[Build Log] The Green machine (lots of pics, have fun 56K)

[Rayce185]

This is a system build log that i released on another forum and have gradually updated over the last few days. It is extremely detailed with lots of explanation and pictures, so bring time with you if you're interested



(July 10th)



Alright, so I finally got all the important parts for my new system delivered.

The specs are:
-XFX 780i SLI
-Intel Q9550 E0
-4GB OCZ Blade Low Voltage DDR2-9600 (5-5-5-18@1.80V)
-2x XFX GTX 275 XXX (670/1404/1180)
-Corsair HX850W (single 12V rail @ 70A/840W, lowest efficiency: 89%)
-Auzentech X-Fi Forte 7.1 (PCIe)
-2x Super Talent UltraDrive ME 64GB in RAID0
-LG GGC-H20LRB BD-ROM
-TRUE + 2x Ultra Kaze 3000

No case yet. We'll get to that later.



Alright, so I prepped everything: got my tools, the item checklist, the procedure list and laid out the hardware for visual inspection. The last thing I wanted to do is waste hours on a built to find out that there was a noticeable flaw in the hardware. You may have heard of stories about bent CPU socket pins...

Tools of the trade. Needed or not, it's good to be prepared.



The lists. Ordered part list (sorted by retailer) on the left, "operating procedure" on the right (split into hard- and software)



Probably the best DDR2 RAM you can buy.



Ultimate droolage. "I shall get thou past 4GHz!"




nVidia reference board. Good enough.



Woo! No bent pins for me!!!



I absolutely LOVE the PSU's bag! I shall find good usage for it.




The actual muscle. I hope the stock cooling is worth a damn.




The drives. Those SSD's look sexy as hell.





"The Rock"



Yeah, the TRUE is infamous for having a rough contact surface. This will definitely have to be lapped as soon as possible.



And of course the contact isn't ideal either. Not the best circumstances for a to-be OC system, but next to lapping the TRUE, I have a special treat for the CPU's IHS. More in stage 2.



Thou shall play along well with my Sennheiser HD600



What can't really be referred to as a "case". You shall see.



After all was checked and looked okay, I prepared the case (A+ GL-3) so I could install the mainboard by screwing in the supports. This is where the fun begins.



The case only came with SEVEN mobo supports! Awesome! How the hell am I to install anything if the manufacturer doesn't even fully equip their products?

But fear not: Ray was smart enough to save all the unused and reused screws over the years, and may have some identical supports in the box...



I was wrong. Not one single support had the same features as the ones form A+. Either the head was too long, the thread was too long, or two small, or all of the above.
Well, I decided to take two of the ones that came closest: the ones with the long heads, and whipped out the trusty Dremel.



So far, so good. They aren't the exact length, maybe half of a millimeter off, but that will do.

Then let's focus on the mainboard.
First, the stock heatsink solution must be removed, and some proper TIM (I use MX-2) has to be applied to all parts. So I get the isopropyl alcohol, a crapload of q-tips and a few leaves of toilet paper and start cleaning the old, useless mush off.




Fugly residue from mushy mushmush. It gets stuck on everything.



Alcohol (and q-tips) FTW!





Then apply the MX-2. I didn't have a small hard haired brush at hand, so I took one piece the loose packing styrofoam that came with the hardware, squished a tip onto it and used it so evenly apply a thin layer onto the mobo. This works way better than an old credit card (or similar) since those are too large and not flexible enough, especially when moving around the MOSFET's.







And since manufactueres mess up the TIM application everytime, let's redo the GFX cards too.
I never knew that GPU's have bcome so huge! Here's a picture comparison between an 8600GTS and a 62mm TEC unit.




As you can see in the above picture, there is lots of that ugly mushy TIM (or similar) on the heatsink, intended to cover the VRAM, shading unit and (probably) the voltage regulator. Unfortunately you can't remove most of them, since then there wouldn't be anymore contact between the chip and the heatsink. So I just did the GPU itself and the MOSFETs.




And of course, I repeated the whole procedure on the other GFX card too.


The rest is pretty straightforward: Install 60mm fan on MCP heatsink, install CPU, add TIM, install TRUE.... WAIIIIT a minute!
The TRUE is so large, that the 60mm fan (mounted slanted on the mobo heatsink) cannot fit. Great.
I guess I'll just use the standard fan adapter form the XFX board until my stage two cooling mod is ready to install.



