Bare-die testing: A delidded 3770k, an H100, and 9 different TIMs

[Idontcare]

I see quite a few posters have chimed in with questions and comments - I promise I will address them all in short order. But first I wanted to wrap up the cliff-hanger I left off with a couple days ago...

And by cliff-hanger I am of course referring to the test conducted with what can only be adequately characterized as a silly-stupid obscene amount of NT-H1 placed on my 3770k :whiste:

Is that an ear of corn I see in there? D:

(I don't care who ya are, that's funny right there)

After testing, naturally I removed the H100 from the mount and naturally the CPU stuck fast to the H100.

^ this happens with all the TIMs, it is nothing unique to this specific test condition, just thought I'd share the photo.

One thing to note here is that despite the ridiculous amount of NT-H1 I added, it still wasn't enough to cause a bunch of TIM to squeeze out and make a mess of everything. There was enough of a gap between the CPU PCB and the H100 that the NT-H1 was able to basically stay within the confines of the CPU PCB.

^ this is the process I went to using for removing the CPU from the H100 if I wanted to preserve the imprint left in the TIM on the H100 and on the 3770k as best as possible with minimal smearing. It looks like I am training to become a dental hygienist "open wide please!"

Once I got the CPU off the H100, here's what the imprint on the H100 looked like:

^ We can seen that despite the copious quantity of TIM used the CPU die itself actually made relatively good contact with the H100 as indicated by the visibly detectable areas of exposed copper in the central rectangle area.

And the die itself?

Not bad. Yeah there is crapload of NT-H1 sitting on top of the PCB itself surrounding the die but you can also clearly make out the CPU silicon die sitting in the middle indicating that basically the compression of the H100's mounting pressure did the intended job of ensuring a reasonably good mount was attained.

And the results:

Interestingly the idle temperatures were completely unaffected by the slathered layer of NT-H1.

No big surprise here in that the excessive amount of TIM resulted in elevated temperatures when the CPU is under a load, we all expected this, but I don't think we expected the magnitude per se.

It is surprising to me just how minimal the impact was in terms of thermal performance. At most we are looking at 6°C increase in operating temperatures at the very high end of the OC'ing spectrum of 4.9GHz.

I expected this test to be the epitome of worst-case scenario, that it would be a disastrous setup which would result in operating temperatures that were far worse than those observed with IC Perihelion or TX-1.

Instead what we find is that even under such a worst-case completely silly over-application of NT-H1 the resulting thermal performance only a smidgen worse than the best results obtained with IC Diamond (within a degree or two) and was still better than the best results obtained with TX-2, AS5, MX-4, IC Perihelion and TX-1.

What this goes to show is that at least for this particular setup and application (bare die mounting with an H100), the amount of TIM you use is only of secondary importance to the specific type of TIM you are using.

Start off putting your best foot forward by choosing to use a quality high-performing TIM and at worst you are going to bungle the application and mount and set yourself back by 5-6°C or so.

That may be news to absolutely none of you, I may be the last person on earth to have come to this realization, but even if that is the case I am still happy to have at least caught up to the rest of you and it only cost me a $10 tube of NT-H1 for the education :thumbsup:

 

[hokies83]

I want to see the Liquid Ultra results when applied correct.

On die u only use enough to coat the die there should be no run off or puddles.. like u painted it on very clean..
Like u see with the IHS i did as an example of how much Liquid Ultra to use.

You can do the same thing with the back side of the IHS very thin coat.. this is how the Liquid ultra works it is different then other tims..

As im sure u could see when u used it.

Try this and i bet temps will drop another 10c.
See look.. this is the IHS a much larger area and i used 1/4 of what u used on the tiny die..

VS

There is a huge difference there.. So im hoping u give it another go!

 

[naisanza]

I see quite a few posters have chimed in with questions and comments - I promise I will address them all in short order. But first I wanted to wrap up the cliff-hanger I left off with a couple days ago...

And by cliff-hanger I am of course referring to the test conducted with what can only be adequately characterized as a silly-stupid obscene amount of NT-H1 placed on my 3770k :whiste:

Is that an ear of corn I see in there? D:

(I don't care who ya are, that's funny right there)

After testing, naturally I removed the H100 from the mount and naturally the CPU stuck fast to the H100.

^ this happens with all the TIMs, it is nothing unique to this specific test condition, just thought I'd share the photo.

