• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

i7 920 first attempts

C

CursedMind

n00b
Joined
Jan 6, 2009
Messages
18
So, hi there guys. This board and the vids were awesome, and by watching and reading along I actually found myself re-doing the orders for components of my all purpose gaming&work rig.

I wasn't into PCs much, as I am an Mac user ever since my first encounter with computers, and by that I've not been looking on components much, neither cooling. Just to explain why I'm so careful in doing the OC.

To get to the point:
I've built and set up a stock clocked machine with

CPU: Intel Core i7 920
CPU Cooler: Thermalright Ultra 120 Extreme 1336
MB: Asus Rampage II Extreme
all inside a Cooler Master HAF 932 Case

GFX: Asus stock OCed GTX280 (wich I mention because its actually the hottest thing I have in that Case when that machine is under load... idle at 47°C)


I did get drivers, SpeedFan and Prime95 and what I'm looking at now irritates me a little:
After 1 hour of 100% load on all cores, HT enabled, no bios settings done whatsoever, CPU temperature is at 39/40°C with 20°C insider the case. ( room temperature at 21°C )

While this is probably something I should be happy about, I'm not quiet sure that I tested the right way. That OC/heat charts done by Kyle show very diffrent numbers with the same CPU cooler being used. Can anyone give me a hint what you guys do to test your builds and get accurate readouts for Core Temps ?
And how careful I should be with this machine when OCing to 3.8 with HT enabled ?

EDIT: When I get back to idle temp goes back down to 20°C in less then two minutes.
 
SO, after having some bluescreens first, Prime95 warnings then, I do have a stable system at 3.2Ghz 160 BLCK with 1.3 CPU voltage and 1.35 DRAM/QPI voltage.
I don't get the voltage thing yet, but as the title says it is my first attempt, and I guess that 1333 Ram of mine just doesn't like higher BLCK then that. temperature is fine at 51°C peak...:rolleyes::confused:;):D
 
Well , nothing wrong till now , temp is excellent.Try 33 multiplies , like 166 , it helps with the dividers since everything is a multiple of 33.
 
Are you setting your memory divider correctly? What about your uncore? What TIM?
 
166 works fine. I did lower the QPI voltage though, it's 1.28something now.

to the memory divider: Can I set that lower to get stable ram with higher cpu multiplier ? Most of the stuff is set to "auto" because I honestly don't know what is reasonable to set there, or worse, because I don't know what it does. Uncore gives me Values instead of a multiplier.

EDIT: All those Ram settings give me a hard time lookin for explanations but I think I may have too low latency settings ?
 
One thing to try is set the memory speeds low, and then see how far you can push the CPU clock. Then jack up the memory clock and not the CPU clock, though that might not be possible except for unlocked CPUs.
 
*sigh
I think I better wait and read a little more before I try now as I don't think I got all the mechanics involved yet.

3.32 Ghz with 20x mutliplier 166 Bus speed, DDR3 at 996 Clock, thats what I know I've got...

Nehalem calculator tells me QPI clock is 2988 and UnCore at 1992, but I don't know where I would change the UnCore Multi (wich at these settings must be 12 but I'd like to be 16) or the Memory Multi (wich would be 6 it says, but I'd like it to be 8)

(calculator download) http://icrontic.com/downloads/nehalem-overclocking-calculator

Thanks for the feedback though, this OCing stuff is pretty nice to have a look at. I'd never have guessed theres so much behind it, and if not for a faster machine, at least I can learn something from it :D
 
DDR3 @ 966Mhz is 1932Mhz. So you have your 1333 running at 1932.

Your bclock * mem multi will give you your mem speed.

example 200bclock with a divider of 2:8 = 1600Mhz.

Also your uncore must be exactly double that of your mem multi. IE 8 * 2 = 16x.
 
It was a LOT easier then I thought when I just set those "auto" values manually.
Looks like all those problems were caused by the Rampage autovalues.