Okay, next up is installing the PSU into the case. But something seemed fishy...



Great. There isn't nearly enough space for the GFX cards. This case is a nightmare. Thank god I didn't pay any money for it!
Ah well. Until I decide what case I' want, I'll just leave the system naked.




I installed a potentiometer for the two UK3000's on the TRUE, since they don't necessarily have to run at full speed (and volume) the whole time.



Unfortunately, I will later find out that this one tiny potentiometer cannot handle the power that is needed for two UK3000's, so I had to split up the line and use two of them instead.



As you may or may not see, the top GFX card is touching the sound card, which is never good. Just to be on the safe side, I did this rubberband mod to keep the cards straight:



So. Everything is mounted together (more or less neatly) and ready to be fired up. First and most importantly: update the BIOS with the bootable disc I burnt earlier today. The ISO was ready-to-use on the XFX website, nice.
Afterwards I used my bootable USB stick to let memtest run, as an initial stresstest for the RAM.



Well, I just let it run while I'm writing this, so this is how far I've gotten. I'll install Windows 7 x64 (build 7201) shortly and run more benches to see what the stock system is capable of. I'll keep you guys posted.

[Rayce185]

(July 11th)



Aight, here we go.

After installing Win7x64b7201 on the array, which was strip sized to 32K, I reconfigured the array to a 128K size and reinstalled everything again.





As stated previously, I reached a boot time of 46secs.
Well, I don't know if it was the reconfiguration or me defragging the small 100mb partition that W7 created (or both), but I got an even better time: 38secs! HURRAY!
(EDIT: I now do times around 28secs...)





So I went ahead and did some minor benches to get warmed up...





Well, I wasn't happy with the GPU idling at around 50°C, so I put my creativity to use again.





Soooo... I added two UK3000's for tunneling and four 80mm Papsts.
Call me crazy (which I most probably am), but it helped!




I then ran PCMark Vantage, and only got a "meh" 9800 points. Definitely worthy of an improvement.
Last but not least I wanted to stress the CPU for a bit, nothing big, so I ran LinX for ~17min (3GB, five runs). I am happy with the cooling performance: It never exceeded 47°C. (I forgot to close the Rivatuner Monitor after PCMark, for those who wonder)




Next up is a wee bit of OC'ing. But I'll have to see how the BIOS ticks first. I had it set to an unlinked 1333MHz/667MHz, which effectively gave me a 1:1 ratio in CPU-Z (as you saw above): effective 333MHz FSB to an effective 333MHz RAM speed. The latter fuggers are specced for 600MHz and have been reported to go over 650, so I'll have lots of fun doing this.




To be continued...





(Later that night)




Since it's a relatively cool night outside (it's 4:30am BTW), I opened two windows to let some cold air flow through here. I put on three pairs of socks, two jogging pants, two long sleeve shirts and a light jacket.... but the results are satisfying:






Let's start OC'ing!

[Rayce185]

(July 12th)




Alright guys, I'm at it again!

I am not as experienced with voltages, so I used the 780i guide's presets for everything except the CPU Vcore and the Vram. I also set the FSB:RAM ratio to synchronous link, put the FSB to 1600MHz and the multiplier to eight. I put the Vcore to 1.3250V and kept the RAM at 1.80, since it's specced to run at 1200MHz instead of the configured 400. The setup ran flawlessly perfect and Linx did 5 runs @ 1792mb without any errors. I didn't bother make a screenie since 3.2GHz is kind of "boring", so I proceeded to a 425MHz FSB and kept the rest untouched. The result is more than satisfying:


Stock specs with sync. link:




8x425 @ 1.3250Vcore (sync link):



I am pleased, a lot. Especialy with the temperatures. But with a crazy open cooling like what I have, I didn't expect any different.
Speaking of which: I "optimized" the setup a bit:



As you see I added a teeny weeny 60mm Mini Kaze under the the TRUE. Even if it doesn't help the CPU cooling, it will definitely help the chipset cooling by blowing fresh air into that direction, instead of it just using the air surrounding the CPU.




I also added a Lian-Li fan triple monitor with three temperature sensors, which I used two of for the RAM bricks and one for the CPU. Unfortunately I must've broken the CPU one, as I don't get any readings from it.





And a quick "peaking" shot from the sexy RAM




Well, the RAM temps never ereached 30°C, which I expected since they aren't OC'ed at all or anything.