One thing to note here is that despite the ridiculous amount of NT-H1 I added, it still wasn't enough to cause a bunch of TIM to squeeze out and make a mess of everything. There was enough of a gap between the CPU PCB and the H100 that the NT-H1 was able to basically stay within the confines of the CPU PCB.

^ this is the process I went to using for removing the CPU from the H100 if I wanted to preserve the imprint left in the TIM on the H100 and on the 3770k as best as possible with minimal smearing. It looks like I am training to become a dental hygienist "open wide please!"

Once I got the CPU off the H100, here's what the imprint on the H100 looked like:

^ We can seen that despite the copious quantity of TIM used the CPU die itself actually made relatively good contact with the H100 as indicated by the visibly detectable areas of exposed copper in the central rectangle area.

And the die itself?

Not bad. Yeah there is crapload of NT-H1 sitting on top of the PCB itself surrounding the die but you can also clearly make out the CPU silicon die sitting in the middle indicating that basically the compression of the H100's mounting pressure did the intended job of ensuring a reasonably good mount was attained.

And the results:

Interestingly the idle temperatures were completely unaffected by the slathered layer of NT-H1.

No big surprise here in that the excessive amount of TIM resulted in elevated temperatures when the CPU is under a load, we all expected this, but I don't think we expected the magnitude per se.

It is surprising to me just how minimal the impact was in terms of thermal performance. At most we are looking at 6°C increase in operating temperatures at the very high end of the OC'ing spectrum of 4.9GHz.

I expected this test to be the epitome of worst-case scenario, that it would be a disastrous setup which would result in operating temperatures that were far worse than those observed with IC Perihelion or TX-1.

Instead what we find is that even under such a worst-case completely silly over-application of NT-H1 the resulting thermal performance only a smidgen worse than the best results obtained with IC Diamond (within a degree or two) and was still better than the best results obtained with TX-2, AS5, MX-4, IC Perihelion and TX-1.

What this goes to show is that at least for this particular setup and application (bare die mounting with an H100), the amount of TIM you use is only of secondary importance to the specific type of TIM you are using.

Start off putting your best foot forward by choosing to use a quality high-performing TIM and at worst you are going to bungle the application and mount and set yourself back by 5-6°C or so.

That may be news to absolutely none of you, I may be the last person on earth to have come to this realization, but even if that is the case I am still happy to have at least caught up to the rest of you and it only cost me a $10 tube of NT-H1 for the education :thumbsup:

This has always been a question in my mind about how much impact too much paste would be. Thank you for conducting this research. You're doing great work.

 

[Deders]

the MX-4 results are not a fluke (I mounted the MX-4 three separate times to confirm the results).

I tend to find that despite what the manufacturers say, the temps do drop significantly with MX4 after 24 hours of use, how long did you leave it before trying again?

 

[tw33k]

I haven't been here for a couple of days but am now caught up. I agree with hokies83 100%. IDC used far too much Liquid Ultra on the die and his temps are too high as a result. The amount shown in the pic of the die covered in Liquid Ultra would be enough for 3-4 applications covering the IHS.

I urge IDC to do the Liquid Ultra again using only a tiny amount and painting the die

 

[Idontcare]

You used to much Liquid Ultra Bro

use 1/4 of that and try again XD

Half a rice grain. <-- and i think that is to much.

Liquid Ultra is made to Spread does not work as well if u let it spread it's self.

http://youtu.be/3N3D1zaeJoU

Watch the video of how much he used on a Huge Amd IHS.

Before I posted my results with the Liquid Ultra I tried two different mounting approaches. (same as I did with all the other TIMs, some I did even more than two mounts)

What I posted were my "best" results obtained with Liquid Ultra.

Initially I tried the "brush on a thin layer" method, per the directions.

But the results were not so good. So then I added more Liquid Ultra and the results were much improved.

Now I can fully appreciate the tendancy to trend towards the obsessive/compulsive side of things (I have my days too) where following the installation procedure without question and without deviation is the law of the land...but I think you are missing a few obvious common sense aspects to Liquid Ultra.

For starters, its a liquid. Do you know what a liquid does when it is compressed between two solid objects? It provides basically zero resistance to the compression because being a liquid it just squeezes out of the way, leaving the two impinging solid surfaces to come to rest in direct contact wherever and however best suits their local topology.

There is no such thing as having "too much liquid in the way". Think about it.

