rereading the overclocking guide... I'm confused about one thing... it tells you that the first step is to lower the all the multipliers in order to be able to establish the highest bclock that is possible and then bump up each of the other multipliers individually etc etc until you're done. The thing I'm confused about is that in the guide... voltages appear to be set after the maximum bclock is established i.e. they are talked about last in the guide and it seems a bit confusing to me because if it's all set to auto, won't the voltages increase with the increases in the multipliers/bclock up to the levels that I almost crapped my pants at? Or do we manipulate the voltages at the same time that we are playing around with the multipliers and the bclock?

You are changing voltages when determining your maximum Bclock. In particular, the QPI (VTT) voltage which aids in stabilizing the higher QPI frequency which is a result of increasing the Bclock. It is possible to leave Vcore at auto when doing this. However, in my experience this tends to give an inaccurate representation of temperatures when determining max Bclock. So long as you adjust the CPU multiplier to have the processor running near stock speeds or lower you could set the Vcore to around ~1.16V and should not experience stability issues.

I'm a bit confused about what bclock is exactly and why it is that we are increasing it seperately from everything else? Like to me, overclocking a cpu means setting the multipliers that you will be using and then increasing the bclock until just below unstable or until you're happy with your overclock. I don't understand why bumping up the base clock has to be done individually? Won't Prime95 pop up with the same errors if we simply overclock at the correct cpu multipliers? I hope that last bit is clear... I'm just a bit too confused at the moment in this respect to say it any clearer. So I guess to sum up, why don't we find the maximum stable bclock with the CPU multiplier at it's end state? I sort of just came up with a possible answer to that... being that not everyone will be able to thermally push their CPU's to levels where bclock is max? Is that the only reason?

For the Intel core CPUs the base clock takes the place of the FSB clock and is what sets the frequencies for the CPU, QPI, DRAM, & Uncore. Hence increasing the Bclock changes the frequencies of more than just the CPU. The Bclock is determined separately for a few reasons. Most importantly, if we are dealing with CPU's with locked multipliers our max base clock effectively sets the limit on the potential overclock. Furthermore, it can allow for fine tuning of the other frequencies which cannot be done with just multipliers so it is advantageous to know just how far it can go. The method you describe, setting max multiplier and bumping up Bclock, is a perfectly valid way to overclock.

Last but not least. I would like a bit more info about this whole voltages issue. The way I see it, if you increase the bclock... you're essentially increasing current? Thereby you're increasing the temps of the CPU? By increasing voltages, you are reducing the current? I'm no electronics buff and I'm sure what I said is wrong but basically, from what I've seen, if you increase voltage, you basically reduce the tempereature of the component?

Also how is increasing voltage for any given level of bclock/frequency going to make the CPU or the memory more stable? Why do the errors occur in the first place when we overclock?

To my knowledge, the current drawn does not change. To answer your question it is necessary to define what voltage and amperage actually are. Amperage is a measure of coulombs per second (C/s) and thus represents the rate at which charge is delivered. Voltage on the other hand is a measure of joules per coulomb (J/C) which is how much electrical work is delivered for every coulomb of charge. By increasing the voltage we are increasing the work that is delivered to the CPU every second. It follows then that, by increasing the frequency of the CPU it takes more electrical energy delivered per second to power the CPU and thus the reason why voltages need to be increased. Now moving to your assertion about temperature and voltage. Temperatures increase when you increase the voltage. The power, or wattage, which is a measure of joules per second (J/s) a CPU receives (or any electrical component for that matter) is determined by P=VI, where P is power V is voltage and I is current. Thus as we increase the voltage we increase the power. Most of the power a CPU receives is lost as heat, ergo as we increase the voltage and hence the power the more heat that is given off by the processor per second.

Hopefully this answers your questions and does not confuse you further. If it not I can try to explain it a little clearer.

 

[Atomic_Sheep]

I think I'm starting to get a clearer picture of all this overclocking. So in other words, what overclocking is all about is raising the bclock and the multipliers to the maximum level that you can/that the components are designed for in the case of memory whilst keeping voltage as low as possible in order to reduce the power consumption as well as reduce the temperatures to the lowest possible levels? By lowering voltages, you are essentially creating a situation whereby in some instances, when the system is maxed, there isn't enough power to enable to chips to do what they are designed to do i.e. information is lost which then results in errors which means you need to bump up the voltage in order to provide sufficient power in those situations?

