An almost complete and most excellent delidding adventure with a factory-crooked IHS; safe vCore per Intel (sort of); stress testing to 5.2 GHz; 3D marks
After recently delidding, I like to express my 'Big Thanks'
to Valgaur, FtW 420, Swag, VonDutch
and many, MANY others that have helped with their postings
Parrt 1 - intro and backgroundWarning: Spoiler! (Click to show)
As part of running a software company, I updated my 'private stock' of five machines for my home office starting late last fall re I7 3770s and various Asus Z77 boards, eventually ranging from P8Z77-V LK boards to Sabertooth Z77 and ROG Max V Extreme.
Four of the machines are filled with 3770 'non-Ks' as they are part of a Virtual Machine (non-K's have locked multiplier / to 41/43 and virtualization; K's have unlocked multipliers but no virtualization). ALL machines run the max 32 GB of RAM and are loaded to the gills with heavy biz software such as multiple instances of SQL (incl. "enterprise") and IIS web servers (set to 'LocalHost') which I mention as it does affect at least vCore in overclocking.
While I used to do a lot of overclocking 'back in the day', I had not done serious OC'ing for about seven or eight years - my, how things have changed - PLL ? :-) All of the 'non-K' 3770 overclock quite nicely to at least 4.5 Ghz, and two to max multiplier x BCLK up to 107, so 4.6 Ghz (all with after market cooling).
That kind of whetted my appetite 'for more'...1st came a decent video card (GTX 670), then a second one for SLI (next to me on my desk is lying the 3rd one....). Then I picked up a 3770K ...which turned out to be from the Costa Rica 3229C series...I ran and posted some benches and had replies from folks who recommended the 'delidding thread'...the chip validated well over 5.3 Ghz with the lid on and with fairly moderate vCore for that speed (with more to come)
When I posted these screens for 100% load at 4.8, 4.9 and 5.0 Ghz, it meant
a.) DELID as this is a good chip that runs very fast on low vCore; just needs better temps at 5 Ghz daily
b.) DON'T DELID as this is a good chip hat runs very fast on low vCore - why take the risk to ruin it ?
c.) Some PM offers to purchase the chip from me
more 100% load / vcore /speeds prior to delidWarning: Spoiler! (Click to show)
The above, PLUS
The delidding itself will be a 3-part affair, with two of the three parts completed so far.
The first step involved applying new TIM to the 3 video cards 'as a dry run', though I skipped CL (the intended TIM) as there were some aluminum bits mixed in with copper parts on the cooler that would have come in contact with the CL (which doesn't play nice with aluminum).
Then I took the CPU out and placed the Hardware-store type blade on one corner...which was followed by a BIG realization: It was not going to work
- using a magnifying glass, I confirmed that the factory-mounted IHS 'was crooked' - on one whole side, you could not even see the black silicon-type glue as the metal of the IHS was basically sitting on the PCB...on the other side, there was a lot of room - more than I have seen in pics by others in this thread.
As I at that time only had the hardware-store 'bigger' blade, I put it all together again...and applied some Coollaboratory Ultra to the IHS (replacing what was a correctly applied Arctic MX4 TIM). Here are some direct comparisons of these two TIMS:
We then drove up to Whistler (long weekend in Canada), but I returned early on my own to get to the bottom of this ! Fellow OCers and especially Delidders' surely know how it is like ! The problem was that I was so tired yet full of hi-po espresso when I had another go at the CPU, this time starting with the crooked side that showed a lot of space and black glue.
Next - the actual delid described with too many words:Warning: Spoiler! (Click to show)
Still, I in the meantime had picked up the thinner 'Wilkinson' blade and I used that first...on the side with a lot of space, this was fairly easy though being really tired and my eyes wanting to shut did not help... yet the blade slid in effortlessly around 65% of the circumference - until I got to the spot where the IHS was virtually sitting on the pcb...working even the thin blade in there meant riding a bit on the pcb so I pushed 'up' as hard as I could - and actually ended up cutting into and notching the IHS (think of it as extra venting now).
Finally, I pulled out the bigger blade again and placed it where there was the most black glue on the pcb on the free side - then twisted it. The IHS flew off and the blade went into not one but two fingers...