System ran 3.3Ghz with 20 CPU multiplier at 166 Bus speed, 8 MemMulti+16QPI Multi to give me a bit slower but 1328 DDR3 8-9-9-24 clocks wich should be within the specs for my cheap memory.

It ran Prime95 for 20 Hours without errors while not getting any more then 53°C At CPU, and no, the window was closed wich means room temperature at about 24°C.

And now I'm happy having not only built my very first PC system, but did a safe, stable OC too. :) Thanks guys for the input.

Lesson learned there:
DON'T set AUTO values. IF you don't know what it does, try to figure it out, but don't set it to AUTO. :):eek::D
 
sorry for resurrection of this thread.

some weeks passed, and I read a lot of people hitting 3.8 on air with stock voltages. I don't even get to 3.6 on stock voltage, one out of 20 tries @3.6 managed to stay stable for 10 minutes of Prime95 before BSODing on me, and I don't consider that stable...

3.3 at 20 multliplier & 166 bus speed, ram somewhere around the 1333 ( as being not OCed) runs fine, but nothing else.

I tried much, and I did boots and some minutes of stresstesting with various popular settings like 19x190 and even some reckless stuff around 4.2 on air just to see what the problem was. Temperature is not an issue there either, none of those clocks went even near the 54°C I get at current *safe* 3.3 OC because I'm not able to stress them long enough before system fails.

So what I'd like to hear from you is some advice on that "easy to get" 3.8 Ghz.
Highes priority though is that a mild stress like Prime95 runs 24 hours without any issues.

Whats the better method: 190 Clock x 20multi Ram @ 1140 ( Is stable atm @ 1333 though)
or
200 Clock x 19multi Ram @ 1200Mhz ?
I did both of those clocks at various voltages somewhere around 1,325 vCore, and while I have the feeling that my CPU is stable at nearly any voltage(Prime95 with minimal ram usage runs at least an hour as far as I have tested that), I get insta-bluescreen as soon as there is some RAM usage wich I cannot understand, because that is even downclocked a little. Any help there ? I think I'm missing something that is so obvious it's not explained anywhere.

The rampage bios does not allow to set uncore multis smaller then twice the mem.clock, And I did always set it that way, and 190x20 would be soo sweet to get... I do a lot of rendering on this machine. :/

EDIT: just a link with some BIOS settings where I can see what voltages have been used for a stable system would be fine too. I'm willing to spend some hours, but I just spent too many on something people consider "easy" :D
 
Use CPU-Z to verify your actual Uncore and QPI speeds. By far the easiest way to hit 3.8ghz is to use 21x181 by enabling turbo. If your RAM isn't able to do 1448mhz (181 bclk @ 8x mem multi) you will have to set the memory multi to 6x for 1086mhz and set the Uncore multi to 12x which is the minimum stable setting for that memory speed. You can increase the Uncore multi by 1 or 2 beyond the 2x memory ratio for a performance boost but may need more DRAM voltage and/or QPI/VTT to be stable.

Also, your board may have an issue with Turbo not engaging past certain clock speeds due to the 'TDP limit exceeded' issue. I'm not sure if there is a BIOS update or what for that yet but I think you might not even hit that issue until 4ghz? I'm not sure but definitely give turbo mode a try since it's like a 'free' higher multi that lets you keep more stress off the motherboard/RAM/IMC if you need to.
 
As far as I have seen, I cannot set consumption levels for Turbo, and the Bios itself states that Turbo is going to be disabled for clockspeeds beyond 3.6Ghz.

But you say nothing is wrong with my numbers ?
 