I'm still not 100% sure about how all this works... mainly what frequency is in a physical sense and how it's used in order to do calculations. A processor is basically a bunch of capacitors?... but how is this frequency related to all this? (that's more of what I was asking when asking about what bclock was). My theory as to what frequency is in this sense is it simply determines how many times a second the capacitor is being able to charge and discharge? So if there isn't enough power to charge the capacitor in certain high intensity situations at the desired frequency, we'll get an error in which case we need to bump up the voltage, thereby increasing the power and thereby enabling the capacitors to be charged and dischared at the required clock speed without any mistakes/misses of charges/discharges?


Anyway... the next question I have is related to my BIOS.

(not my current settings nor is it my bios... I'm still stock until I'm more confident with what I'm doing)

I don't quite understand why there are 2 columns for the options for the voltages. For me... the ones on the left show the voltages... and on the right, you can make the selections... with mine currently set to Auto, I'm guessing that just because the left column may say DRAM voltage of 1.5 which mine does at the moment, it won't be limited to that figure if the system so choses i.e. the left column isn't the upper limit?

Also, same goes for the memory timings. Mine are set to auto also... the left columns are showing 7 7 7 something (20 maybe... but doesnt matter)... and the right one is showing 9 9 9 24 2 I believe with the right column greyed out as it's on auto as I said. So what does the left column showing?

EDIT: Also... if I want to see whether I can run my memory @ spec... i.e. 9 9 9 24 2T @1600mhz & 1.5v, I would need to increase the memory multiplier at the standard bclock of 133 and also crank up the uncore to at least 3200mh? So I would set my memory voltage to 1.5v in the right column as opposed to auto? Is there an uncore voltage setting? After I've got my memory up to spec, I would then something like memtest or something to see whether it's stable at those volts? And obviously, the bclock will remain the same 133 as we are not checking for its stability and the QPI will remain stock also? This way I'll be able to tell if my memory can run at spec?

Lastly:

http://www.maximumpc.com/article/fe...ing_guide_we_push_nehalem_its_limits?page=0,1

On this site... on page 2 it says:

"The QPI is important to watch because ramping it up too far can kill your overclock."

What does it mean that it can kill your overclock? How is increasing QPI past the stock 4.8 going to make things worse? Can it make the system unstable if it goes too high?

 

[Atomic_Sheep]

Hi guys, I'm currently sitting on x36 QPI, x6 Mem, x14 CPU and x14 Uncore
bclock 200
vcore 1.25v
qpi/vtt 1.155v
dram 1.5 (what my mem should be at) and 999 24 auto timings
IOH is set to auto...

Anyway... doing prime95 I get no errors after around 5 mins or so... had none at slightly higher qpi/vtt (what I'm playing with atm)... I'm currently decreasing the qpi/vtt with each pass of 5 minutes of prime95... why am I going so low in terms of this setting? In the i7 guide, they basically say to increase this value up to around 1.3v straight up and keep increasing if there's instability.

Also... as I decrease my qpi/vtt my temps during prime95 don't seem to be dropping... does that sound right? Steady at around 55c

 

[EarthDog]

vtt will affect CPU temps, yes.

 

[MIAHALLEN]

Hey man, good job so far....I'm glad the guide has been helpful for you.

Your QPI/VTT is VERY low....but that's not a bad thing It just means you have a very good chip. Forget dropping it lower, you're low enough already, and you'll likely have to raise it again anyway once you move on to OCing your memory. If I were in your shoes, I'd probably set it to 1.25V to make the memory step a little easier

Temps are going to follow voltage changes in an exponential fashion....in other words, the chage between 1.3 and 1.35V will make a very noticable change in your temps, but the change from 1.1 to 1.15V will not

Doin great man, keep us posted

 

[Atomic_Sheep]

vtt will affect CPU temps, yes.