I could see that on the 'problem corner', there was NO black glue left
on the pcb to scrape off, and there were some drag marks of the blade on the pcb, but they had not gone 'through the top green layer' just marked it. I claim divine intervention because I was half asleep when I did that, so it could not have been me
Of more concern though were various copper like spots on other parts of the pcb - did I scrape right through there with the bigger blade ? Nope, they turned out to be pieces of copper from when I forced the blade into the IHS and notched it. Close call
After cleaning up the IHS and pcb, it was time to apply some CL-U, and I think I got just the right amount on it. I placed the CPU carefully into its socket and then very gingerly put the IHS on top - set back a bit to account for the 'forward-slide' when closing the metal clamp even with my finger pushing onto the IHS center top, though I know that I am about 1.5mm too far back still
- don't think that it will matter much as the die is not touching any sides and the IHS is otherwise seated firmly on the die with the correct amount of CL.
But there was another problem - or two. For now, I have been using a closed loop Thermaltake 2 Extreme water cooler (the one with the Asetek block / pump and the 240 mm rad). This is a really good solution, other than its mobo back 'plate (rubbery plastic') and retainer 'soft metal nuts' are very flimsy...this set-up has wandered across three different motherboard types, and on two of those it has been 'on and off' at least 5 times.
By now, some of the spacer rubber pieces are only half the depth they used to be, and on one corner, one of the retainer nuts has worked itself deep into the rubber / plastic backing plate - meaning that even on full lock it will still have a bit of play and thus not be perfectly seating the water block on the IHS. Furthermore, one of the bandages had come off and I was literally bleeding onto the IHS / mobo - time to act fast and rational...
Given the upcoming 'next step' per below which includes a new Koolance block / custom loop in just a few days, I decided NOT to use CL P on the IHS for just a few days as I had read that it can be a bit hard to get off. Further, I knew that the current water block fit would be marginal at best, so I plastered far more MX4 on it than I normally would, in part to make up for the bit of play at one corner, and knowing that it just had to last a week, because:
...paraphrased per "Jaws" We gonna need a bigger boat
...it was difficult enough to wedge an eATX Max V Extreme into the Antec 302 case, complete with two video cards, but there is no way a 3rd will fit plus a custom cooling loop
...so here is the 'bigger boat'
(still in dry dock, also will be the home of the IvyB-E in the fall with dual GTX Titans)
Warning: Spoiler! (Click to show)
"Early / Temporary Results"
Per above qualifiers (CL U on die, but too much MX 4 on ill-fitting water block), have a look-see at these early results for 5, 5.1 and 5.2 GHz 'delidded'
:5.1 and 5.2 GHz:
Warning: Spoiler! (Click to show)
I hope the new loop with 360 double-thick rad w/5 120mm fans, IHS to Koolance 370 SI CPU block with CL Pro in between will improve these already impressive temp gains per delid and partial CL. I always could run 5 Ghz 24/7 with for that speed 'modest' vCore, but temps were a bit of an issue - not anymore, even before the 'big' cooling upgrade. Per above, even 5.2 Ghz is very stable at benching.Safe vCore - ?
My experience with max vCore is : Not to post it as that topic has already seen some 'heated' discussions in this thread (isn't that right, VonDutch
That said, I tend to use what I consider an excellent overclocking tool in addition to the posted OC for IvyBridge Bios settings on this site - said tool is Intel's own 'Extreme Tuning Utility
'. While it has a nice option to raise vCore all the way to kingdom-come, it won't actually apply any more than 1.525v (+ - a small variance, depending on PLL). I take Intel's wisdom to heart, as much as I have other tools that can raise vCore all the way to 2vs. I might validate beyond the 5.32 Ghz I already did way back at slightly more vCore, but in the end, unless I go to LN2 later with this chip and board once IvyBridge-E has arrived. I take Intel's wisdom seriously.
Finally (if you made it this far
), some 5 GHz 3D benchmarks...including some 'extreme' settings...even with the less than perfect temp cooling solution, no core ever went past 72 C on the most intensive CPU tests
more pics / bench results
Warning: Spoiler! (Click to show)
Edited by Joa3d43 - 2/12/13 at 9:00pm