CursedMind said:
SO, after having some bluescreens first, Prime95 warnings then, I do have a stable system at 3.2Ghz 160 BLCK with 1.3 CPU voltage and 1.35 DRAM/QPI voltage.
I don't get the voltage thing yet, but as the title says it is my first attempt, and I guess that 1333 Ram of mine just doesn't like higher BLCK then that. temperature is fine at 51°C peak...:rolleyes::confused:;):D
Click to expand...

really? i got teh same speed ram you have and only set the BLCK to 165 and did a PRIME test for 7+ hours last night with temps going as high as 63 on all cores with no BSOD and no tweeking with ratio's memory and cpu voltages at all and its running stable at 165 BLCK. i might be brave enough and push it abit higher :)

how do i know what my voltage is set to? i have a asus p6t delux mobo
 
ignore those old statements above, I was running with near random voltages there. But what you say is entirely correct, as long as you stay at 166 blck theres absolutely nothing to worry about, no matter what you set or don't set. The Ram timings are just something I forgot to undo before returning to my old OC.
 
cec1l said:
You can increase the Uncore multi by 1 or 2 beyond the 2x memory ratio for a performance boost but may need more DRAM voltage and/or QPI/VTT to be stable.
Click to expand...

QFT, I just found this out the hard way yesterday ...
 
CursedMind said:
ignore those old statements above, I was running with near random voltages there. But what you say is entirely correct, as long as you stay at 166 blck theres absolutely nothing to worry about, no matter what you set or don't set. The Ram timings are just something I forgot to undo before returning to my old OC.
Click to expand...

how do i know how much voltage my cpu si running at?|
 
I figure you should be able to see some number in the Bios screen, if not, at least Core voltage is shown in CPU-Z
 
Hi, figured this was a better place to give my first impressions of 920 OCing...instead of starting a new thread.

My 920 won't run 4GHz stable without 1.4V or more, which isn't enough for my HSF (that barely fits in the case anyway.) :)

I'm running 3.8GHz (190x20) on air with core temperatures in the 70s.
I just changed my voltage settings to 1.275 core, 1.4 QPI, 2.0 PLL, 1.66 DRAM (1540MHz/8-8-8-24).
Prime95 has been running in the background for a while; core temps are in the mid-70s.

screenshot

FWIW, my 920 is from batch 3839A and I'm running a Vigor Monsoon III HSF.
 
Thanks, that looks promising. And I think I know now why I've run into that much trouble on my own... I'll try a little tonight when I'm done working, and post what I get then.
 
MechGeek said:
Hi, figured this was a better place to give my first impressions of 920 OCing...instead of starting a new thread.

My 920 won't run 4GHz stable without 1.4V or more, which isn't enough for my HSF (that barely fits in the case anyway.) :)

I'm running 3.8GHz (190x20) on air with core temperatures in the 70s.
I just changed my voltage settings to 1.275 core, 1.4 QPI, 2.0 PLL, 1.66 DRAM (1540MHz/8-8-8-24).
Prime95 has been running in the background for a while; core temps are in the mid-70s.

screenshot
Click to expand...

Currently testing with PLL & QPI voltage that looks a bit more like those that you have, but no screen yet, I'm still not entirely stable, but this is -way- better then when I started considering I have time to write this before prime gives me some blue color all over that screen :D

Actually, I even messed around 4 Ghz, but there is just no way my RAM is ever going to make that happen,... temps looking fine with "want to make sure" 1.44 vCore, going to work on that as soon as I'm stable. 1.275 did definetly -not- work for me.

UPDATE:
Working myselfe upward beginning with 1.275, I found that anything below 1.32 does not even work until login screen for me, with anything below 1.42 giving me BSOD after a couple of seconds in Prime :/
I did steps of 0.02 Volts for this, it looks like I'm either doing something wrong here, or my CPU is just one that does not like to be overclocked much. :(

2nd Update:
Looks fine. I may have to go to 1.44 again if I run into any errors (going to run 2xPrime @3GB ram each for 6 hours now), but I don't think so.
What I noticed though, is that Speedfan and Everest each give me different numbers for my core temps. Why could that be ?
ocscreen3800ve5.jpg
 
For 3.8GHz, I also:
- Set CPU differential signaling voltage to 900mV
- Disabled CPU & PCIE spread spectrum.
- Locked the multiplier at 20x (i.e., no SpeedStep or Turbo mode)

Are your memory timings too fast? My RAM rated for 8-8-8-24 @ 1600. If you're using cheaper RAM, try the next lowest DDR multiplier setting (should be around 1290-ish @ 190Bclk?) and 9-9-9-24 timings.

I was actually able to lower my QPI and DRAM voltages a little. BIOS screenshots here in another thread
 
MechGeek said:
For 3.8GHz, I also:
- Set CPU differential signaling voltage to 900mV
- Disabled CPU & PCIE spread spectrum.
- Locked the multiplier at 20x (i.e., no SpeedStep or Turbo mode)

Are your memory timings too fast? My RAM rated for 8-8-8-24 @ 1600. If you're using cheaper RAM, try the next lowest DDR multiplier setting (should be around 1290-ish @ 190Bclk?) and 9-9-9-24 timings.

I was actually able to lower my QPI and DRAM voltages a little. BIOS screenshots here in another thread
Click to expand...

Could anyone explain to me what I'm doing when I disable CPU &PCIE spread spectrum ?
I know where the option is, but I don't switch anything off (or on) when I don't know what I'm doing... :D

Ram is fine btw. although its 1333 :p
Turbo is off, multiplier locked differential signal on auto because I don't know what it does exactly, or what it can do, setting wrong values.

Found this using google:
Spread Sprectrum will alter/normalize the voltage spikes on your motherboard......... this will in turn reduce Electro-Magnetic inteference.
But yeah, nowadays a lot of devices are build to resist EM inteference, plus overclocked PC will fail if Spread Sprectrum are activated. The reason is that overclocked CPU are very sensitive to voltage spikes, so any spikes sent by the Spread Spectrum function will result in instability!
To put it bluntly - turn it off
Click to expand...

This sounds like it could be the reason I have to go with such high voltages, so the lower end spikes still deliver enough for the OCed CPU to work ?

But isn't this some kind of safety mechanism that is supposed to keep normal spikes that went through the PSU from making my machine crash ?
 
It's not a safety mechanism per se, it's more about reducing peak EM interference on a given frequency band by spreading the computer's stray radio emissions across a larger spectrum.

The problem with spread spectrum being enabled is that every single clock cycle is a tiny bit slower or faster than the previous one. My rationale for disabling spread spectrum is:
- One of the "faster" clock cycles might push the OC too far, crashing at an otherwise stable speed
- Clock jitter makes it difficult for all the clocks' users to stay in sync with one another, especially if they have PLLs of their own or have considerable physical distance between the source and receiver. (remember, at 4GHz, electricity in copper only travels 4" per clock cycle) - a clock triggering a bit too early or too late somewhere in the system could flip a bit and cause a crash.
 
Thanks a lot, I don't think I ever learned that much about hardware in such little time :)
Just got a Prime warning, so I'm back to bios. Going to check those settings now, and what room they give me in core voltage.;)

one final question:
The speed at which a digital circuit can switch states - that is, to go from "low" (VSS) to "high" (VDD) or vice versa - is proportional to the voltage differential in that circuit. Reducing the voltage means that circuits switch slower, reducing the maximum frequency at which that circuit can run. This, in turn, reduces the rate at which program instructions that can be issued, which may increase run time for program segments which are sufficiently CPU-bound.
This again highlights why dynamic voltage scaling is generally done in conjunction with dynamic frequency scaling, at least for CPUs. There are complex tradeoffs to consider, which depend on the particular system, the load presented to it, and power management goals. When quick responses are needed, clocks and voltages might be raised together. Otherwise, they may both be kept low to maximize battery life.
Click to expand...

This refers to that "differential signaling voltage" ?
 
Differential signaling amplitude is a tradeoff. If it's too low, the signals being sent may get lost in the background noise. But if it's too high, the inductance/capacitance will slow it down. I lean more towards the high side because it'll theoretically send a stronger signal, with a quicker, cleaner voltage rise (or drop) to whatever thresholds determine 1 or 0 on the other side. I'm speculating out the a$$ here so take that for what it's worth. I don't really know if it makes any difference or not.
 