Temps are going to follow voltage changes in an exponential fashion....in other words, the chage between 1.3 and 1.35V will make a very noticable change in your temps, but the change from 1.1 to 1.15V will not

Yer, I noticed 1c drop from a high of 57 when I dropped down to my present 1.115v. So I was was kinda expecing at least a mild drop... just didn't experience it after many drops so got kinda curious. I wasn't planning on going any lower as 1.1 is stock? But overall I'm happy with my 200hz bclock and so not going any higher. Pretty sure I won't be having a 4.4Ghz overclock. When I overclocked everything the stupid way (leaving all the voltages on auto), I was getting temps in the mid 80s and that was at I think 3.8Ghz (total with boost) in prime95. I found out today that I haven't finished screwing the heatsink the way it's meant to be screwed in so that may be the reason for the high temps.

I have started to try to get my memory up to 1600 speed but I immediately encountered BSOD and failure to boot and loops during startup and system checks so going to start increasing the qpi/vtt back up until I'm stable (I did crank up uncore from x14 to x17 with the memory multiplier going to x8 rom x6 so it's definitely low qpi/vtt. Don't think I'll be overclocking the memory at the moment at least... so when I get up to spec, I'll get straight to increasing the multipliers.

EDIT: Well... looks like I managed to boot @ 1.195v for qpi/vtt... will do a memtest and prime95 test tomorrow hopefully so see how it goes

EDIT2: Had to increase it to 1.215v when I started running Prime95 CPU only.

EDIT3: Btw... do you have to increase the CPU multiplier back to 21 or can you leave it lower? I don't see any reason why you shouldn't be able to leave bclcok @200 enabling you to have your memory run @ spec rather than underclocked or overclocked slightly and have the CPU run at an unorthodox CPU multiplier which gets you to 4Ghz with turbo i.e. this would mean a multiplier of x19? Anyone see any issue with this? Just seems so nice working with a bclock of 200!

 

[MIAHALLEN]

Your final CPU multi and bclock values can be watever passes and you like There is no rule that you have to use any certain values.

 

[Atomic_Sheep]

Hi guys, well I've been increasing the CPU multi and I've hit a bit of a wall at 20. I'm increasing vcore steadily... up to 1.31250v but still getting BSOD during Prime95 (takes longer for it to BSOD with each increase in vcore). Is core voltage the only thing I should be worried about? I know it says that once you're up to the CPU multi stage, you shouldn't have to worry about QPI/VTT but

http://www.maximumpc.com/article/fe...ing_guide_we_push_nehalem_its_limits?page=0,1

suggests that some say the QPI/VTT should be at least equal to the vcore to ensure stability. Any comments? (I tried bumping it up by 1 notch @ vcore 1.2750... didn't appear to make any difference but on that site it says you need a higher QPI/VTT for extreme overclocking)... getting 4Ghz is proving more difficult than I thought after the initial success... the temps are getting too high for my liking also @ these voltages... @ ambient of around 17c I'm hitting 70c during Prime95 which is basically a 5 degree increase over 3.8Ghz, so talk about exponential! I'm starting get the feeling 3.8 is my overclock.



Also, the standard QPI speed is 4.8Gt/s... how can I tell what mine is set to? All I can see is that my QPI is at 3.6Ghz although in BIOS its got 7.something Gigathreads.

 

[Draizien]

In my experience it is not always about raising only Vcore. What kind of BSOD are you getting? Knowing this can help to diagnose your instability issues. How are you going about increasing your frequency? Holding Bclock constant and changing multiplier or vice-versa?

 

[Atomic_Sheep]

Well @ 1.25v vcore and 1.215v qpi/vvt, I was booting half the time, other times was getting bsod during boot, other times getting bsod just as I would start prime95. As I kept increasing vcore, I stopped getting any boot issues but continued to get BSOD during Prime95... the higher the vcore that I would increment to, the longer Prime95 would run for until I can't remember what the vcore was when I've started to BSOD pretty much after 3-5 mins of running Prime95 (@1.325v I got a 9 minute run of Prime95 @74c max temp and got a BSOD so looks like it's still increasing stability but at this rate I'll need ridiculous voltages and temps to have a 4Ghz overclock). And yes, keeping everything else constant... i.e. bclock and all the other voltages and multis... just varying the cpu multi... well... I'm trying to get 4Ghz happening @ x20 but @ 1.325v it's still BSOD'ing. Just seems to be a wall... after 0.1v increase, I see no stability @ CPU multi x20. 3.8 happened really easily, so I thought that 4, shouldn't take much more.

