Overclocking i3 530 @ 4.7Ghz

forteh

forteh

forteh

Associate
Joined
18 Jan 2010
Posts
224
Location
Lichfield
I am trying to push my cpu as far as I can in order to maximise performance in arma3, I'm not interested in synthetic benchmarks :)

I've got an i3 530, originally supplied as a 4.02Ghz bundle from ocuk and it's been rock solid but arma3 has been bringing it to its knees, severely limiting fps at times. Now ideally I would get an i5 760 and overclock but I'm on an absolute minimum shoestring so pushing this cpu is the first thing to do.

I have got it stable at 4.62Ghz but looking for advice on going further.

Machine spec is: -
i3 530
Gigabyte GA H55M UD2H v1.0 Bios F11
Thermalright truespirit 140BW
2x2Gb Corsair XMS3 1600 (pc10700) slots 2 & 4
2x4GB Crucial 1600 (pc12800) slots 1 & 3
Gtx660

OC settings so far are (off the top of my head): -
21x multi
220 bclk
8x RAM divider (1800Mhz, 11-11-11-28-39)

LLC enabled
CPU Vcore 1.35v
QPI/Vtt Voltage 1.39V
PCH Core 1.3V
Dram Voltage 1.5V
All other voltages set to auto.

32-gigabyte-ga-h55m-ud2h.jpg


I have proved the bclk stable up to 225mhz (with a 10x multi).

The ram appears to be stable at that speed/timings, it has passed multiple memtest86 without error, the sticks are fairly warm but voltage is stock (infact undervolt on the corsair sticks). The timings are taken from the 1600Mhz spd given by the crucial sticks, the 1333mhz corsair sticks seem to be keeping up ok though. I want the ram frequency as high as possible as this makes a major difference to fps in arma3.

This setup is prime95 stable with a peak temperature spike of 79°C and typical average temperatures around the 74-76°C mark. TJmax is 105°C.
The Vcore reported by cpuz is around 0.2V higher than the bios setting, I guess this is the DviD bumping it up, I tried disabling it but that caused massive instability.

If I try to push up to the 225Mhz for the 4.725Ghz I need to push the Vcore up to 1.4v (reads as 1.42V in cpuz) which obviously adds more heat with momentary peaks up to 84°C and typical average temperature of 79-80°C.

4.725Ghz isn't 100% stable yet though and I'm not sure if I have just reached the limit for the IMC (reported to not be that strong on the clarkdales). Should I be pushing more volts into the QPI/Vtt, if so what sort of limit should I stick to?

If I put the multi back up to 22 it should will be running at 4.95Ghz but I'm not sure what voltages to play with to achieve this, is it purely Vcore that drives the multiplier stability or does QPI/Vtt need to get involved aswell?

At some point I intend to upgrade the cpu, ideally to a quadcore with 8mb L3 cache (i5 750/760, i7s). I am investigating possibly using a xeon x3430, you can pick them up for less than 30 quid and as I have a stable 225bclk I should in theory be able to put it straight up to 4.05Ghz (18x multi) From what I have read, the IMC on the i7/xeon is very strong so should stand a higher bclk than the i3. Am I correct with these assumptions?

TL : DR I have 4.62Ghz stable but I want to go faster :D
 
Last edited:
vtt is needed if your running a lot of memory or high ram speeds 1.4v would be my limit,the imc controller is on the mb and not built into the cpu like on x58

to use less vtt you could try a slower ram speed?

if you wanted to upgrade the i7 860 would be a good choice due to it having 4 cores/8 threads which would help a lot in cpu heavy games like arma 3

you can try a 22x cpu multi aswell
 
Ah I was under the impression that the IMC was on die for the clarkdales (as per http://www.overclockers.com/3-step-guide-overclock-core-i3-i5-i7/).

Switching to a 6x ram divider would give me 1350Mhz @ 225bclk, this would give more overhead but lose the significant increase I get from having the higher frequency.

i7 860 is well outside of my budget, the i5 is to be honest :(

What should I be looking at to get the 22x multi stable though? Obviously I don't have much headroom on the QPI/VTT voltage without slowing the ram which I would rather not do if possible.
 
more vcore/qpi/vtt I would have thought

some cpu's are more stable on an odd multi or an even multi aswell
 
Aye, had read about the stability improvements on odd multis, also read that the clarkdales don't have problems with it! Now I have proved the ram and bclk stable to 225 I might try dopping it back to 200Mhz and bunging the multi back up to 22 and see if that will go any higher :)
 
I upped the QPI/VTT to 1.41V (nearest increment from 1.39V) and ran it at 4.725 (225x21) but prime95 kept crashing after 20 minutes or so, not an error like normal but a full on crash and windows error reporting box. I installed OOCT and it runs without issue peak temps up to 84°C with 79-80°C typical.

I tried to get 4.62 with 210x22 on 1.4vcore & 1.41QPI/VTT and prime immediately fell over, the lower multi is far more stable at lower voltages.

There shouldn't be a requirement for more VRAM should there? It passed memtest @ 1800mhz on 1.5v without issue.
 
should be ok if theres no errors

prime blend tests a lot of ram,so could need touch more vtt rather than dram

you can try 1.55v dram though it wont hurt
 
Ill up the vram a touch and see if it stabilises, I am somewhat suprised that the 1300mhz corsair ram is running at 1800mhz on a 1.5v undervolt!
 
With vram at 1.56v it ran 20 mins of prime before a bsod, it's subsequently run over an hour of blend without a hitch!

And after all this I'm looking to get a Xeon in the near future :D
 
Nope not prime stable @ 4.725 even with 1.425 vcore, 1.41 QPI/VTT and 1.56 vram. Considering it's stable @ 4.62 with 1.35 vcore, 1.39 QPI/VTT and 1.5 vram I guess I've reached the limit of the processor :(

Think I'll pick up a Xeon x3430 and see how far that goes :D
 
Do a Cinebench R15 single and multi core run!

(Quick tip, set it to run with high priority in task manager as that prevents background processes from slowing the run down).
 
Single core cinebench run is 127, multi core run is 333, for some reason it's reporting the cpu speed as 4.84Ghz rather than the 4.65Ghz it actually is :/

Are those scores good or not? :D
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom