How do I decide which 3770k to keep?

Hello fellow overclockers

So I bought two used 3770K's and I want to use one of them in my main rig and sell the other.

Id like to keep the one that OCs the best. Im planning on delidding the keeper.

So Ive been testing them out, but at 4,7 at 1.375v I hit like 95c. It does not crash and when its delidded it may prove to be very good. So what should i do? Just delid the coolest at lets say 1.3v? See which one will boot at 4.6ghz with least volts? Maybe 4.8 at least volts?

Thoughts?

Ive never had this good a dilemma before, hope you all have some nice tips.

I dont wanna sell a delidded cpu to someone who doesnt know much about it.

Welcome c0olking, If you're just looking for a good daily overclocker, see which one will run at the least voltage for X Oc and keep it. If you want to extensively test them to see which one runs at X Oc with the least voltage. Set an Oc say 4.5 and stress it using Aida 64, Prime 95 or IBT. See how far you can get each chip and on what voltage. It may take a little patience but you'll figure out which one likes more voltage.

Mandrake4565 said:

Welcome c0olking, If you're just looking for a good daily overclocker, see which one will run at the least voltage for X Oc and keep it. If you want to extensively test them to see which one runs at X Oc with the least voltage. Set an Oc say 4.5 and stress it using Aida 64, Prime 95 or IBT. See how far you can get each chip and on what voltage. It may take a little patience but you'll figure out which one likes more voltage.

Click to expand...

What I have to do is set the ratio fx at 46 and then see which cpu will boot into windows with least vcore or should I attempt to do a stresstest aswell? Then do the same at 47 and 48 to see if a pattern appears?

Im just thinking one cpu may be able to run, lets say, 4.8. But it will require alot of vcore.
Then the other may not even be able to boot at 4.8, no matter what vcore I throw at it.

Im running a custom loop and planning on delidding, so the temps would be taken care of, but if the chip wont run Xghz, then my custom loop and delidding wont make any difference??

Making any sense??

c0olking said:

What I have to do is set the ratio fx at 46 and then see which cpu will boot into windows with least vcore or should I attempt to do a stresstest aswell? Then do the same at 47 and 48 to see if a pattern appears?

Im just thinking one cpu may be able to run, lets say, 4.8. But it will require alot of vcore.
Then the other may not even be able to boot at 4.8, no matter what vcore I throw at it.

Im running a custom loop and planning on delidding, so the temps would be taken care of, but if the chip wont run Xghz, then my custom loop and delidding wont make any difference??

Making any sense??

Click to expand...

You're on the right track overall but certainly do some stress testing - Some chips will act OK until stressed to show themselves for what they really are under fire.

As for your cooling, if speaking of higher MHz with better cooling that's neglegible with Sandy and Ivy Bridge chips at least.

These chips will only go so far depending on the chip itself and even sub-zero cooing if used won't net but perhaps 200 or so more MHz from it, that being something the Witchdoctor himself said once and I believe it. Sub-zero would be to simply keep it from burning up under load, not for it to go much higher as it normally is with an AMD. Voltage needed obviously will be different with each chip but know some chips that run well need more voltage than others so I'd base it more on temps seen rather than voltage itself when checking things out.
Voltage is a variable you can control, temps are harder to manage but you can still influence it with how you tune things.

Delidding of course helps and will allow more stability under load with higher MHz used and thats a move in the right direction.

I'd have to say go ahead and delid both, stress test and choose the one that does best at higher speeds in MHz under load - Thats the thing you're wanting to find for your useage.

Also, in reality if I had two of those I'd keep both just in case something should go wrong and would need a replacement - That's just my opinion if I had two of them for myself.

The 3770K I have now is a good one, not delidded but tops 5.8 GHz under sub-zero cooling and I know it might do a little more if delidded but not wanting to mess with it, being used ATM in my everyday rig and posting with it right now in fact on air and it's doing fine.

Hoping this may help some at least and good luck!

McCoy are the Ivy bridge multi locked like the Sandy bridges? I have a 2500k that is locked to 5300 Mhz, it will do it all day long benching on Air at 5.3 w 1.48 V but even under DICE will not boot to windows most of the time it will not even post at any thing over 5300 Mhz. I've even tried 52*102 and nothing.

Mandrake4565 said:

McCoy are the Ivy bridge multi locked like the Sandy bridges? I have a 2500k that is locked to 5300 Mhz, it will do it all day long benching on Air at 5.3 w 1.48 V but even under DICE will not boot to windows most of the time it will not even post at any thing over 5300 Mhz. I've even tried 52*102 and nothing.