As I'm writing this, my 3.6GHz stress test completed. Here's the result:



I R happeh!!! I'll probably lap the TRUE next, make a few temp readings after 20 Linx runs and then install Stage 2. Here's a sneak peak of what you will be awaiting:




And when it's all set, I'm going to show you how my original rig, Old Faithful, will get a new revision:






Stay tuned!





(later that day)





Here's my newest mishap =/



Vcore dropped to 1.272V under load.






(continued)




And I just redid the same test as above, but with a BIOS Vcore of 1.3850... and an idle CPU-Z Vcore of 1.352, under load it was 1.288:





And this is my current "setup", *ROFL*












(later that day)




I'm struggling with the TRUE lapping.

I got this "lapping kit" on ebay since it has basically everything in it that one would need: Different grit sandpaper, three grit sizes of lapping compound and a buffing cloth (small piece of micro fibre =/ ).

Well the roughest sand paper is 220 grit and it's worn off already as it is maybe a quarter of the size than the paper you'd normally get in a hardware store, and stores are closed on Sundays over here...

For reference again: The crappy OEM surface.




My sanding panel. Neoprene ("anti slip mat"), glass, sand paper.




30min passed. Either the paper isn't rough enough, or the nickel coating is really hard. But the progress is SLOW. Also, signs of the TRUE's unevenness are showing




60min passed. Getting there, but still very slowly. Signs of the sand paper's wear are showing.




70-90min passed (can't remember the exact length as I was still quite drunk from partying shortly before). This is how far I got before falling asleep on the floor. The paper is hardly sanding anymore.
I am in distress.




Session two, four hours later (Don't ask me why I woke up so early). About 20min of sanding, the difference little to nothing.
I am not happy. Sand paper feels as rough as newspaper and the friends are still as sleep (want to ask if they may have any paper).




As you can see, the TRUE's base is incredibly uneven. I still can't figure out why Thermalright didn't give it a proper treatment, as it isn't really considered a cheap cooler. Even my 25€ Scythe backup cooler came lapped, what gives?





(continued)




So I also have a new toy:




The two sensors have a range of -50°C to +200°C, should be enough for my purposes

I had the fancy idea of shoving one of the sensors under an edge of the CPU's IHS. It was good in theory...






I also shoved one under the small heatsink of the top set of MOSFETs.





As said, the theory was good, but the practise wasn't so much:





Under load it never exceeds 32°C, whereas Realtemp is showing up to 55°C. Ah well, at least the MOSFET sensor should be fairly acurate, since it's directly ontop of one.





(more on that day)





Okay... so can someone help me out? I'm kinda stuck when going at higher FSB speeds.

475x7.4... 3.56GHz
Vcore idles at 1.376 and drops to 1.304 under load
Temps stay in the mid 50's
Linx crashes within 35s on a 15x1024mb run.

Other voltages:
FSB 1.4V
RAM 1.85V
SPP 1.45V
MCP 1.575V
HT SPP-MCP 1.35V


It runs flawlessly at 450x8 = 3.6GHz, but as soon as I up the FSB, things go to hell. at 500MHz I can't even boot the BIOS.




Running on 3.825GHz now (450x8.5) with RAM set to 4-4-3-15 CR1 on a 1:1 ratio. Vcore drops to 1.304 under load, which is acceptable. CPU temps never exceed 59°C, which is good for a half lapped TRUE *LMAO*


I'm beginning to wonder: If I am able to get the VTT issue resolved with XFX and I can set it to 1.5V, how far will I be able to push the FSB then? Because with the fully prepped cooling setup, this will be my next limit.






This is the highest I can go at this time, until the issue with the VTT has been resolved. I will do an overnight run tonight.

[Rayce185]

(Last night)





Call me clinically insane, but I have to pat myself on the back. I didn't know my balls were that big and hairy that I would have been able to pull off what you are about to read. But let's start from where we left off, shall we?





As you may or may not know, there is a risky little procedure called "IHS removal" that basically leaves you with a buck nekkid CPU. Most of the current Intel processors are soldered to their IHS, making the removal not just very difficult, but insanely dangerous as you risk tearing off the die from the PCB if you are not careful enough, or simply clumsy. The benefits are obvious: The heatsink has direct contact to the die, giving you better cooling and the possibility to OC even higher.

Well, Teh RheY decided that this was a good way to fugger around with his brand new 200€ CPU.