Now then, why does Coollaboratory recommend its users "paint" on the Liquid Ultra? This is an equally easy to answer question but perhaps only so because I have experience with another liquid-metal TIM: Indigo Xtreme.

The reason Coollaboratory wants you to pain on your Liquid Ultra is because it is electrically conductive and the last thing Coollaboratory needs on its hands is a bunch of DIY enthusiasts frying their mobo's and CPU's because they put way too much Liquid Ultra on the CPU and it runs over the IHS and down into the socket. (a secondary reason is that they get to claim, rightly so, that you have many many more mounts per tube this way as well, makes the TIM seem all the cheaper on a per-mount basis)

Force the user to use so little Liquid Ultra that it can't possibly accumulate enough excess to create a droplet that would run down the side of the IHS and Coollaboratory is safe from developing a bad reputation as a CPU killer

I wasn't worried about that for my situation because (1) my mobo/cpu are horizontal in their current testing setup, not vertical where gravity can cause problems with excess droplets of Liquid Ultra, and (2) I'm running without the IHS so the extra Liquid Ultra has plenty of room to slop down onto the PCB where it is still electrically isolated from the electronics.

Now then, how and why does Indigo Xtreme factor into this discussion? Because for starters Idigo Xtreme is a solid that must be melted after installation, but before it melts the Indigo Xtreme TIM is a whopping 0.67mm thick

Now how is it that Indigo Xtreme can get away with having users put such a thick wedge of TIM under the HSF and yet they deliver bar-none the best cooling performance? Because the TIM turns into a liquid and flows like a liquid with all the excess squeezing out (the same that Liquid Ultra does).

But Indigo Xtreme went one step further than Coollaboratory went in terms of ensuring its users wouldn't be at risk of frying their CPU from the excess electrically conductive TIM - they packaged the TIM in a plastic that melts and solidifies around the excess TIM so that it doesn't run all over the place.

^ that packaging drives up the cost of Indigo Xtreme, by a lot! $10 per application.

But the advantage of that package is that Indigo Xtreme ensures their users aren't using too little of the product (as I found out can be the case with Liquid Ultra and IC Diamond) while at the same time ensures the users aren't at risk of frying their electronics and giving Indigo Xtreme a bad rep.

Coollaboratory didn't go to those lengths, so instead what you see are tedious application guides designed to prevent users from getting too much Liquid Ultra on their gear at the risk of getting too little on the IHS.

For IHS applications too little is probably tough to do. For my application with the bare CPU die getting too little on there was easy to do. There was an optimal amount, how much I don't know exactly, but brushing on a super thin layer was not the ideal amount of liquid ultra as far as operating temperatures were concerned.

I wasn't concerned with going to all the effort of detailing and documenting this before because I didn't think people would actually have issues getting a liquid TIM to mount correctly in the event of using too much. It is a liquid after all, offering the least resistance of all the TIMs tested here when it comes to squeezing itself into the thinnest layer possible.

It makes sense why Coollaboratory wants you to use as little of it as possible, from a liability and economic sense, but it doesn't make sense to argue that using too much of a liquid TIM would inhibit thermal conductivity once the mounting pressure is brought to bear.

 

[Idontcare]

I haven't been here for a couple of days but am now caught up. I agree with hokies83 100%. IDC used far too much Liquid Ultra on the die and his temps are too high as a result. The amount shown in the pic of the die covered in Liquid Ultra would be enough for 3-4 applications covering the IHS.

I urge IDC to do the Liquid Ultra again using only a tiny amount and painting the die

Regarding temperatures - all my reported temperatures are normalized to 25°C. I track the ambient air temperature that is being fed into the H100 radiator in real-time and adjust the recorded CPU temperatures accordingly when compiling the data tables.

I use a room fan to ensure minimal localized heating is occuring in the vicinity of the mobo itself, and I place remote temperature sensor directly in front of the H100 radiator intake so I know how cool the radiator is running.

The remote temperature sensor is displayed by the Temperature monitor, which I setup in a bank of external sensors that included my kill-a-watt, voltmeter, and sound meter.

The bank of external sensors is then positioned to be captured by a webcam in real-time fashion.

The video from the webcam is displayed on the desktop along with the other applications that are of interest (CPUz, task manager, realtemp, Linx, etc), and an automated screenshot utility is running in the background which captures a screenshot every 15 seconds.

I run all my tests for a minimum of 5 cycles of Linx using problem size 43122 (~14.GB) and only report the peak CPU temperature.