Click to expand...

The chip isn't locked, it just won't do anymore period when it reaches it's limit, that's how it is.

Ivy acts the same as Sandy in this way.
My 2600K won't get into the OS past 5200 for sure even though it acts like it wants to. It actually gives the Windows logo during bootup at 59x but stops there no matter what I do with it.

Gonna fill you in..


Testing area:

3770K "A" did 4.6Ghz 10min prime blend run and needed 1.3vcore and got 78c.
3770K "B" did 4.6Ghz 10min prime blend run and needed 1.23vcore and got 75c.
"B" doing "A's" first successful run (at 1.3vcore) got hotter than "A", at 82c.
"A" would not even boot into windows at 1.25vcore.

My conclusion is that at equal vcore "B" gets hotter, but "B" requires alot less vcore.

Things are looking good for "B".

Remember, Im not saying this is stable in any way.
This is a test to figure out which one to keep.

General settings used in bios, same for all runs:

PLL Voltage: 1.7
PLL Internal: Enabled
LLC: Extreme

(thoughts?)

Once you delid them temps will change for the better - Don't forget how they were made and the TIM that was used under the lid causing temp problems, for some chips it's worse than others.

Could be chip A is one of the leaky ones that does require more voltage BUT has the potential to go higher than chip B. When you hear the term "Leaky Voltage" it's used to describe a chip that demands more voltage to clock but gives better clocks in return with said voltage. As long as you can keep temps under control such a chip tends to be a winner vs a chip that clocks well with lower voltage and runs cooler for extreme OC'ing. However since you're talking about everyday use I don't believe that would be a factor to worry about so much.

I'll ask some of the others to weigh in on this - Guys, how about it?

I agree with what McCoy is saying.

For everyday usage under ambient cooling it seems B is the better chip for two reasons. First it uses less voltage for X Oc and if it is running a bit warmer then A it will likely benefit from the delid.

I've had to make a similar decision once before.

Though my decision was with a different CPU on a different socket type.

In my case it was between two Q6600's. #1 had a VID of 1.325v, and #2 had a VID of 1.25v.

At 3.0GHz #1 was at 1.25v to remain stable, whereas #2 was at 1.152v. And at 3.2GHz the differences are even greater CPU #1 needed 1.38v to remain stable, whereas #2 only needed 1.168v to remain stable.

However, CPU #2 always runs hotter even when it is at a considerably lower voltage for the same overclock. Say by as much as 3-5°C at idle, and 5-8°C under load.

I ultimately decided to go with CPU #2, because although it runs hotter I was able to reach a much higher clock speed while still using much less CPU voltage (CPU #1 stopped at 3.3GHz (1.432v), whereas with CPU #2 I could go all the way to 3.6GHz and still be at much lower voltage than CPU #1 (1.328v)).

I guess in my case it might be a bit easier, since the voltages required were so vastly different.

Personally, I'd go for a minimum 2-hour stability run of prime95 to find the minimum stable voltages for the two chips before you decide which to keep.

It's possible once delidded the trend with temps could reverse with them - B could then start running cooler than A once the lids are off and they are properly TIM'ed.
The extent of the mess they made inside each chip won't be known until the tops are off, I'd do that and retest as I suggested earlier to be sure.

Dr. McCoy said:

It's possible once delidded the trend with temps could reverse with them - B could then start running cooler than A once the lids are off and they are properly TIM'ed.
The extent of the mess they made inside each chip won't be known until the tops are off, I'd do that and retest as I suggested earlier to be sure.

Click to expand...

Agreed, from what I've read part of the issue with the 4X70k chips, regardless of whether it's the original Haswell or the DC, is some of the IHS's sit higher off the die then others. Accompany that with the crap Tim and/or improved crap TIM under the IHS and you get some chips that run a lot hotter then others, even at the same voltage.

I have a particularly hot 4770k, when I first purchased it I did a little comparison to another forum members 4770k. I was on a custom loop with a 360 Thermochill Pa 120.3, 6 fans in push pull cooling the cpu only, he was on a Thermaltake Water 2.0 AIO unit. Both running the same motherboard and ambient temps were 22c. At 1.23 v on the cpu I was running nearly 20c hotter then he was at the same voltage, while running prime blend. That said under subambient cooling mine has always Oced a bit higher then his.