The procedure is pretty basic: Cut the glue between the processor's PCB and the IHS, then desolder it with heat. Sounds pretty easy, huh? Well if jamming razor blades into your CPU and lighting it with fire is your thing, you should give it a try!


I will write this as a quasi-guide for anyone who is crazy enough to do this too. There are various ways to do the cutting, just as there are different ways to desolder the IHS itself. This is the way I felt was best for my needs.



DISCLAIMER:
What you are about to encounter may be considered to be painful to your computer. Be sure to turn your screen away from it so it must not endure watching this butchery.

PROPER DISCLAIMER:
If you try this at home, I am not to be held responsible for any damage you make. If you fupped uck, it's your own damn fault for listening to a crazy German on an overclocking forum. If I jump from a bridge and survive, this may not necessarily apply to you either.




So here's my run on it:

Unfortunately I forgot to make pictures of how I started cutting the PCB glue, I guess I was too excited. Anyway, you take a prepped razor blade (see pics below) and start in one corner, slowly cutting back and forth until you are approximately behind the little IHS step, or rim. MAKE SURE that you try and keep the blade pointed slightly upwards, or at least not downwards. The PCB is EXTREMELY sensitive and even the slightest cut into it may fataly damage your CPU.
You have then successfully made your first cut through the PCB's rubber, yay! You have also just voided your warranty and are eligible to any fudd-up you do from here onwards. But let's continue...

With the blade still in place, slowly cut your way jigsaw style to the next corner. be sure not to cut too deep, try and stay maybe 1-2mm behind the rim at the deepest point. If you are unsure how far that is, take your razor blade, place it ontop of the IHS and align it parallel to the rim. Then take a waterproof pen and mark that line on the blade. When cutting, try not to exceed said line by too much, but be sure to cut the glue all the way through. You will notice this when the blade can go along the rim easily with not a lot of pressure.

Anyway, you're slowly moving around to all four corners of the IHS. After that, prep three more blades the same way and position them like this:




Yeah: Add a thick layer of TIM on the CPU. You'll see why in a bit.
Take a large piece of duct tape and stick the blades together, like raising walls on a house building site. This little construction should look like a cube with one side open:



This will add pressure to the IHS and will make it pop off faster when the solder is liquid. Note that the solder must be liquid ALL THE WAY or you may risk tearing off the top of your die by this procedure. You can avoid this by adding a high temperature for a short while, instead of slowly waiting for the solder to liquify with low temperatures.

Next you will need a flat pot, or a pan.
I will first show you how not to do it, and how I nearly tore off the die like I said before.

I read in another guide that the solder liquifies at 80-90°C. So I set the heat on the stove to a low setting and measured the temperature:





DO NOT DO IT THIS WAY WHEN USING THE PRESSURE PROCEDURE LIKE I DID!!!
I had the CPU in for well over two minutes and nothing happened, luckily. As said if the solder melts too slowly, it will be gooey and you risk tearing the die. I cannot repeat this often enough.


How to do it right: MAX THE STOVE TEMP, WAIT A FEW MINUTES AND PUT IN THE CPU.
Before putting it in, I measured the temp again, but that small sensor can only go up to 200°C, everything above will be inaccurate. But I measured 240°C, but it could be well over 300.

THE IHS WILL SEPARATE VERY FAST!!! It popped off in about 2-3 seconds after laying in the CPU, so be prepared to burn your fingers by QUICKLY getting your CPU out of the heat. Your kitchen will smell a bit of burnt TIM. The TIM was to get the heat faster and more even onto the CPU. I'm sure it will work without it, but I wanted to be more safe than sorry. I'm sure you would be too.

This should be laying in the pot, cooling off:




And this should be your CPU:




Note: I did this yesterday at around 4am. I had no idea if the thing even worked anymore until I threw on the system about an hour ago :lachen:
But let's stay optimistic and continue.

You will want to have the dies as smooth as possible, so start scraping off the remaining solder CARYFULLY with a razor blade.




I bent the blade slightly to have a more accurate cutting point, so you can be more precise and go less risk of doing any possible damage.




When done, this is how it should look like:




Now I needed to relax my nerves (I was still shaking from the instant pop when the IHS flew off), so I left everything as it is and continued with other parts of the build in the meantime.

