Overclocking

#81

Untitled-65.png1495x836

Just a quickie showing current OC at current voltage. I’d consider this being one of the poor i7s out there (not as bad as C0 920s, but worse than D0 920s). Could of been predicted though due to its chances of being one of the first retail batches of 940s in the country.

29.4’C ambient temp

QPI/VTT 1.260
IOH Core 1.120
DRAM 1.58
QPI PLL 1.16
LLC is on as always.

Anything not listed there has not been touched, such as CPU PLL.

Currently running a test with 0.016 less vcore (1.296 in windows) and running 1.24 QPI/VTT. So far so good, will post the result tomorrow. It seems that bumping up the QPI PLL has allowed me to lower some other voltages that I thought were the original cause of late linpack errors.

#82

nice OCs guys

#83

Running at 3.8 atm 4 is doable but with a lot more volts…Happy with 3.8 :yes:…

Vcore 1.325
Vtt 1.3
PCH 1.080
Cpu Pll 1.82

#84

Good stuff Mac

#85

Untitled-66.png940x870

Managed to drop the Vcore & QPI/VTT down a step and still remain stable.

29.6’C ambient temp

Actual Vcore is in the SS
QPI/VTT 1.240
IOH Core 1.120
DRAM 1.58
QPI PLL 1.16

It was a surprise honestly as it seemed as if no amount of QPI voltage I used under what I previously had would get a stable OC (would get LinX errors but no BSOD), seems that I had to tweak a certain aspect of the QPI which has allowed me to drop other votlages. I’m going to see if I can go any lower.

[COLOR=#000099]QPI PLL VCore[/COLOR] (default: 1.1v, <1.4v is pretty safe)
What it does:[INDENT]Keeps on-chip memory controller in-sync with bclk.
[/INDENT]When to raise QPI PLL VCore:
[LIST]
[]Try raising this along with Vcore and VTT, but in smaller increments.
[]Helps stabilize higher CPU Uncore frequencies and QPI frequencies (in CPU feature)
[]Try raising this when you increase memory clock speed via multiplier.
[]Try raising when LinX produces errors after a few minutes without BSOD
[/LIST]

http://www.evga.com/forumsarchive/tm.asp?m=100494809

That thread is very handy for X58 owners, and most points are still valid to P55 owners.

#86

nice mac … bout time you threw a vantage bench up too imo :stuck_out_tongue:

To save the chatbox oc overload lol…

John : i think my error lied in all the chatter about vtt and its need to be boosted to get a sniff of 4ghz , or 191/200 bclock

So because i had that high … everything else had to go up …

id check your bclock with vtt at lowest(staying within .5 of ram) … then chip the bclock up and vtt if needed … and grab 10mhz jumps… i did it but thoguht prime would suffice to check it … so now i know the stupidity of that … ill be ibt testing it to see what 191/200 requires …

plus ill reseat the hsf when i get home to check temp difference… pity its gonna take so damn long just to reseat it lol

#87

very nice Nip .

thats something i am looking into now … cpu pll … i have had it set at 1.8 ( lowest) and am seeing what effect it has in the grand scheme of things

while they are different things … its one thing i havent really read on … just locked it in and not thought about it

#88

CPU PLL is a real funny votlage. Some people actually find greater stability lowering it, I don’t need to touch it thankfully.

[COLOR=#000099]CPU PLL VCore[/COLOR] (default: 1.800V, spec range: 1.65V-1.89V)
What it does:[INDENT]Keeps CPU clock in-sync with bclk.[/INDENT] When to raise CPU PLL VCore:
[LIST]
[]May help with stability while increasing the bclk or CPU multiplier.([COLOR=Red]or may make it worse[/COLOR])
[]May help with stability past 210 bclk if you observe that during runtime the QPI Link (found in E-Leet) bounces too much.
[][COLOR=Red]Not a commonly raised. May actually cause instability. Test this variable alone.[/COLOR]
[/LIST]
You know CPU PLL VCore s too high when:
[LIST]
[][COLOR=Red]Its possible you could actually gain stability by lowering this[/COLOR].
[/LIST]

[COLOR=Gray]One of those voltages that nobody really has a firm grasp on, lol.

