The Sandy Bridge OC Thread

Status
Not open for further replies.
yashchitale

yashchitale

Discoverer
Hello,
My i5 2500k has been running stable at 4326mhz(103 blck* 42 multi) at 1.288v vcore and 70*C under prime95 with the Asus P8P67 Pro mobo. This was done with auto tuning in AI suite ii. Is there any possibility i could bring the vcore voltage down? according to what i read at hardforum i have to adjust the VRM frequency to 350. Should i leave the LLC on standard or keep it high atleast?
 
you certainly can and should bring down the voltage. I have an i7 2600k and I run it @ 4.2 Ghz on stock voltage (1.174) as my 24/7 configuration and for benching i go upto 4.5 Ghz @ 1.24 v. Also refrain from OCing the bclk and using AI suite to OC. Instead use high vdroop & LLC, enable PLL overvoltage, disable EIST and just adjust the multi to 43 or 44 and gradually reduce your voltage from 1.288 in 0.001 v steps until you are not stable anymore. Then just go up to the last stable voltage & leave the system running prime95 blend test (amount of RAM to be used= System RAM- 1GB) or IBT (Maximum stress test) for atleast 3+ hrs. Save the OC profile in EFI and you are done.
 
Yep, up to 4.3GHz should be possible with stock voltage itself. I am running my i5 2500k @ 4.5GHz 24/7 on 1.2V. The voltage you are giving looks way too much for that OC.

Rather than relying on the auto tuning feature, just try to OC directly from UEFI. All you have to do is to set the Turbo multiplier to desired value and leave everything else at default. Just keep a check on the voltages while running prime 95 and tune further accordingly.
 
I have HT, VT and Speedstep and all other stuff enabled.

I am not OCCT stable at 1.27 at 42x100 on my Core i7 2600k.

At 1.330 I am stable at the same setting.

< 1.4 is ideal

1.4-1.45 is borderline and the max you should think of

>=1.5 NOT recommended, stay off that limit

As long as you don't touch 1.4 and have a good OC and your full load temperatures with OCCT don't cross 60-70C OR around 50-60C with gaming/Prime95, you should be fine :)
 
looks like you got a bum chip mate. 1.27v for 4.2 Ghz is way too high for the i7 2600k. I also have all the stuff like VT-x, turbo, HT etc enabled and I have disabled EIST by choice as it doesnt drop the voltage at idle, just the clocks. I can do 4.2 at stock voltage and my gaming temps are 54-56 C. Mind you i dont have AC and my room's kinda toasty at 33-34 C.
 
This was done with auto tuning in AI suite ii. Is there any possibility i could bring the vcore voltage down?
Click to expand...

...Yes you could bring it down thru BIOS, AI Tuner is a good rough guide, and does over volt a bit to ensure stability IMHO.

according to what i read at hardforum i have to adjust the VRM frequency to 350.
Click to expand...

I have set mine to 350, they say this is needed for higher clocks.

Should i leave the LLC on standard or keep it high atleast?
Click to expand...

Really up to you, I was able to achieve 5.1Ghz with either Auto or Extreme.

51ghz.jpg


7.348 @5.1Ghz

CPU-Z Validator 3.1

1789227.png




NOTE: The above is on Stock Cooling :cursing: !!

Just a quick test run to test the chip out. Waiting my Cooling systems to come in.

Please do not try this out you will surely burn out your chip in no time if you are on stock cooling with this !!! I have brought it back to 4.4Ghz for now.

:)
 
Just curious, @Infected: Did you get the new Tower cooler with your 2600k or just the punny cooler like us lesser mortals on 2500k?
 
Guys.. worked like a charm.. tried all sorts of combinations.. In the end left EIST on.. Left LLC and VRM freq to auto.. Used offset instead of manual voltage.. offset voltage for 4.2ghz was -0.095v and for 4.3ghz -0.070v.. I dont wish to OC more.. Temps looking great.. And Prime95 ran stable for 2 hours each on both profiles..

Overall result: Temperatures dropped almost 10 degrees Celsius..