Once the run is complete and I see what the peak temperature was, I then go back through all the screenshot images until I find the one screenshot that shows when the CPU got that hot. Then I note the reported ambient temperature from the remote sensor as well as the measured voltage on the voltmeter.

In Excel where I log all the data runs I then record both the ambient temperature and the reported CPU temperature, to then compute the normalized loaded temperature by normalized the ambient to 25°C.

How much of a difference does normalizing for ambient temperatures make? In my case, with the Liquid Ultra, my reported CPU temp (from realtemp) at 4.9GHz was actually a mere 79°C, but once I compensated for the ambient temperature I reported the temperature as 86°C. (which then makes it apples-to-apples comparable to all my other test results, which were also normalized to correct for ambient fluctuations)

I know it is important to rigorously account for test-setting differences in the lab, you don't get out of graduate school without learning this. But I also know that not too many people on the web go to such efforts when reporting how teh awezome their results are. Folks like to get wrapped in being with the "in crowd" and they want to report results that agree with everyone elses reports.

Unfortunately my results don't seem to jive with the popular opinion out there on how great Liquid Ultra is. That could be because I royally screwed up the testing, or it could be because other people aren't going to the lengths I go to in order to ensure the data is actually apples-to-apples comparable rather than just making assumptions that it is.

I understand that people are excited by Liquid Ultra, and I don't want my tests to be perceived as taking away from that. Liquid Ultra is the best TIM I tested so far for bare-die mounting. And even though NT-H1 comes close in my tests, I wouldn't use NT-H1 for this application outside of just testing it because it has the pump-out effect working against it.

Liquid Ultra is also going to pump-out, but being a liquid and having decent cohesion strength it will just pump right back into the space between the CPU and the HSF as the system cools (the pump-out effect is reversible).

 

[Idontcare]

Thanks for the excellent TIM "review" IDC. Looks like I'll have to get some Liquid Ultra/NT-H1. Is it possible to delid a i7 3820? I would love to reduce my temps, from what I read is about 20 degrees lower?

Are you trying to keep your warranty intact for the performance tuning plan?

I only ask because if the answer is "no" and you aren't worried about your warranty but you do want to maximize the lowering of your temperature then I recommend you lap the IHS and then use Indigo Xtreme.

Why didn't you use your calipers? Doesn't it have a pin that extends out as you increase the measurement? Use that part.

I want folks who read this thread to be motivated to think they should try something like this for themselves and at the same time I don't want them thinking they have to be all tooled up with the latest gadgets in order to accomplish the same. My gadgets are just for robustly proving my points, but they were not necessary.

Comparing washer heights for example is easy to do by eye, a toothpick makes it easier. Digital calipers makes the work more precise but does not increase the quality of the results, just the confidence in them.

When folks read my thread they need to have confidence that I did things right, so I use digital calipers, but when they do it for themselves I don't want them thinking it can't be done unless they too have digital calipers and so on.

I used the toothpick just to show that everyday household items are good enough for DIYers to get this stuff done. The calipers are nice but unnecessary. Necessity is the mother of invention

How permanent is your washer setup? Maybe you can glue them together as a stack? Get your desired stack prepped in a clamp (maybe cover clamp in a plastic baggie) and then drip some super glue gel into it.

Actually now that you mention it, what I should have done was run a ring of tape around the outside. Black electrical tape would have held fast and done the job quite nicely. Ah well, maybe next time

Could you have saved time, and/or got more consistent results, using an air-based cooler?

I imagine the water cooler you used may introduce some unpredictability because you don't know what temperature the water is - wouldn't it start off cool, then warm up over time? So if you check the temp of the CPU too early, maybe the water isn't fully warmed up? I guess if you wait long enough you can ensure the entire water loop is fully warmed up and then check the CPU temp, but wouldn't it be very time consuming compared to just using a big air cooler that would reach operating temp much faster? Sorry I'm ignorant about how water cooling affects temp readings over time, but can't help think the water complicates things.

I ran LinX in a loop for a minimum of 5 cycles before I recorded the peak temperature. I do the same whether I'm testing with air-coolers like the NH-D14 or with these toy water-coolers like the H100. It gives consistent results. With LinX I don't see a change in peak temperature between runs 2 and 5 but I run 5 cycles anyways. All the heating up is accomplished in the first cycle.