Out of curiosity, what has made you go from LinX to IBT? Do you just prefer the GUI or was LinX bugging up for you? Only reason I ask is because LinX and IBT both use linpack to stress the system, they are the same program with a different interface.

It is also fairly accepted that running a high CPU PLL (as in over Intel’s max) will fry a chip quicker than high Vcore.
[/COLOR]

#89

making my 2.8 > 3.4ghz seem so low

#90

sorry … i use linX … just get used to calling it IBT as near on every thread i read on oc they ue it lol

does the same thing though i would assume.

yeah cpu pll is a bit unknown … much like when we were looking at it for the 8400 lol … just whack it at 1.56 and see if it does anything lol.

i do remember seeing that mine goes in .1 jump … so 1.8 … 1.9 … which im not keen on … will look into it again but will test the bclock with imc volts first.

as im on the very bottom side of the .5 range … i might need to sneak it up 1 notch to 1.168( from 1.1625) just to keep it in spec perhaps … as my ram is specced for 1.65v

#91

10 linX runs and 7 hrs prime overnight … a good speed.
the temps are only prime… linx was 72 hottest core from memory

now to try for 4 lol

#92

Yeah, they are the exact same program with different GUIs. LinX is popular over at XS as that is where it was developed. A lot of OCAU boys also use it as many are also members of XS. You can’t go wrong with either program imo. LinX lets you be a bit more specific in terms of setting parameters though.

Nice voltage and temps there, I don’t see any 10’C variances? I’d say that you have an acceptable mount. No need to reseat unless you want to chaise getting 1 - 2’C differences between each core which may be possible, though it could take a lot of reseatings until you get lucky. Anything from the amount of paste, to how evenly you tighten down each corner will play its part.

#93

WOOT! Made 4GHz, tho hardly in style :stuck_out_tongue:

Vcore - 1.382 or so in BIOS. This is the kill, i’m bout 0.7V up on what made 3.8 stable.
Vtt - 1.290. Couldn’t get 200 bclock with 1.27, seems to be needed.
vram - 1.64. Haven’t refined this, rams running at default stock/timings
PCH - 1.12
PLL - now at 1.9. Will mess with this more, was at 1.8V til I got to 4GHz, will test 4 at 1.8 now too.

Definitely going back to 3.8 for 24/7. A lot lower volts and a hell of a lot cooler!

#94

Gpu @ stock…

#95

nice johnny, tis a bit warm though :smiley:

#96

OK! So, nothing else changed except PLL moved down to 1.7V.

Still stable at 4GHz! BUT, max temps are now 81/82/80/84 (compared to 87/86/873/87 when PLL = 1.9)

Huge temp change for same stability!

edit: i’m thinking it’s a bad assumption to think that’s all the PLL voltage’s doing. Another test and i’m back up near 85. Still, 1.7 works at least

#97

Most of these chips(i5-i7860 etc) seem to have a wall@3.8 anything above requires a shit load more voltage to hit 4.0 I,m happy with 3.8 with less volts better temps to run 24/7… :wink2:

#98

Like said before, nobody really knows the effects of CPU PLL.

One other tip that may effect P55 CPUs like it effects X58 CPUs, even multipliers are really unstable compared to odd multipliers on the same voltage.

If I were you John, I would leave Turbo mode on like az has, so you get a 21 multi and then put your base clock at 191MHz.

The only draw back of doing this is that your ram will not be running at 1600MHz but you will make up for it by tightening timings. It may substantially drop your Vcore if the 750s suffer from the same problem as the 920/940.

EDIT: seems like it’s still true.

Pushing the board as far as we can we shot for 4.37GHz at a 230MHz base clock and a 19x multiplier: remember the Lynnfield CPUs are reported to overclock better with odd multipliers.

#99

turbo mode is off im pretty sure … but you can force the 21 multi … most can do it now.

Yeah … after 3.8 it just gets inefficient … so much voltage to gain 200mhz … im happy down in the cool zone for 24/7 … unless i manage ( as talking to john) to get a miracle combo for settngs

#100

Either way is good then, using turbo mode or just forcing the 21 multi, I would be using the absolute max multi in an attempt to lower voltages across the board.