Thanks, yes I see what it does then. I'll leave it untouched for now, because (Oh wonderous things have happen) my PC is actually booting fine, and currently running prime again, with some vCore values that I like much much more then those 1.42V I had. *ggg*

I really think turning Spread Sprectrum off should be no.1 on the OCing to-do list, this seems to have insane impact on stability.
 
Hey, in the screenshot you posted, I noticed CPUZ says your DRAM frequency and FSB : DRAM ratio are the same as mine. Have you fixed that? You want 2:6 FSB : DRAM ratio for your memory sticks at Bclk=190.
 
I have fixed ram issues with carefully adjusting voltages and latency. Was a bit tricky though, and I don't know why it worked, but it did. But I actually was not able to get much below that 1.4vCore I wanted sadly. It seems pretty stable, 1.32 upwards, but it isn't.
Got several bluescreens that said: "A clock interrupt was not received on a secondary processor within the allocated time interval". going to 1.376 core Volts seemed to fix that though.

Memory is on 2:8 FSB:DRAM 9-9-9-24 commandrate 2T ( is the 2N latency Option in Bios)
but I have a strange thing happen here with everest:
Everest does seem to think that my cpu Multi is going up and down between 6x 12x and 20x oO CPU-Z does not.
 
Good thread, so I'll throw mine in here. I just started with this cpu, so much to learn.



Ran P95 for 1hr 20mins before I stopped it, and temps were eventually up to 70-73c.

Now I do have Load Line "enabled" and wonder if others has messed around having LL enabled, or leaving it disabled and just increasing the vcore to compensate for the droop, like I've usually done.
 
CursedMind said:
Was that Cl 7 DDR3-1600 ? Nice OC.
Still know temp on NB after prime ?
Click to expand...

Thanks.

The memory is the Corsair C8 6Gb kit. Its awesome ram. I also have a OC setting for the memory running at 1740ish at the 8-8-8-21 timings, but with the CPU at 3460mhz. It's still very quick at those settings. I have not messed with the timings yet.

The NB temps after an hour plus of P95 was at 49--50 degrees.
 
tweaking my 3.8 oc... got the memory running 7-7-7-20 at DDR3-1520. It was Prime95 stable for 12 HOURS before BSODing. sooooo close. :( gonna try 7-7-7-24
 
Mech, I take it your running the "full blend" of prime 64bit version? Also, 12 hours is quite a long time at 100% core usage. How long are you wanting it to run until you feel stability is achieved?

I've read good things about your memory, how do you like it so far?
 
Yep, running the full blend and 8 threads. My rule of thumb is if it can run a Prime95 full mix for 24 hours, it's 100% stable.

I like this memory. It's rated 8-8-8-24 at 1600MHz... since I'm underclocking it a little, I'm trying to get latencies down.

To try and speed up the Prime95-crashing process, I dropped DRAM voltage to 1.62V so I can get a quicker idea of which latency is holding me back. So far, here are my results at the lowered voltage:

7-7-7-20: Prime95 hangs within a few minutes
7-7-7-24: Prime95 calculation error within a few minutes
8-7-7-20: Stable for 20 minutes...possible solution?
8-7-7-18: Testing right now
 
I bet you'll be good at the 8-7-7-20 setting. I'd be curious to see the differences in dropping those two to 7 does versus 8 on some memory bench tests like Everest and such.
 
tRAS went down to 16 with no short-term ill effects (8-7-7-16) so tRAS=20 won't be a problem. Now running Prime95 with huge FFTs (4-32M) ...


Hey, CursedMind, did you miss my previous post about your FSB : DRAM ratio? At Bclk=190 and FD=2:8, your DRAM is running at 1520MHz. Running 1333-rated memory at 1520, regardless of timings, may be what's causing you grief. Try running at 2:7 (if you can?) or 2.6 (DDR3-1140MHz) and let us know how that goes.

I tried running 2:10 at a lower bus speed (DDR3-1750 or something) on my 1600MHz memory and it wouldn't even POST, even at 10-10-10-28 timings. But it'd do 1666 at 9-8-8-24 (my original overclock was 166x21 turbostep).
 
You must log in or register to reply here.
Back
Top