~Intel Core i3/i5 32nm Dual Core OverCLARKing Thread~

[radaja]

just so i understand,you are setting the vcore manually in the bios?or are you using auto?
i haven't experienced this vcore increase from changing multiplier.i know these 1156's behave
differently when LLC is enabled than the 775's did.vcore increases under load,where as with
my 775's boards vcore usually dropped slightly under load with LLC enabled.im going to give
the multiplier thing a quick go to check.

EDIT works fine on mine,although when i used the dynamic vid control and left it to
normal instead of manually setting the 0.625 increment to add what i wanted,and it
booted at 1.38v for no reason.i guess its like using auto

[PiLsY]

233 bclk failed despite bumping both vtt and vddr. Im finding it fails if I increase vtt at all - don't know if this is related to vcore:vtt ratio or if the chip just doesnt like over 1.34vtt (maybe until I get off stock cooling?).

Anyway, not willing to risk getting vddr>vtt+0.5v yet, so left vddr at 1.78v and dropped vtt back to 1.34v. Im guessing I have some vdrop/vdroop, so bios difference of 0.45v is my limit. Will see if its stable when I get home in about 8 hours...

[WhiteFireDragon]

no voltages in bios is on auto, everything i set manually. i guess it's just my mobo then. it's weird how biostar lets us change vcore by adding or subtracting vcore instead of entering in an actual number. for example, if i want a resulting vcore of 1.35v, i need to pick "+0.15v", or if i want 1.50v, i need to choose "+.30v". i guess this is what's causing my voltage differences. really bad voltage regulation design IMO

233 bclk failed despite bumping both vtt and vddr. Im finding it fails if I increase vtt at all - don't know if this is related to vcore:vtt ratio or if the chip just doesnt like over 1.34vtt (maybe until I get off stock cooling?).

Anyway, not willing to risk getting vddr>vtt+0.5v yet, so left vddr at 1.78v and dropped vtt back to 1.34v. Im guessing I have some vdrop/vdroop, so bios difference of 0.45v is my limit. Will see if its stable when I get home in about 8 hours...

233 bclk is a pretty respectable number. even high end x58 and p55 have a hard time reaching this, so it's mobo limitation. after a certain point, increasing voltage won't do much. btw, vdimm at 1.78v is pretty high. i think you should tone it down a little.

[PiLsY]

Close your eyes when it gets under water with 1.7vtt / 2.1vddr then for some 2200mhz action....

There's more bclk in this board. I've used no voltage adjustments yet aside from cpu, vtt and ddr. PCH, QPI etc all at default.

Got bored so remote desktop'd to my pc at home and bumped the bclk more. Testing 235 now, looks to be at least close to stable (30 mins p95 blend so far).


Anyone know how uncore works on these? What multiplier is it and is it just tied to bclk, or can mem ratio or qpi ratio reduce/raise the multi? Just wondering as im needing 1.7+vddr to get over 230bclk no matter what I do with mem ratio or timings (indicating uncore requires the extra vddr, not ram).

[radaja]

heres my latest,temps were 38c during the run.sub 9 score
http://valid.canardpc.com/show_oc.php?id=1042059
SuperPi 32m and everest cache & memory benchmark
20x235@4700 w/1.42v DDR3-1880 9-9-9-24 1t



Edit,i just notice in my bio when i leave trfc on auto it says 74
but in cpuz it shows 134.so i set it manually to 72 and it now shows
corectly in cpuz.and my everest score is much better now even at a
lower cpu clock.
manually setting trfc to 72 in bios

[TinTin]

Here are my result with a i5 661

200X23, 1.4V set in BIOS, air-cooled by TrueCopper, Ambient Temp ~ 22C, DDR2000, prime95 for a few minutes, seems not too bad.....



A quick 5G Validation...


247bclk: air-cooling+air-con



200X24, 4.8G SPI32M stable:


200X25, 5GHz SPI32M stable:



240X20, 2:8, DDR3 1920 CL7 SPI32M stable:

[WhiteFireDragon]

almost 1.7v pumped through that thing is crazy. i'm surprised that fragile chip can handle that on air. TinTin, what's your vtt at for 5ghz 32m run?

[TinTin]

almost 1.7v pumped through that thing is crazy. i'm surprised that fragile chip can handle that on air. TinTin, what's your vtt at for 5ghz 32m run?

~1.45V VTT is for 240bclk, for 200X25, I use 1.325VTT...
CPU skew ~ 500ps DELAY,
CPU PLL ~1.9V seems have something help - air-cooling

I find for higher BCLK with this MB, 2:8 ratio is much easier, with 2:6 what I got is 237.5 bclk

[PiLsY]

Same here, 8x better than 6x for bclk, cpu and mem clocking. Gigabyte P55 UD5 and i3 530 - looks like this may be the case across various platforms?

[TinTin]

almost 1.7v pumped through that thing is crazy. i'm surprised that fragile chip can handle that on air. TinTin, what's your vtt at for 5ghz 32m run?

SPI 32M is not so stressing.
I just worry about running high Vcore with LN2 and doing some Wprime32M/1024M benching.....the easiest way to kill clarkdale

[radaja]

TinTin,very nice OC's.i wish i had an i5-661.

heres an updated everest score with trfc set manually to 72
at 20x235@4700 DDR3-1880 9-9-9-24

[KanGaXx]

Hi guys,

which is the max BLCK or BUS that you can handle with 660/661/670 with *10 moltiplier on ram ? for example, i found that i can do 210bus with mems @ 2100 ( *10 molti ) but 211 no boot, which is your experience ?
i have a p55a-ud6 bios f7.