Today I finally bought a sheet of 180 grit sanding paper, so I could finally continue my TRUE lapping.
Oh yeah: Remember that I had the brilliant idea of using a sheet of neoprene as an anti-slip mat under the glass? Well it turned out to be a DUMB idea:



I accidentally stepped on the glass when walking around, well I was jumping around dodging all the parts and junk lying in my apartment, and the softness of the neoprene gave in and popped the glass. I am sure this wouldn't have happened if I had used the glass flat on the ground.
Dumb idea, part two:



Duct tape loves neoprene. A lot.



I was able to remove most of the duct tape, luckily. So it's still usable to about 80%.




Let me describe to you how you properly sand a heatsink. I haven't read one proper guide on any forum yet, and this bothers me.

You put the sanding surface on the ground, not a table. Reason is that you can then kneel on the floor (take a pillow), with the workplace between your legs, so you can add an even pressure of your upper body onto the heatsink. If you just hold it infront of you, you will not apply even pressure. The further you move away from the part, the less pressure you are applying on the side farthest away from you. Simple logic, isn't it?

Here is how you hold the TRUE:



Use both hands and do a circular motion. Do not do linear moves, no crosses or any other fancy move. You may look cooler doing it linear, but your sanding 50€ worth of copper, not scratching a turntable.
Here is why: The same logic applies with the linear movement as with the body position: When pushing the heatsink away from yourself, you apply uneven pressure to it, resulting with a crooked surface in the long run. And we're trying to get away from the unevenness, aren't we?
The right way: Do a small circular movement WITHOUT APPLYING EXTRA PRESSURE! The heatsink's own weight will do the trick. Pushing down on it will result with more pressure being applied on some points than the other. It may be a solid piece of metal, but you aren't bending it, you are just moving it around with uneven applied pressure.

Count your circles. I'm not joking. Do fifty circular motions, then turn the heatsink by 180° and rotate it in the same direction for another fifty moves. You do this because even if you aren't applying pressure, the circular movement has a momentum that gives the surface a SLIGHT uneven pressure balance. You balance this out by rotating it and doing the same amount of circles in with the heatsink back to front.
If you are too lazy to count, then you can use a stop watch. But as you are never doing the same speed, counting is the more acurate way of doing this.

Wash the sand paper regularly, or at least wipe it off to get the copper out of it. Not only will the sanding performance improve, but you will also have less metal scratching itself into the heatsink's surface.


Now that you know how it goes, here are my procedural results:





And the final result, after applying the finest 1200 grit paper and the three different lapping polishing compounds that were supplied:




I am happy.




Now, let's get the CPU all shiny too, shall we?

Remember you are sanding off solder from silicone, so be VERY CAREFUL! Wash ANY metal off of the paper, keep the paper as flat as possible at ANY TIME (better yet: glue the paper to the glass so it doesn't bulge when getting soaked), do small circles and move SLOWLY. I only counted to a maximum of twenty in each direction so I could check the surface as often as possible. Since the solder isn't 100% flat before sanding and you don't want to risk damaging the silicone, rotate the CPU by 90% if the solder get's removed unevenly.
Like this: 20 circles, 180°, 20 circles. Check surface. Rotate so that the corner with the most solder is at the front of the circle movement, the repeat.

In the end, it should look like this:




And after applying the polishing compounds, like this:




I have a lapped CPU die. How many people do you know who can say that?



Okay. So now that everything is ready, it's time to install everything.
Since the CPU is significantly lower now, you cannot use the standard CPU retention module from the socket, as the heatsink will not touch the die. So we remove it:




Next we add some padding so that you minimize the risk of the heatsink landing crooked on the die. The last thing you want is to tighten the screws and the die cracks because the heatsink was ontop with an angle. This is where the 3mm thick neoprene comes in handy. Lay it underneath the CPU and use a scissors to cut it slightly smaller than the CPU:



Then use a marker and a knife and make an aligned hole for the die to touch the heatsink:




Perfect fit.




Take it back off, apply some TIM and put it on the socket.




So not only is the CPU flatter, but the heatsink too. So the TRUE's mounting brackets aren't effective anymore. It is recommended to add a washer or something anyway, even if you don't lap the TRUE or remove the IHS.

I did multiple steps to ensure maximum pressure.





Yes, those are two metal washer that I stuck together with superglue. :lachen:


Carefully put the TRUE in place. Don'T worry about smudging the TIM, it isn't touching. You can even pick it up again and look.

Remember the piece of neoprene you cut out of the spacer? Use this now:





It will work as an anti-slip mat when putting on the washers and tightening the screws.

I also added washer to the screws to add even more pressure.




Then tighten the whole thing.




Next, you take it back off.

