Just de lidded my 3770k

Yaayuh! · 1 Aug 2012 at 20:39

buzzlightweight said:
4.9GHz and 5GHz is out of my reach, not the temps thou, its the volts, 4.9GHz @ 1.48v crashed prime after 5 mins, temps were 71c!!!!

I cannot get over how good the liquid pro is and the de lidding!!

Have you tried other voltages?
Also, what is your ram running at? You may need more VTT the higher you push it.

sadbuttrue · 1 Aug 2012 at 20:47

Makes you think - Did Intel use low quality TIM to prevent a huge number of fried chips due to insanely high voltage. The average overclocker would just keep increasing volts until temps got too high.

ince · 1 Aug 2012 at 20:53

buzzlightweight · 1 Aug 2012 at 20:56

Rossi~ said:
Have you tried other voltages?
Also, what is your ram running at? You may need more VTT the higher you push it.

ram is running @ 2133MHz 1.55v 9-10-10-28-1 vtt @ 1.1v

sadbuttrue said:
Makes you think - Did Intel use low quality TIM to prevent a huge number of fried chips due to insanely high voltage. The average overclocker would just keep increasing volts until temps got too high.

thats what i think

ince said:
Nice

ty

buzzlightweight · 1 Aug 2012 at 21:09

found this on another forum, i will stick at 4.8GHz 1.41v

What are you talking about?
Yes that is what I am referring to when i say intel recommend MAX, that is why i recommend max of 1.45v for air OC for any 24/7 OC. I think 1.52v is the maximum SVID range which could be possible for SVID to apply on its own maybe if you were under LN2 and higher frequencies were possible without setting the vcore.
1.52v isn't so harsh for the CPu anyways I usually go to 1.52v or 1.53v to find max air clock for fun, without damage. However I don't do it all the time.
That is why recommended voltages were provided, if someone else knows better and wants to provide their own numbers they can go ahead and do it. But I wont change my numbers unless there is some actual proof. So far they have worked pretty well for those using them.
If anyone wants to see what all cores are running at, right-click the CPUz screen and it will show you the speed of all the cores, as well as allow you to select which core you want it to show the frequency of.
That is a good OC hanghang, idk what your issue is what my table, they set maximum ranges.

This is after 40mins of prime, gonna leave on all night

Just click to enlarge

panyan · 1 Aug 2012 at 21:25

buzzlightweight said:
IBT done and under 75c

i believe for intel burn test to work properly there has to be more than one run as the outputs are compared to deduce the system is stable

Muel · 1 Aug 2012 at 21:29

When you say de-lidded, do you mean you removed the lid completely and bolted the cooler directly to the chip, or did you replace the stock compound with better stuff?

buzzlightweight · 1 Aug 2012 at 21:35

panyan said:
i believe for intel burn test to work properly there has to be more than one run as the outputs are compared to deduce the system is stable

i will have a go after i have finished prime, ty for pointing that out

Muel said:
When you say de-lidded, do you mean you removed the lid completely and bolted the cooler directly to the chip, or did you replace the stock compound with better stuff?

replaced stock ***p for liquid pro and refitted IHS

Twirlywoos · 1 Aug 2012 at 22:15

great results, can't wait to see what you get when it's cured.

buzzlightweight · 2 Aug 2012 at 21:16

so max temps after 10 hrs prime was 75c but i had to up the vcore to 1.424v.

gonna work on my turbo volts and current limits to drop volts

wazza300 · 2 Aug 2012 at 21:44

you wanna be carefull how far you push voltages,it might degrade the cpu,temps are fine but personally i wouldnt go above 1.4v cpu voltage

joeyjojo · 2 Aug 2012 at 22:19

Very nice results.

For IBT, as above, you'll want to do at least 5 loops, and you could do as many as 50 if you wanted to be certain.

Setting memory to "max" is good as you're getting close to peak performance of the chip which would be 32*4.8 = 153 GFLOPS. However the 111 GFLOPS in the previous screeny is a little low, probably just because it's the first run and hadn't got into the swing of it. You really want to see about 80% of max, or 123 GFLOPS.