If not visible, Vcore @ 4.2Ghz - (1.168v in screenshot, goes upto 1.176v), 4.3Ghz - 1.192v

Currently using 4.2ghz profile..
 

Attachments

  • OC 1.jpg
    OC 1.jpg
    91 KB · Views: 147
  • OC2.jpg
    OC2.jpg
    92.1 KB · Views: 156
Lord Nemesis said:
Mine doing 4.5GHz @ 1.2V.



Uploaded with ImageShack.us
Click to expand...

Excellent thermals. By the way you keeping speed step and auto management features on or off..?

I think I got a good chip, should be able to get 5GHz with 1.3~1.35V.[/QUOTE]
 
@dev: Thanks..
2 questions:
When testing with Intel Burn test chip passes under all leves staying below 67*C. But when testing with Prime95 i get an error saying expected result was 0.4 and i got 0.5, hardware failure.. Please tell me what should i incur from this.. And there is an option round of checking.. Should that be checked? It is disabled by default..
Secondly, I am using TIM which came with my Hyper 212 plus. It is grey and thick and was hard to spread. Should i leave this be or get mx-2? Also on benchmarkreviews it is written one should flatten out the surface using wet sand paper.. Should i do this?

Benchmark Reviews: Performance Computer Hardware Tests
 
asingh said:
Excellent thermals. By the way you keeping speed step and auto management features on or off..?
Click to expand...
Speedstep is off currently. Other than that the only manual settings are the turbo multiplier and the CPU voltage. Everything else on auto.
yashchitale said:
@dev: Thanks..

2 questions:

When testing with Intel Burn test chip passes under all leves staying below 67*C. But when testing with Prime95 i get an error saying expected result was 0.4 and i got 0.5, hardware failure.. Please tell me what should i incur from this..And there is an option round of checking.. Should that be checked? It is disabled by default..
Click to expand...

You should infer that Prime 95 is much more stressful than the Intel Burn Test. :P

To tell the truth, there is not much point in running these super stress tests (Like Prime 95 or OCCT) insanely for 24 hours like some people like to do. Even 3 hours is an overkill. If you can run Prime 95 for like 30 mins with safe temps, rest assured that nothing you throw at your CPU during your regular use would push it over its limits. If the OC is too unstable for 24/7 use it will show up in the first 10~15 min itself.

yashchitale said:
Secondly, I am using TIM which came with my Hyper 212 plus. It is grey and thick and was hard to spread. Should i leave this be or get mx-2?
Click to expand...

In my opinion the stuff CM bundles with their HSF's is crap. At least that was how it was when I got my last CM HSF. Use that stuff only as a last resort. If you can get better stuff like MX-2 or Sunbeam TX-3, TX-4, go for it. I shaved around 5 deg when I moved from the CM stock paste to Arctic MX2 using the same HSF.

yashchitale said:
Also on benchmarkreviews it is written one should flatten out the surface using wet sand paper.. Should i do this?

Benchmark Reviews: Performance Computer Hardware Tests
Click to expand...
That process is called lapping. It can done on the CPU IHS or the HSF contact surface or both. Frankly speaking its a lot of hard work for shaving off a few degrees of the thermals. Its done by serious overclockers who are already using the best HSF's and still want more out it. Its definitely not an exercise for the practical overclocker who is looking for a quick performance boost for 24/7 use. Finally it requires a lot of patience to use sand paper for slowly grinding away the warranty from the back of your 10k+ costing CPU.
 
okay.. Thanks for that.. So the expected 0.4 from prime95 and the actual result 0.5 means i need to ramp up voltage more? And also about the Round off checking option? Keep that off or enable it?
 
^^ Yes, increase it a wee bit more. Since you are running on 1.19V, increase it to 1.2V and see how it goes. If every thing is fine, see if you can raise the multiplier to 44 and test. As for the round off setting you are talking about, I don't know about the exact setting, but in general, the goal of running the benchmark is that your OCed CPU should not be running into any computational or rounding errors during the run as that would be an indication of an unstable CPU.
 
Status
Not open for further replies.
Top Bottom