Yes, you heard right. I'm taking it off again. You have to know if the pressure is enough for the TRUE to squash the neoprene spacer and apply enough pressure on the die to be effective:



See? It wasn't. Now what do you think would have happened if you turned on the PC? Fried silicone, anyone? ALWAYS check.

I added some plastic washer to the screws, as I ran out of metal ones. This time two per screw:




I didn't make a picture of the removed TRUE again, sorry. But it worked flawlessly.


Now lets' put this monster into a case!



No, I didn't buy a new case yet. I took out the old PC and put it into the small HTPC case and am now using the stripped out Chieftech tower :lachen:

I'm using black thumbscrews to tighten the mainboard, as I will be taking it out fairly often in the next few days.



Next to the practicality, they look sexy as hell!

I bought 3.5" frames for the SSD's:



You can mount two 2.5" drives into one frame, and also add an 80mm fan at the bottom. But since I'm doing a one-per-frame setup, the fans won't be needed.


Say hello to the TRUE triple UK3000 setup!




Done.











And now, the moment of truth...











It's alive!



I am happy.





I am, however, not satisfied with the temperatures it produces.



The whole point was to lower temperatures. And they are supposed to be easily 5°C lower with a lapped TRUE and an IHS-less lapped CPU.
I have concluded three possibilities for this:

a) The pressure isn't high enough
b) The neoprene is warming up
c) The TIM needs to cure.

You can decide what the reason may be. I have ordered a proper backplate for the pressurized retention mod found in the air cooling subforum.

I call this stage 1.5, as this is not what I was talking about when referring to stage 2 earlier in this build log. Stage 2 will be following this week, hopefully, and you will be surprised about it at least as much as with my suicide IHS mission.




Anyway. I shall retire for now. It is almost 10am, I have slept less than five hours in the last three nights combined and I have classes to attend to in two hours. I wish you all good riddance. And I hope you enjoyed the read so far.


This isn't done yet. Not by far.






(Later, this morning)





Ghetto cooling mods FTW!







Plus I had to move the case. The sun was shining in and the floor next to the PC was 35°C hot!!!








This concludes my current progress on my system build. As soon as I receive the new backplate tomorrow, I can begin with my stage two cooling setup. And as soon as my VTT issue has been resolved with XFX (or I buy the EVGA 780i FTW) I will be able to go past the 4GHz limit.

[Rayce185]

Sorry for the larger screenshots. I'll thumbnail them now. I just wanted to get everything posted first.

EDIT: All huge screenshots t-nailed.


Thanks for reading!

[perkam]

What was the temperature decrease after lapping TRUE + IHS off?

Also. Dude. One word: WATER.

Water's not for me because I'm not that nitpicky about cooling, but holy heck 0_o You need it desperately.

Perkam

[Rayce185]

As you can see in the screens, there was even a temperature increase

It's all described in detail.

[ToTTenTranz]

So.. you call your system with two GTX275 and an overclocked quad-core "Green machine"?


I wonder what you'd call a "Red Machine".

[TheCarLessDriven]

WOW, talk about thorough!

One thing though, you are putting on A LOT of thermal paste on the CPU. This is probably why you are not getting good temps.

Oh and why last gen Mobo, CPU & Ram?

[Rayce185]

So.. you call your system with two GTX275 and an overclocked quad-core "Green machine"?


I wonder what you'd call a "Red Machine".

ATi doesn't make Mobo chipsets :|

WOW, talk about thorough!

One thing though, you are putting on A LOT of thermal paste on the CPU. This is probably why you are not getting good temps.

Oh and why last gen Mobo, CPU & Ram?

Really? On some areas you can slightly see through to the die... so I should apply even less?

I'm still dumbfounded on why the temps have worsened, but I was told that the neoprene has no influence on this. And the system has been running all day stress testing, so it can't really be the TIM curing anymore either.

So it's down to either not enough pressure, too much/little TIM, or both.




Thanks for the comments guys! Keep em coming!

[TheCarLessDriven]

ATi doesn't make Mobo chipsets :|

Really? On some areas you can slightly see through to the die... so I should apply even less?

I'm still dumbfounded on why the temps have worsened, but I was told that the neoprene has no influence on this. And the system has been running all day stress testing, so it can't really be the TIM curing anymore either.

So it's down to either not enough pressure, too much/little TIM, or both.




Thanks for the comments guys! Keep em coming!