I can understand some hesitation about this point, but I'm wondering how much of an effect the residual layer after cleaning of either compound affects future TIM application performance. You've been testing hardware that's mirror smooth to begin with, but I haven't gone through the effort of lapping either my i7-3770K IHS or Megahalems cooler. In my efforts to get a sufficent amount of TIM under the heatsink, I had to re-apply IC7 a few times and have suffered from the staining on both surfaces.

That staining is a problem, unfortunately, because it invalidates your warranty even if you purchased the extra performance-tuning warranty.

Any physical alteration to the CPU package - be it from lapping, delidding, or marring of the IHS surface as the stains effect - invalidates the performance warranty

IDC did you do any testing with HT on/off? Was wondering how much HT effects temps on IB's.

I've never messed around with disabling HT on my computers. From what I've seen on the web it results in lower peak temperatures by just a smidgen, maybe a degree or two at most.

Question - how thick is the top of the die? Could you lap it to clean it up?

The die is protected by a thin (microns thick) layer of silicon nitride, which is an extremely durable material. I am sure I could buff the surface with my dremel and a buffing pad.

I may try that actually but I also suspect the topical marring is entirely superficial when it comes to thermal performance.

With that said, I am not opposed to him retesting with your suggestion - I have my doubts that we will see any real difference, but I am always anxious to learn. Also keep in mind that mounting on the die is different than mounting on the IHS. There is a much more surface area on the IHS and because of this, it is actually is easier to get 'too much' as it would require more pressure overall to push it out the sides.

There are other differences too, as Yuriman discovered in delidding his 3770k, the chip-to-chip variation is rather large with these Ivy Bridges and the magnitude of temperature reduction varies by quite a bit from sample to sample.

What doesn't vary is the relative ordering of the improvements, delidding is always better than keeping the original CPU TIM, and removing the IHS entirely is always better than putting it back on. That just leaves us enthusiasts to argue over "how much better"

i'm calling scam on this thread... this was actually just clever marketing for digital calipers

You caught me out Now I regret having quit my day job :hmm:

I tend to find that despite what the manufacturers say, the temps do drop significantly with MX4 after 24 hours of use, how long did you leave it before trying again?

Never for more than a day at most. I did, however, leave the "ridiculous amount NT-H1" testbed to sit for 3 days and retested to see if the result changed...they hadn't. Deadnuts the same. But maybe that is what MX-4 needs? A little time to settle in?

 

[Kenmitch]

I've never messed around with disabling HT on my computers. From what I've seen on the web it results in lower peak temperatures by just a smidgen, maybe a degree or two at most.

I realize one buys the 3770k to take advantage of the HT and it's silly to disable it to achieve a higher overclock....It was more geared towards what a person could expect with a 3570k and to see what kinda temp hit HT makes on IB's

 

[Idontcare]

IDC : slightly off topic but any reason the 3.5ghz voltage(1.118v) is higher than the 3.9ghz (1.106v) and 4.2ghz (1.103v) .
Not that it matters for test as it was used throughout with each paste .

Any reason you didn't do a 4.6ghz?

The specific voltages and clockspeeds I tested at in these tests were selected based purely on the desire to retain test conditions that were apples-to-apples comparable to the test conditions I first started with in testing my 3770k out of the box in its stock condition.

In the beginning I wasn't testing 4.6GHz, I was testing 4.5GHz and 4.7GHz. Only later on did I start testing higher clockspeed as the thermal improvements enabled them, first 4.8GHz and then ultimately 4.9GHz. I did go back and test 4.6GHz for the final shmoo plots, but it never became a standard testing clockspeed for all my other tests.

The 1.118V at 3.5GHz for example is the "stock voltage" for the CPU when the CPU is running fully loaded at its stock clockspeed.

At that clockspeed and voltage, with the stock IHS in place and using the stock HSF with stock TIM, the peak temperature was 91°C.

From there I then went on to test the temperature at those same parameters while changing the TIM, changing the HSF, lapping the IHS, removing the IHS, etc.

So the specific clockspeed and Vcc combos is more of traditional precedent than for any specific reasoning. I held them constant so the comparisons would be apples to apples and nothing more.

 

[Ed1]

Ok, Understand, vcore at 3.5 was set on auto , I get about same thing (1.12v @3.8 on auto ).

 

[hokies83]

IDC how can i have much better temps then u have?

Im running 1.55v Liquid pro on die liquid pro on IHS..

