Haswell i5 4670K/i7 4770K batch and o/c results

[aussie-revhead]

Guys, I saw something weird tonight with my 4770K. I have spent the last few days priming different combinations of settings trying to reduce vcore at 4500 and while I was watching cpuz my cpu downclocked to 39x but it was not too hot, it was only about 65 at the time, so what else can make it throttle?

[Alex-Ro]

Disable cpu thermak feedback and vrm fault management in bios,also cpu integrated vr efficiency mangement should be set to high performance

[aussie-revhead]

Thank you, just seems odd I had temps of 100 degrees before delid and it didn't throttle (that I saw). Now I was testing working down from 1.30v, I have a screenshot of 16hrs priming still holding 4500mhz so I dropped the volts a little further to 1.250v and this happened....

Weird.

Before bumping up system voltages it needed 1.30v to be stable, now its down to 1.250v and still tesing.

[LardArse]

you should remove vttddr from your speculation since it only goes to the dram not even to the cpu so how can it hurt the cpu?

[Phaseshift]

See I don't think it is cache voltage for me, as I have not used more than 1.6v under LN2 and input on the last dead one was 2.5, then only thing I have in common is high vdimm/vttdr i.e. >2.00v, so maybe thing a connection there?

I've killed 2 chips already; one was simply because I tried 1.5v to see if it would boot into 50x multi. with vccin on 2.2-2.25 (this was on an h100i cooler lol)

The other chipped died on 1.25v but vccin was set to 2.2-2.25.

[cstkl1]

So far priming with c-states which requires higher io voltages n very picky on the dram clock thingy... Both proc surviving.

Before this was full manual with i/o voltages can run at 1v up to 2266.

[gravimag]

i don't get it.. i tried x50 and i left VRIN (Vccin) on AUTO and it was still holding at 1.8V. why u guys trying to bring it up? i tried 2.1V for my OC it didn't make it more stable...

[chispy]

you should remove vttddr from your speculation since it only goes to the dram not even to the cpu so how can it hurt the cpu?

Hi Peter and thank you so much for clarifying this , My good 4670k I thought it was dead or IMC was hurt just to find out is working great again on M6E and Z87-Pro. After last night benching at -115c 3 hours into the benching session everything was going great , so I was as happy as a dog with 2 tails , I did what Dumo suggested me , and I did not run the Ram ( Sammies TX LV cas 9 ) at very very thight timings second and thirdiaries. It ran flawlessly even with uncore at 5120.7 as before it was not stable with that uncore so high , I was running at 5925Mhz 1.88v. core and I tweaked all other voltages for stability and find the sweet spot it felt good I saved a lot of profiles already due to so many tweaking on the voltages and checking for stability. After 3 hours into benching flawleesly 3D at those speeds I got a BSOD and then when I re-boot code 55 , clear settings re-boot again code00 , re-boot again code 55. At this point I hit the memory OK button "w00t" amazing feature in Asus boards , I was able to boot again at same settings and cpu speeds but , I could only run the Ram at stock speeds or Auto otherwise code 55. I ran a few more benchies at 5924Mhz cpu but with Ram at auto. sometimes It would get stuck at multi x8 only in windows. Me been a curious guy. I shut down the rig and started looking with a magnifiying glass at the board M6E , I took one piece out of the board at a time vga all ok , memory slot had water in it due to condensation eventhought I did an awesome job ( or so I think ) in the insulation , even put some Vaseline in the Ram contacts. To end the Story , all the troubles I have had with the good 4670K Haswell was due to condensation and water finding its way to the Ram slots , I have no idea how it got there but that is findings why cpu was acting crazy. I guess tonite when I go cold again will put lots of Vaseline in the Ram slots and try to insulate that part of the board much better. Asus Memory OK button feature saved my life , great feature on Asus boards.

Example #1 , CPU running flawlessly 01 in the first 3 hours: CPU at 5925Mhz @ -115c , Uncore at 5120 ,Ram at 1338 cas9 thight. GPU was old , eVGA GTX 260 216sp running on chilled water.






Example #2 , CPU started to act weird , code 55 , code 00 and sometimes cpu multi stuck at x8 inside windows that was after 3 hours subsero. from this point on forward I could only bench at Ram speed on Auto settings but same cpu speeds.




Conclusion , the problems was not the cpu or the motherboards , The problems was water into the Ram slots . So take this as an example guys , if your cpu act weird and you think like me ( cpu dying ) , please check and verify your Ram slots for water or condensation around Ram slots.

[acebmxer]

isnt vrin(Vccin) supposed to be 0.400 to 0.500 higher then vcore?

[Alex-Ro]

I've killed 2 chips already; one was simply because I tried 1.5v to see if it would boot into 50x multi. with vccin on 2.2-2.25 (this was on an h100i cooler lol)

The other chipped died on 1.25v but vccin was set to 2.2-2.25.

I think you guys are setting VCCIN too high for air,i saw no gains above 1.9 honestly,on air...

Played with dice and used 1.7 vcore and 2.2 vccin,high dram voltage,tight latencies,whatever,cpu survived for 6+ hours at -70.