Doing more runs might make it stabilise at 123 or more, but if not it might be paging, in which case you should set it to custom and use a bit less than max memory until you consistently see 123 GFLOPS or more.

IBT isn't as easy as just pressing "run", but it's extremely powerful.

panyan · 2 Aug 2012 at 22:28

joeyjojo said:
IBT isn't as easy as just pressing "run", but it's extremely powerful.

also remember to right-click the "start" button and enable the extreme mode

vectisitch · 2 Aug 2012 at 22:44

would you think this would be worth doing on my 2700k??? opinions please. i would'nt mind having a go if worth it

wazza300 · 2 Aug 2012 at 22:52

50 loops of ibt Is way to much,20 loops max

2700k chips are soldered so no need to de lid

If you turn off multithreading your gflops will increase,its a well known issue

buzzlightweight · 2 Aug 2012 at 23:11

wazza300 said:
you wanna be carefull how far you push voltages,it might degrade the cpu,temps are fine but personally i wouldnt go above 1.4v cpu voltage

you are right, i have dropped to 4.7GHz 1.33v, i didnt see the point in 100MHz and 0.10v MORE :eek:

joeyjojo said:
Very nice results.

For IBT, as above, you'll want to do at least 5 loops, and you could do as many as 50 if you wanted to be certain.

Setting memory to "max" is good as you're getting close to peak performance of the chip which would be 32*4.8 = 153 GFLOPS. However the 111 GFLOPS in the previous screeny is a little low, probably just because it's the first run and hadn't got into the swing of it. You really want to see about 80% of max, or 123 GFLOPS.

Doing more runs might make it stabilise at 123 or more, but if not it might be paging, in which case you should set it to custom and use a bit less than max memory until you consistently see 123 GFLOPS or more.

IBT isn't as easy as just pressing "run", but it's extremely powerful.

will do but where did you get 32 from? now im at 4.7 should i be looking at 120 / 150 gflops?

vectisitch said:
would you think this would be worth doing on my 2700k??? opinions please. i would'nt mind having a go if worth it

hell no you will kill your cpu!!!

joeyjojo · 3 Aug 2012 at 00:06

wazza300 said:
50 loops of ibt Is way to much,20 loops max

If you turn off multithreading your gflops will increase,its a well known issue

I used to think this but when it came up another member (forgotten who) claimed to have seen failure for 100 loops or something. He's built a lot of systems so I believe him.

buzzlightweight said:
you are right, i have dropped to 4.7GHz 1.33v, i didnt see the point in 100MHz and 0.10v MORE

will do but where did you get 32 from? now im at 4.7 should i be looking at 120 / 150 gflops?

That's wise.

32 comes from the number of floating point calculations possible per clock cycle. It used to be 16, AVX doubled it to 32 with the new IBT.

There should really be a sticky for how to use IBT.

wazza300 · 3 Aug 2012 at 02:10

20 runs is fine imo,each to their own,prime95 will find any further instability ibt/linx is a cpu killer

Yaayuh! · 3 Aug 2012 at 07:08

Tbh 5 runs is fine then run Prime. IBT has been known to degrade chips, it's far too harsh and even the best of Overclockers steer clear of it.
Best only to be used for a quick stability check, Prime 95 for overall system stability.

joeyjojo · 3 Aug 2012 at 10:25

Rossi~ said:
Tbh 5 runs is fine then run Prime. IBT has been known to degrade chips, it's far too harsh and even the best of Overclockers steer clear of it.
Best only to be used for a quick stability check, Prime 95 for overall system stability.

Yeah right, source? :rolleyes:

A CPU is designed to perform calculations. IBT just gets it hot by using more of the chip than other stress tests.

Here is the thread for anyone interested where Justintime says he's seen "numerous fails in the 50s, 70s etc." so recommends using 100 loops.
http://forums.overclockers.co.uk/showthread.php?t=18410424