Hottest i can get is 82c on core 4 beside the igpu..

I am using an H-100 With cogars in push/pull

My room is about 80 degrees.

I do know Liquid pro beats Liquid Ultra by 3 - 5c.. but we are more then that apart on temps

 

[supremor]

Reseated my Hyper212 Evo today and changed from the CM TIM that came with it to NT-H1 and temps across all cores are 3-5C lower. Not bad for 30 minutes worth of work (moving a fully loaded HAF-X is a pita).

Keep up the good work IDC! :thumbsup:

 

[Idontcare]

IDC how can i have much better temps then u have?

Im running 1.55v Liquid pro on die liquid pro on IHS..

Hottest i can get is 82c on core 4 beside the igpu..

I am using an H-100 With cogars in push/pull

My room is about 80 degrees.

I do know Liquid pro beats Liquid Ultra by 3 - 5c.. but we are more then that apart on temps

OC'ing is always a YMMV situation. For all I know I gots me a crap chip of a 3770k.

When I read other forums like XS I see folks are really running down the batch numbers and so forth to get good temperatures when OC'ing their 3770k's.

If it isn't clear, and if it needs to be said, I absolutely believe you (and everyone else) when you (and they) claim to be getting the temperatures you/they claim.

I just can't get there. And I've tried. Just nothing I do will break my cpu down to that next tier in temperature that you are getting to.

But you are also running 5.1GHz whereas my chip isn't LinX stable at 5GHz with any voltage. 4.9GHz is the max I can get stable with LinX. I took my 3770k all the way up to 1.5V trying to find 5GHz stability and it was not to be found. It was semi-stable, but not 5-cycles of IBT stable.

So I chalk this up to me getting a lemon of a 3770k.

That said, I think I killed the lemon, and the lemonade tastes funky :\

In an attempt to find the elusive "extra 5°C lower" temperature you (and I) were excited to see with Liquid Ultra, I did another run with a medium amount of Liquid Ultra (the goldilocks test, not too little but not too much)...alas I killed my 3770k

I am pretty sure me killing the chip had absolutely nothing to do with these tests in terms of delidding my CPU and so forth. I live in Allentown PA, roughly on par with NY city temperature-wise, and it has been cold and dry here lately. Cold and dry is a recipe for static electricity, and I've been double-layering my warm clothing lately - naturally all the more static generating.

And so I was busy cleaning my CPU the other day, getting ready for another test, and it didn't boot afterwards. I'm usually pretty good at ensuring I discharge myself on various metal connections around the house because I really do build up a static charge...but I know I did not do that at all the other morning when I was focused on getting that goldilocks test done.

I'm irritated with myself for having ruined my cpu with such an noob mistake, at the same time I had already accepted the likelihood that my 3770k would die at some point (I killed my QX6700 too) so its not entirely unexpected. I just wish I could have finished out my tests to your satisfaction, that is disappointing to both of us I am sure

On the other hand, apparently all I've done is kill a lemon of a 3770k, so in some small way the world is now a better place, right? :whiste:

 

[hokies83]

OC'ing is always a YMMV situation. For all I know I gots me a crap chip of a 3770k.

When I read other forums like XS I see folks are really running down the batch numbers and so forth to get good temperatures when OC'ing their 3770k's.

If it isn't clear, and if it needs to be said, I absolutely believe you (and everyone else) when you (and they) claim to be getting the temperatures you/they claim.

I just can't get there. And I've tried. Just nothing I do will break my cpu down to that next tier in temperature that you are getting to.

But you are also running 5.1GHz whereas my chip isn't LinX stable at 5GHz with any voltage. 4.9GHz is the max I can get stable with LinX. I took my 3770k all the way up to 1.5V trying to find 5GHz stability and it was not to be found. It was semi-stable, but not 5-cycles of IBT stable.

So I chalk this up to me getting a lemon of a 3770k.

That said, I think I killed the lemon, and the lemonade tastes funky :\

In an attempt to find the elusive "extra 5°C lower" temperature you (and I) were excited to see with Liquid Ultra, I did another run with a medium amount of Liquid Ultra (the goldilocks test, not too little but not too much)...alas I killed my 3770k

I am pretty sure me killing the chip had absolutely nothing to do with these tests in terms of delidding my CPU and so forth. I live in Allentown PA, roughly on par with NY city temperature-wise, and it has been cold and dry here lately. Cold and dry is a recipe for static electricity, and I've been double-layering my warm clothing lately - naturally all the more static generating.

