Overclock.net › Forums › Intel › Intel CPUs › Delid 3570k without a razor blade
New Posts  All Forums:Forum Nav:

Delid 3570k without a razor blade

post #1 of 62
Thread Starter 
G'day all,

First post, first intel processor since my 286, first overclock and first de-lid - so be gentle.

I'm doing a lot of work with Autocad virtualised, so my Piledriver server just does not have the single thread performance I'd like, thus I bought a new Asus P8Z77-V PRO/TB and 3570K with the intention of overclocking it.

I appear to have bought a dog as it requires about 1.42v to stay stable at 4.5G. I've got a CoolerMaster Seidon 240M on it, but I was bumping into throttling during Prime95 runs.

I had a bit of a read about de-liding the proc to help temps, but I did not have any blades that were thin enough so I tried a different tack.

The HS on this chip has a step in it, so I grabbed the top part of the HS with it inverted in a bench vise. I then knocked the HS loose by placing a block of pine against the PCB and gently thwacking it with my calibrated hammer. This dislocated the glue enough for the HS to just peel off with a bit of help from a swiss army knife blade.

No damage to the PCB, and a mild scuff to the sides of the HS.

Never thought to take pics. It's back in the machine now and running ok. Unfortunately I only had Arctic MX-3 to hand, so it has some of that both under and on the HS.

Interestingly, I had to re-seat the HS 3 times before I could get the system to boot. It appears that if the HS gets mis-aligned while you clamp the CPU back down it must interfere with something. I thought perhaps the pads on the edge of the board, but it seemed to be far enough away from those. Anyway, loosen the clip, move the HS and re-clamp and its all good.

I did a few searches before posting this, but the only references I could find to using a hammer to de-lid appeared to be in jest.

I just wish I could get it stable at 4.6G.
post #2 of 62
Thread Starter 
So, I got it "stable" at 4.6G. I thought I'd try removing the HS entirely, but found the CPU clamp got in the way, so it had to go also.


Off with its head !

I found that without the HS I had to try quite a few shuffles of the whole assembly to get it to boot, and then it ran *hotter* and far less stable.

I put that down to not enough pressure on the waterblock as it was designed to press on the extra height of the HS. Rather than fart around modifying the waterblock I just put the HS back on, but did not bother with the clamp.




Now we are "stable" at 4.6G with a voltage offset of +.185 (gives me 1.44V under load mostly).

I had to use a much smaller LLC (it's on the 25% setting) and a higher overall offset otherwise I'd get MCE's as the load came on and off.

My load test consists of Prime95 blend with 29.5G of ram allocated, and every 5 minutes an additional make -j2 kernel compile. I've found it will look stable with Prime95, but chucking a big compile job in to hit the disk as well tends to show up instability a *lot* faster.

Will let this run overnight and see how it goes I guess.

(Yes, I really did de-lid that chip with a vice, a hammer and a block of wood)
Edited by fnarfbargle - 3/12/13 at 3:58am
post #3 of 62
Nice work, I would of loved to have seen a video of you taking a hammer to a 3570K tongue.gif
post #4 of 62
yeah same here GeforceGTS, that would have been awesome. good job fnarfbargle. thumb.gif
post #5 of 62
Thread Starter 
Well, well.

Prior to the delid, and running at 4.5G I was sitting at between 85-95C under load with the rare spike into the limiter (kernel tells me when it drops into and out of throttling). I could not get any stability above that no matter what I tried.

After the delid I sit at about 75C under load, but bumping up to 4.6 & 4.7G with a vcore of 1.46V does not raise the temp any. Looks like a significant win for thermal conductivity.

Bytemark says the system is getting faster, and it certainly needs a bit more vcore (nudging 1.46v) but I'm still getting ~75C core temps under load at 4.7G. I wonder how much harder I can push it?

Cheers lads. Next time I do something agricultural like that I'll be sure to get some snaps or a vid.

Initially I was going to just grab the HS and twist the board off with a pair of multi-grips. I figured the glue would have very little shear strength so it'd pop right off, but it turns out the glue has so little shear strength that a good lateral shock was all it took to break it free. Bonus for not scratching the board or die.
post #6 of 62
Nice job man! Thats some scary stuff! Then again, taking a blade to a ivy bridge is scary too. It all depends on the hands that are performing the action I guess. I don't know if you have hear of Coolaboratory Liquid Ultra, but that is what you should use as a TIM on your chip and combined with your water cooling you might be able to see temp drops around 25c at the LEAST. It is basically liquid metal and is decently easy to brush onto your die and IHS. The price is a little steep, $22 after shipping, but it is worth every penny.
post #7 of 62
Yikes...good stuff. Although i wouldve sold that CPU and tried to get another (if return/swap was not possible) - maybe that could hit 4.7Ghz, then delid it wink.gif
post #8 of 62
Thread Starter 
Quote:
Originally Posted by jerrolds View Post

Yikes...good stuff. Although i wouldve sold that CPU and tried to get another (if return/swap was not possible) - maybe that could hit 4.7Ghz, then delid it wink.gif

Yeah. Unfortunately (or fortunately actually) I live in Australia and there is SFA market for second hand processors, so it's not that easy to just run through a few until a good one pops up.

I had another crack at direct-die cooling this afternoon, but it turns out the cooler is sitting hard against the edges of the socket. Looks like I'll need to machine up a thin copper shim if I want to try that properly.

I've just ordered myself a 3770K, so hopefully that will clock a bit better. I have a solid rocks stable system at 4.5G with the 3570K, but I could really use the extra L3 and it gives me an excuse to buy and try another CPU.

Ooo.. there's the SMS telling me I can go and pick up the new proc. I promise if I end up knocking the top off this one I'll get a vid of me & the hammer.
post #9 of 62
Thread Starter 
Ok, so the 3770K will clock to 4.7G but bumps against the thermal limiter. 1.48V. Goody. An excuse to de-lid the new chip.

How to de-lid an Ivy Bridge in under 15 seconds :http://www.fnarfbargle.com/private/130315-Delid-vid/IMG_0081.m4v
I used MX-3 on the die and HS (I meant to use NT-H1 but must have grabbed the wrong syringe).

Dropped about 10C overall temps.

Now to fiddle with moar speed
post #10 of 62
Quote:
Originally Posted by fnarfbargle View Post

Ok, so the 3770K will clock to 4.7G but bumps against the thermal limiter. 1.48V. Goody. An excuse to de-lid the new chip.

How to de-lid an Ivy Bridge in under 15 seconds :http://www.fnarfbargle.com/private/130315-Delid-vid/IMG_0081.m4v
I used MX-3 on the die and HS (I meant to use NT-H1 but must have grabbed the wrong syringe).

Dropped about 10C overall temps.

Now to fiddle with moar speed

Ah awesome, thanks for the vid, it makes it look so easy!

I lost feeling in the end of my thumb for two days from the amount of pressure I had to put on the razor blade when I did mine, blade was to thick and I scratched it redface.gif Wish I had seen this method and had access to a vice :l
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › Delid 3570k without a razor blade