 

[fix24311]

Hey guys. I'm new user of Core i7 and never overclocked anything before. I don't want to push my CPU to the limit, just ramping it up a bit is enough for me. Recently I got 3.6Ghz speed with idle temp of 48c and full load temp of 60c to 70c. I disabled C1E, SpeedStep, and TM functions, my BCLK is 173. All other setting are set to AUTO. The stock v-core of 1.19v made my system froze, so I just up it a notch to 1.2v. Ran LinX for 20 runs and all success.

My question is, do you think I can up the speed more with the same temp? I'm using CoolerMaster Hyper N520 as CPU cooler. My specs are:
Core i7 930
Corsair CMP6GX3M3A1600C8 2Gb DDRIII RAM times 3 (so a total of 6Gb)
Rampage II Extreme
GeForce GTX275
Corsair HX750W psu

I'm using HAF932 casing with 8 chassis fan. 6 of them are 90cfm airflow, 3 are exhaust, remaining are intake (4 on the side intake, 1 front intake, 2 on top exhaust, 1 on back exhaust)

*EDIT: If you guys can suggest the best config I could go with my spec is also fine by me ) I'm very much thankful for your help

 

[Atomic_Sheep]

Don't know how good or bad that coolermaster N520 is, but those idle temps seem quite high for that overclock to be honest. The full load temps are pretty decent... if they are mid 60 @ hot ambient conditions e.g. 25-30c or something. Considering you're in Malaysia and all. But other than that... with my cooler @ 3.8Ghz I'm currently sitting on idle temps in the mid 30s. Prime95 I'm getting around 63-64c with ambients around 22-24c I would say.

As for upping the system at the same temp... well you can certainly try and see what happens. I've had my CPU up to 85 for very short while... crank it up and stop the CPU stress tests if the temps reach levels you're not comfortable with.

 

[Atomic_Sheep]

Hmm... looks like I'm not stable at 3.8Ghz. Dowgraded my comp to x19 @1.25v vcore and 1.215v qpi/vtt and I got a BSOD during FSX in a similar fashion to my other BSOD's that I'm getting whilst running Prime95. I tried FSX @ 4Ghz and it BSOD'd after a while just like it just did. Time to see how long Prime95 lasts at this overclock.

 

[fix24311]

Thanks for your reply. So I need to run Prime 95 for until I got BSOD? Do you think I need to tweak other options and not just set it to auto?
BTW, I don't think N520 are good anyway,hehe... It's very cheap, I bought it for my old system which is Pentium D because the stock cooler fail on me after 4 years. I've just recently switch to Core i7, about a week ago.

p/s: I'm using Tuniq TX-2 at the moment for thermal paste

 

[SamSaveMax]

Hmm... looks like I'm not stable at 3.8Ghz. Dowgraded my comp to x19 @1.25v vcore and 1.215v qpi/vtt and I got a BSOD during FSX in a similar fashion to my other BSOD's that I'm getting whilst running Prime95. I tried FSX @ 4Ghz and it BSOD'd after a while just like it just did. Time to see how long Prime95 lasts at this overclock.

Try this.

CPU Multiplier: 19*200bclk OR 20*190bclk
vcore 1.34v
qpi/vtt 1.31v
dram 1.5 and 999 24 auto timings
leave the rest on auto

If that stable...then try to reduce vcore bit by bit.
When you get to know more...then start to play with other settings.
What CPU cooler do you have?..... i7-930 runs pretty hot above 4GHz.
What is your goal for this overclock?

 

[Atomic_Sheep]

Thanks for your reply. So I need to run Prime 95 for until I got BSOD? Do you think I need to tweak other options and not just set it to auto?
BTW, I don't think N520 are good anyway,hehe... It's very cheap, I bought it for my old system which is Pentium D because the stock cooler fail on me after 4 years. I've just recently switch to Core i7, about a week ago.

p/s: I'm using Tuniq TX-2 at the moment for thermal paste

I'm getting BSOD for my own reasons... I think you're meant to get plain errors. Have a read through the thread and see if you can gain info that will help with your overclock.

Also have a read through this:

http://www.overclockers.com/3-step-guide-overclock-core-i3-i5-i7/

I personally wouldn't set it to auto. When I set stuff to auto, I was getting memory voltages of over 1.7v... so that's not something I want to be running at for sure!

If you have any other questions, I think you should start your own thread. It's a bit hard to manage having 2 people asking quite specific things in the same thread.