And so I was busy cleaning my CPU the other day, getting ready for another test, and it didn't boot afterwards. I'm usually pretty good at ensuring I discharge myself on various metal connections around the house because I really do build up a static charge...but I know I did not do that at all the other morning when I was focused on getting that goldilocks test done.

I'm irritated with myself for having ruined my cpu with such an noob mistake, at the same time I had already accepted the likelihood that my 3770k would die at some point (I killed my QX6700 too) so its not entirely unexpected. I just wish I could have finished out my tests to your satisfaction, that is disappointing to both of us I am sure

On the other hand, apparently all I've done is kill a lemon of a 3770k, so in some small way the world is now a better place, right? :whiste:

Sorry to hear about the dead chip that sucks..

I have a Extra Psu i test fans with on my desk i constantly touch when handling Cpu/gou/mb..

Thats a big thing i think with ppl delidding there Ivy's there not killing them with the razor but with static charges..

We really got some new ppl over on OCN doing it and this is there first builds.. so u have to coach them up a bit.

But on the bright side maybe the next one will be a golden chip?
Guy over on OCN killed his he sent it for a RMA intel sent him back a 3770k made of gold...
5ghz 1.34v stable...

 

[ArchAngel777]

But on the bright side maybe the next one will be a golden chip?
Guy over on OCN killed his he sent it for a RMA intel sent him back a 3770k made of gold...
5ghz 1.34v stable...

So the guy on OCN ruined a chip by delidding and fiddling with it (which voids the warranty) and then sent it in to Intel? Not sure people realize, but this is the type of behavior that drives up costs. Not cool, in my opinion.

 

[Kenmitch]

In an attempt to find the elusive "extra 5°C lower" temperature you (and I) were excited to see with Liquid Ultra, I did another run with a medium amount of Liquid Ultra (the goldilocks test, not too little but not too much)...alas I killed my 3770k

On the other hand, apparently all I've done is kill a lemon of a 3770k, so in some small way the world is now a better place, right? :whiste:

That sucks! Hope your next one is a better chip

 

[Yuriman]

I assume you've tried reseating it multiple times but I have to ask anyway. Last time I put my chip in I got "55" error on my board, reseated it and still got 55, it was only then that I noticed that one of the little paper washers I have in each corner between the springs and the block were missing (probably 1/4 as thick as one of your washers), and that slight unevenness in pressure was causing the chip not to make good contact with the socket.

Replaced the washer and all is well in the world.

 

[Idontcare]

Sorry to hear about the dead chip that sucks..

I have a Extra Psu i test fans with on my desk i constantly touch when handling Cpu/gou/mb..

Thats a big thing i think with ppl delidding there Ivy's there not killing them with the razor but with static charges..

We really got some new ppl over on OCN doing it and this is there first builds.. so u have to coach them up a bit.

But on the bright side maybe the next one will be a golden chip?
Guy over on OCN killed his he sent it for a RMA intel sent him back a 3770k made of gold...
5ghz 1.34v stable...

I agree with you. It stands to reason that folks are zapping their chips while handling them to get the lid off. The chip images I've seen from folks killing their chips don't usually show the kind of physical damage I would expect to kill a chip. I hadn't put it together until you mentioned it, but it makes sense now in retrospect.

These 22nm finfets could well be sensitive to ESD in a way that 32nm wasn't.

And 5GHz at 1.34V? I wonder which 22nm fab that one came out of. Sounds delicious Early adopters always run the risk of getting the lemons, oh well.

That sucks! Hope your next one is a better chip

I may end up going with an AMD 8350 for my next chip. The $200 pricetage is right but I need the mobo price to make sense too.

I assume you've tried reseating it multiple times but I have to ask anyway. Last time I put my chip in I got "55" error on my board, reseated it and still got 55, it was only then that I noticed that one of the little paper washers I have in each corner between the springs and the block were missing (probably 1/4 as thick as one of your washers), and that slight unevenness in pressure was causing the chip not to make good contact with the socket.

Replaced the washer and all is well in the world.

I've tried ~2/3 of my list of "try this and cross your fingers" options, so far nothing has worked.

I'm not dead certain the cpu is dead, just 66% sure. In another week I will have iterated through my remaining test matrix of options. But it certainly is NOT looking good at this point. Either the CPU is dead or the mobo is dead, the remaining 1/3 of the tests will isolate which it is for certain.

