Suggestions for DC overclocking :)

[Splave]

Please try using lower vccin!

In my quick testing, 5200mhz can run 32m at 1.500vCore 1.600vccin on air cooling. 1.700vccin hangs, 1.800vccin hangs, 1.900vccin hangs. Seems a good starting point is around 1.5 to 1.55.

 

You must keep vccin atleast .080v higher than vcore or it will lower your vcore under load.

 

Havent tried this theory yet on ln2 but its working on air so far.

 

Sharing in hopes that others will try it, post results below!

Edited July 13, 2014 by Splave

[l0ud_sil3nc3]

Please try using lower vccin!

In my quick testing, 5200mhz can run 32m at 1.500vCore 1.600vccin on air cooling. 1.700vccin hangs, 1.800vccin hangs, 1.900vccin hangs. Seems a good starting point is around 1.5 to 1.55.

 

You must keep vccin atleast .080v higher than vcore or it will lower your vcore under load.

 

Havent tried this theory yet on ln2 but its working on air so far.

 

Sharing in hopes that others will try it, post results below!

 

Good guy Splave

[Vivi]

Wow splave! You are awesome .

 

This makes alot of sense that they maybe worked on the pwm a little.

 

Maybe those guys can unlock their 6.1ghz lol

 

Ill add a link to this thread from mine, how to oc dc

[Leeghoofd]

Interesting facts... so under LN2 it could be we need max around around 2.0-2.1 with these pentium CPUs...

[Massman]

Good guy Splave

 

²nd that

[Vivi]

Yes leeg someone should try it on pentium aswell. Its haswell refresh so not sure if it will be the same as devils canyon but maybe this new pwm is on all coming cpus

[steponz]

I wondered this myself... I wonder if vccin helps though for cb?

 

Haven't tested pentium under cold yet. But I will this week. Ill post my results.

[steponz]

Tested 4790k yesterday, Was running 1.5 vccin with 1.35 volts 5ghz 32m np. Seems we not need to raise vccin anymore.

 

I wonder if my issues on cold were related to vccin where it would just power down for no reason at a range of temps.

 

Ill have to get some ln2 today to test out the theory.

[nedernakker]

verified with my G3258. chip loses clocks when lowering from 2.250 to 2.0 with 1.85V Vcore.

 

So for my chip it isn't working

[dinos22]

Put some system spec up also fellas

[zzolio]

Tested 4790k yesterday, Was running 1.5 vccin with 1.35 volts 5ghz 32m np. Seems we not need to raise vccin anymore.

 

I wonder if my issues on cold were related to vccin where it would just power down for no reason at a range of temps.

 

Ill have to get some ln2 today to test out the theory.

 

i have had the same issues and low vccin help me

[Splave]

verified with my G3258. chip loses clocks when lowering from 2.250 to 2.0 with 1.85V Vcore.

 

So for my chip it isn't working

 

Didnt work on my pentium or 4770k either but 4790k it is helping

 

My Pentium likes 2.3-2.5 and 4770k 2.7 all day...

[der8auer]

verified with my G3258. chip loses clocks when lowering from 2.250 to 2.0 with 1.85V Vcore.

 

So for my chip it isn't working

 

Intels said the golden rule for VRM is to have at least VCCIN = 0.4 V + Vcore.

 

My 4670K doesn't like low VCCIN at all. For best results I need 2,55 VCCIN

[Xtreme Addict]

I tried lower VCCIN on DC and on some cpus it was better (on air and ln2), on some worse. I saw the same with 4770K, most need 2.5-2.7v for high clocks, but I saw cpus which prefer 2.1-2.3v

[Vivi]

Yes i suppose its chip based. But allens discovery is still awesome, to be alongside shamino in thought .

 

If you look at z97 gene mobo you can see an option called "keep vccin as close to vcore as possible". Shamino smelled it!

[sin0822]

Looking at it from a different point of view:

Higher delta (VIN-VCore) = more work for the FIVR = higher temps in the FIVR = could explain the helpfulness of higher VIN for CBs as it would raise some die temps, espcialyl in the wierd substarates they used in the FIVR that could be the root cause of the CB on haswell.

 

However the opposite is true as well, lower delta = less work for the FIVR = lower die temperatures which is why it helps on air.

 

They did some improvements to the FIVR from Haswell >>> DC, they added capacitors to the CPU's FIVR, these are on the underside and most likley input capacitors, meaning the FIVR is now being provided cleaner power, which could explain lower VIN needed to achieve staiblity.

 

I had 3 DCs, one needed 2.1v VIN to be stable at 4.875ghz, I am talking like 2 hour stress test stability, not wPrime or SPI, those could differ. THe other two really didn't care if VIN was higher or lower.

 

I tested 2 pentiums over the weekend, one loved high VIN, it woudl keep climbing in frequency only if I pushed both VIN and VCore, the other HATED higher VIN, I was stuck using 2.2v VIN.

 

I think the gains you are seeing are from the lower temperatures that the smaller delta provides, the deltas which are made possible because of the better input power, but that is just speculation, I am not the engineer who designed DC or the FIVR(nor do I want to be the latter).

[l0ud_sil3nc3]

Yes leeg someone should try it on pentium aswell. Its haswell refresh so not sure if it will be the same as devils canyon but maybe this new pwm is on all coming cpus

 

Pentium eats VCCIN, I have not found a chip that needs less than 2.5 for 5.8+

[steponz]

Pentium needs high vcinn.

4770k needs high vcinn.

4790k, keeping vcinn low, helped with my cb.

 

When vcinn was high, I had cb at -95, when low -105.

[sin0822]

maybe only the DC has the better FIVR

[nXXo]

for my 3258 2.65 2.7v only if i put 1.9v or 1.95 for1.88v of vcore bsod immediatly

[Vivi]

Ye its only for DC because it has new resistors at back for cleaner power. So wont make sense for pentium.

 

My 4770k also needs alot of vccin. If the balance is out i get typical 124 bsod

[Splave]

you guys can try lowering System Agent even below stock added alot of stability for me in threaded benchmarks as well

 

x57 would not pass HWBP at 1.25v, can loop back to back at .780v please try this too if you have time.

 

Edited July 17, 2014 by Splave

[MaJ0r]

Pentium needs high vcinn.

4770k needs high vcinn.

4790k, keeping vcinn low, helped with my cb.

 

When vcinn was high, I had cb at -95, when low -105.

 

I don't think that 4770k need high vcinn. Maybe I'm wrong but I reach 7.1GHz with 4670k(I know it's not 4770k but i think rules are the same) with vcinn about 2.3-2.4V. I don't remember how high vcinn I can set on Z87 OC Formula but for 6.6 Pifast i used 2.3 vcinn and increasing didn't help me.

[Vivi]

Yes varies from chip to chip

[CL3P20]

vccin on some batch of Pentium K will help a lot.. others will hurt. Definite "sillycone"

 

3418B7xx - CB @ -120c ;scaling under cold with 2.92v vccin, up to 1.91v vcore - 5.75ghz through R15 and XTU

*stable at above volts and temps through rash of multi-thread bench, back/back

 

3418B9xx - CB @ -105c ;scaling under cold with 2.4v vccin, up to 1.66v vcore - 5.65ghz through R15 and XTU

*stable at above volts ONLY at ~ -86c through multi-thread bench.. colder = random reboots, regardless of volts

 

 

**going back to 'Auto' for SA and IO's, when pushing for stability at next multi is proving helpful for me... if everything posts, good then try for more vcore. Clocking these little pentiums can be fun.. just need to 'slap' and 'tickle' all the right spots with finger on the GO button!