Thanks

[radaja]

heres my latest.
http://valid.canardpc.com/show_oc.php?id=1043603
21x229@4809 DDR3-1832 9-9-9-24 1t
SuperPi 32m and everest cache & memory benchmark

[TinTin]

TinTin,very nice OC's.i wish i had an i5-661.

heres an updated everest score with trfc set manually to 72
at 20x235@4700 DDR3-1880 9-9-9-24

raise the QPI freq...performance will be better...

[TinTin]

Hi guys,

which is the max BLCK or BUS that you can handle with 660/661/670 with *10 moltiplier on ram ? for example, i found that i can do 210bus with mems @ 2100 ( *10 molti ) but 211 no boot, which is your experience ?
i have a p55a-ud6 bios f7.

Thanks

for Clarkdale, 2:10 is poor for bclk 200+......I can't even get 210bclk, DDR3 2100

[KanGaXx]

for Clarkdale, 2:10 is poor for bclk 200+......I can't even get 210bclk, DDR3 2100

hi thanks! i can boot at 210x10 but at 211 it's stuck, so its normal...hope they will fix in the future

[radaja]

raise the QPI freq...performance will be better

TinTin,do you mean use the x40 so its at 4700MHz?

[WhiteFireDragon]

one thing i notice clocking a clarkbar on a p55 mobo: you can't change the GPU/video frequency, so how high can the bclk be before the GPU limits the OC?

btw radaja, nice clocks on the memory. these clarkbars have sucky IMC's, so it's good to see you have it pretty high. are you increasing vtt every time you're trying for higher memory speeds?

[WhiteFireDragon]

*oops, double post *

[radaja]

WhiteFireDragon,thanks.yea i just did some
real quick testing this is what i found for my board
200-220bclk 1.25v QPi/Vtt
220-235bclk 1.31v QPi/Vtt

[radaja]

TinTin,thanks for the advice about QPI link speed
i went to x40 and got an increase in superpi and everest.

19x235@4465 DDR3-1880 9-9-9-24 1t w/QPI at x36@4230


19x235@4465 DDR3-1880 9-9-9-24 1t w/QPI at x40@4700


20x235@4700 DDR3-1880 9-9-9-24 1t w/QPI at x36@4230


20x235@4700 DDR3-1880 9-9-9-24 1t w/QPI at x40@4700


boost in performance is very nicesince memory performance
isnt that great to begin withnow if i only had some cas8
DDR3-2000 ram i would be golden,i could push it even further

.....................................

[tranngocminh]

System

• MSI H55M-E33 (BIOS 1.5)
• Intel Core I3-530 // StockFan
• G.SKILL PerfectStorm 2x2GB 2133CL8
• Nhiệt độ môi trường: 24*C

Result

• OC Switch 20%: I3-530 @3511MHz (22x160BCLK)
  
  • LinX 0.6.4: 20 round, passed
  • Prime95: BlendTest - 2h, passed
• Manual: I3-530 @4016MHz (22x183BCLK)
  
  • Prime95: BlendTest - 3h, passed

Screenshot

• OC Switch 20%: I3-530 @3511MHz (22x160BCLK)
  
  • LinX 0.6.4: 20 round, passed
    
    
    
    
    
    
    
    
    
  • Prime95: BlendTest - 2h, passed
    
    
  
  
• Manual: I3-530 @4016MHz (22x183BCLK)
  
  • Prime95: BlendTest - 3h, passed
    
    

[TinTin]

TinTin,thanks for the advice about QPI link speed
i went to x40 and got an increase in superpi and everest.

19x235@4465 DDR3-1880 9-9-9-24 1t w/QPI at x36@4230


19x235@4465 DDR3-1880 9-9-9-24 1t w/QPI at x40@4700


20x235@4700 DDR3-1880 9-9-9-24 1t w/QPI at x36@4230


20x235@4700 DDR3-1880 9-9-9-24 1t w/QPI at x40@4700


boost in performance is very nicesince memory performance
isnt that great to begin withnow if i only had some cas8
DDR3-2000 ram i would be golden,i could push it even further

.....................................

You are welcome! Nice boost in performance, did you try 21X...5XXX QPI will be great!

[KanGaXx]

someone with p55-ud6 or p55a-ud6 ?

i have some problem to manage memories cas 7 2:8 over 222bclk.
if someone can give me tips.

( voltage tweak do not help either subtimings )

i have bios F8, with F7a the situation was worst, max was 205, so i notice an improvement but i am not satisfacted yet.

[mzs_biteme]

http://valid.canardpc.com/show_oc.php?id=993326


that's my highest with both cores and HT enabled.

chip: i3-540
board: Gigabyte H55M-S2H

My highest with only 1 core enabled and no HT is somewhere north of 5ghz with 1.6 vcore and I forget the qpi. still haven't gotten below 8 seconds on spi1m but that's all good I'm not worried about it. my 24/7 speed is 200x23@4600mhz(1.375vcore) with default QPI(x32 I think?) and def ram values. I'm pretty lazy lol.

the diff between this one and my e6400(I ran it @3.5ghz) is HUGE. FPS on l4d2 and css went up massively. great cpu I highly recommend for the price.



oh and hey do you guys run with HT on or HT off? I can turn my 24/7 oc up like 100mhz with HT off and I don't notice any diff in performance with it on...

EDIT:: LOL, posted my e6400 instead of my i3 540...

Hehe, got the same setup from Fry's last week... $125 after MIR...
Here is what I got so far without even trying...



Great to see others have great results, especially considering the price of the combo....

And oh yeah... I upgraded from e7300@3.8GHz and this setup (i3) totally p0wns compared...