At this point though I can definitely rule out all the obvious stuff like extra washers or debris of some sort

I've got another motherboard to try it in (Asrock xtreme 4+ or 6+, I forget which) and a 2600k to test in the mobo's themselves as well.

Oh well, this is the risk we take with our hobby toys. Would have been nice to have tried a few more attempts with the Liquid Ultra, and I was holding out one packet of Indigo xtreme to fuse the IHS back onto the CPU to make it like a 2600k...but those tests will probably not happen now as I can't imagine buying a replacement with Haswell around the corner.

 

[GPz1100]

Sorry about your possible loss.. Microcenter still has the 3770K's on sale through sunday, $230+tax.

 

[Revolution 11]

Any chance you can cover these tests you are performing to determine which component has failed?

 

[Idontcare]

Sorry about your possible loss.. Microcenter still has the 3770K's on sale through sunday, $230+tax.

:hmm: There is a microcenter about an hour away in Philli...have to see if the wife is interested in taking a day trip down there.

edit: bummer, I just check the Philli store and they don't have any in stock.

Any chance you can cover these tests you are performing to determine which component has failed?

Sure.

So far I have:

1) remounted the HSF, confirmed CPU was correctly seated in socket

2) repeated step (1) three or four more times

3) Cried, drank beer, cried some more :'(

4) Removed the entire mounting bracket apparatus, spent time counting washers and measuring thickness to confirm I didn't lose a washer somehow along the way (confirmed all the washers are there, cannot confirm I didn't lose a marble or two in the process )

5) Cleared the CMOS, backed off on the H100 mounting bolts (loosened the H100 basically) and kept checking the reboot as the pressure got less and less

6) Got wicked pissed, cleaned up wicked piss, wickedly, vowed to never do that again unless I'm in the shower :whiste:

Remaining tests to perform:

1) Put IHS back on 3770k, attempt to boot to see if socket connections are the problem (the IHS pulls double duty as more uniformly apply the down-pressure on the PCB to properly get the LGA matched up to the pins in the socket)

2) Put 2600k into the MIVE-Z to confirm the MIVE-Z boots just fine.

3) Put 2600k into the Asrock Xtreme <insert random numer>+ to confirm the Asrock boots just fine

4) Put 3770k into the Asrock and see if it boots.

...right now I am fairly convinced it is the CPU that died, this is how my QX6700 died too. Unfortunately there is no video of the "did it dieded?" to upload to liveleak :|

 

[GPz1100]

^^Maybe they have the 3570K instock? That was just under $200 for me the other day at the chicago store. No OC/delidding on that one though. Going into someone else's machine which needs to have a valid warranty.

In fact, the combo deal was pretty good. 3570K + p8z77-v lk for ~$280 after rebates.

 

[BonzaiDuck]

So -- IDontCare -- I've caught up scanning through this latest thread.

And after all it has been through since mid-summer (? or earlier?) -- you finally killed your CPU.

I took notice where you said a page or two earlier in this thread that you'd pick the ICD for your final choice of a 24/7/365 TIM. Is that still your final choice?

The better results with Liquid Ultra versus ICD are about 2+C better than I would've imagined, but the data basis for my best guess were based on tests including the IHS.

Basically, it seems that you've proven a set of guidelines for an Ivy Bridge mod for use specifically with a water-block, or more specifically with the H100 cooler. I think some others here have used their heatpipe coolers in configurations that could vary from yours. If I were to do that, I think I'd avoid mounting fans on the cooler and avoid the extra torque. But the washer-standoff approach would seem to provide a stable approach, anyway.

Having read (or scanned) through all of this and the earlier thread, it has been an incredibly revealing and disciplined project.

 

[SKORPI0]

Thanks for the review and sorry for your 3770K loss.

If you/friend works in specific retail outlets, you can get on this deal.

*RETAIL ONLY* Intel retail edge holiday deal 2012 3770k for $105 or less! and more! *LIVE*

Intel 3rd generation 3770k processor

3770k pricing:
Member: $105
Rockstar: $95
Rock Legend: $85

3770k AND Intel® Desktop Board DZ77GAL-70K
Extreme Series Bundle Pricing

Member: $265
Rockstar: $255
Rock Legend: $245

There is also a Intel 520 series 180gb for a decent price

All members: $145

They also charge $15.50 shipping for those in the US, $38 for Canada.