For testing cpu/binning under cold i would start with a vcore of 1.7 and 2.2 vccin as a base,then see if increasing vccin actually gains something,no point in jumping to 2.5 vccin from start...Start lower with voltages and see where the curve gains/voltage ends,unless of course you some old QE6S ES chip that can handle 2.5 vcore no problem

Also don't forget about temperature compensation option from ln2 mode

[hersounds]

4770k 5 Ghz 1.31 Vcore - WC - NO IHS Mod

[Net]

hersounds : Hmm happy man came across too cpu.

[gravimag]

in my opinion changing VRIN makes no difference... no matter what i do with Vcore and ratio, VRIN (Vccin) when left on AUTO stays at 1.8V... kinda strange, i know.. but MOBO designers must know better then me :P
as far as i know, and tried, VRIN has no impact on stability whatsoever (and i tried 2.1V), Vring does. for 4.8GHz at 1.27V i need 1.3Vring to get this going.. less then that and BSOD.
i tried and made x50, with all auto except Vcore, but checked with voltmeter what is VRIN (Vccin) and Vring is... i was surprised to see VRIN staying at 1.8V, but Vring going up to 1.34V.

[chispy]

4770k 5 Ghz 1.31 Vcore - WC - NO IHS Mod

Well Done Hernan , very nice CPU

[C-N]

4770k 5 Ghz 1.31 Vcore - WC - NO IHS Mod

Yes vnice CPU to even run 5G @ 1.31 but I cant believe your numbers.... gflops are 1/2 what they should be & your temps are almost idle temps for 1.31v! Linx cannot be loading that CPU mate.

[l0ud_sil3nc3]

Yes vnice CPU to even run 5G @ 1.31 but I cant believe your numbers.... gflops are 1/2 what they should be & your temps are almost idle temps for 1.31v! Linx cannot be loading that CPU mate.

It has to be no AVX instruction set, so no SP1 or the wrong version of linx, either way it looks a decent chip.

[Praz]

Yes vnice CPU to even run 5G @ 1.31 but I cant believe your numbers.... gflops are 1/2 what they should be & your temps are almost idle temps for 1.31v! Linx cannot be loading that CPU mate.

4 minutes per run for 6.5k memory usage is also not right for the reported CPU frequency.

[PolRoger]

New Bios on UD5H runs lower vcore

4.5GHz L308B202

nice and at uncore 1:1 at 4.5ghz.

whats ure vccin , digital/analogue , vssa voltages??

btw ures delid?? ure min to max is pretty good if its antec kuhler 920.

Thanks mate, its a good enough 24/7 clock for me.

vccin/vrin = 1.9 ioa & iod = +.050v vssa = stock

Yes its had a roof chop & has liquid ultra on the core & yes its on the 920. Normal silent idle is 35 ~ 40 then the fans kick up after 40oC

@ C-N,

How much CPU Cache/Ring voltage for your sample to run uncore 1:1... 45x?

[SonDa5]

Yup

only tested 1.35vcore, I am sure it can go lower.



seems like a good batch, do you live in California?

Got same L312B152 batch from Frys in Burbank today.

Don't plan on benching on super cold but will do ambient temp water cooling for gaming. Hoping for 5GHZ.

[Phaseshift]

I think you guys are setting VCCIN too high for air,i saw no gains above 1.9 honestly,on air...

Played with dice and used 1.7 vcore and 2.2 vccin,high dram voltage,tight latencies,whatever,cpu survived for 6+ hours at -70.

For testing cpu/binning under cold i would start with a vcore of 1.7 and 2.2 vccin as a base,then see if increasing vccin actually gains something,no point in jumping to 2.5 vccin from start...Start lower with voltages and see where the curve gains/voltage ends,unless of course you some old QE6S ES chip that can handle 2.5 vcore no problem

Also don't forget about temperature compensation option from ln2 mode

Yeah I think I'm going to leave vccin on auto for now since I didn't have a dead chip until I started messing with vccin. I still don't get why 1.25v vcore and 2.2 vccin killed my 2nd chip though.

[C-N]

@ C-N,

How much CPU Cache/Ring voltage for your sample to run uncore 1:1... 45x?

Hi mate,

I had it at 1.3 & it looks like it needs all of it for 45/45 1:1. I arrived at that figure asthats what it needed to occt 4.8 with stock uncore. I found i needed more & more ring voltage as i walked the chip up from 4.5,6,7 etc.

[SonDa5]

4770k 5 Ghz 1.31 Vcore - WC - NO IHS Mod

Well Done Hernan , very nice CPU

That is awesome. I just bought my 4770k with the intentions of doing the same mod. Hope to have some success as well.

[Rob94hawk]

@ C-N,

How much CPU Cache/Ring voltage for your sample to run uncore 1:1... 45x?

Hi mate,

I had it at 1.3 & it looks like it needs all of it for 45/45 1:1. I arrived at that figure asthats what it needed to occt 4.8 with stock uncore. I found i needed more & more ring voltage as i walked the chip up from 4.5,6,7 etc.

Any benefit to running 1:1? I thought is wasn't necessary?

[andressergio]

4770k 5 Ghz 1.31 Vcore - WC - NO IHS Mod

Very nice Hernan !!! great CPU !!!

[C-N]

Any benefit to running 1:1? I thought is wasn't necessary?

LOL about 1" inch on my e-peen my friend that's all

Seriously though nothing you'll notice in daily use. When benching raising the uncore closer toward the CPU speed you will see small benefits.