Thermal paste should be spread out very thinly and evenly starting with a uncooked rice grain size. Its to fill extremely small gaps between the CPU top and the base of the cooler for better contact, not large physical gaps like you showed in the picture with the CPU laying on the cooler.

[Rayce185]

Small update:

As you saw before, I was having temperature issues after the IHS removal, mainly caused by too much TIM and an uneven alignment.

Well I took it all apart again (god that takes long with a case) and re-polished the HS and the die:





I then applied about half as much (or less) TIM as previously.




You still need to spread it out though, as there is no heat spreader on anymore and the entire die surface gets hot.




I then added even more washers to the TRUE.



GIVE ME PRESSURE!!!



Well the results are better, but not outstanding. Load temps:




As you can see, the contact on the first two cores is better than on the second set. I am sure that if I get them both to equal heights, the temperatures of both cores should decrease again, since the contact would then be complete.

Does anyone know which die (when looking at it) is the second one?

[ToTTenTranz]

ATi doesn't make Mobo chipsets :|

Erm.. it was a pun.. about power consumption.

Get it? Green machine, low power?


Anyways, nice rig.

[Rayce185]

Erm.. it was a pun.. about power consumption.

Get it? Green machine, low power?


Anyways, nice rig.

Ohhhh! Okay, gotcha

[Quad-Damage]

Nice read.

I agree with the water.

I have a E8500 @ 4.2ghz on water, never seen over 52 c full load.

[Rayce185]

Ah well. I haven't completed Stage 2, and I wanted to keep it a "secret" until I was ready to bench it, but it may take a while for me to use it (you'll see why) and I want to split up my progress into smaller parts. Having gazillion word long logs is not only a strain for me to write, but for you to read. ("Oh no, not another 20x PgDown post! I'll just skip it and look at the pictures.")



So I didn't want to stick with an air-only cooling, but want to make a hybrid.
Some of you may have read about Muffy's custom TEC plates, so I bit the bullet and bought a set off of him, along with an order for the ebay 62mm 545W peltier unit and a Meanwell S-350-12 (12V 350W, duh).

Let's concentrate on the Meanwell first.
To make sure that I get proper values, I first measured all the voltages: VIn, VOut(Min) and VOut(Max).





Then I market the sides of the Peltier. You don't want the hot side facing the CPU, do you?




I needed to extend the TEC's cables...




And harden the ends to the PSU with some solder.



Next up, one of my new favorite hobbies (*chuckle*)


Polishing!


I got sick and tired of doing everything by hand, so I got a polishing kit for my Dremel 300. After trying it out on the Q9550's IHS I did a test run on the TEC:



Not bad, ey?

Let's do a full side comparison:





Both sides done:






Now let's check out the TEC plates.

Okay, Muffy, this is no disrespect to you, I love your plates to death and you did a GREAT job finishing them for a ready-to-install usage, but I'm sort of a perfectionist. And I love Dremel Polishing :lachen:

Here's how the plates looked after having them lie around for about a week:




And here's my polishing job:




God I love doing this





So yeah. I did all four contact sides.
Then it's time to TIM the TEC (sounds funneh)...





...position it...




... and sandwich it!




I pulled the screws REALLY tight, as I read that they work most effective when under high pressure.
I just hope it's still okay, since I heard a few minor cracks when tightening...




This beast weighs as much as the TRUE! So with two 120x38 fans I should have well over two kilograms strapped onto the CPU! Poor little bastard...



Next up: Sealing it!
i found this small cartridge of silicone in a shop. Practical, I don't need that huge one with the metal frame.




Then I applied it...




...and removed the excessive silicone with a piece of cardboard.




That's pretty much how far I've gotten! Now I'll let the silicone harden out, which is said to take about a day at that thickness. Does anyone know how to accelerate this?





Next to modding my PC to hell, I have found another device to toy around with. This will be done over the weekend:



-Adding a 110x60x30 alu box for the switches ontop
-Adding three switches with green LED's for the VOut ports on top(12V/30A switches)
-Adding an input power switch to the side (250V/6A switch)
-Mounting a 0-20Voltmeter to said box

-I already replaced the stock fan with a Scythe Mini Kaze 60mm




If it all goes well, I may have the TEC setup installed by tomorrow evening

[jpennstar]

Just WOW.

That looks like ALOT fun too.

Thanks for taking the time to show us how to do some xtreme mods.

[mcmeat51]

Duct tape loves neoprene. A lot.

Haha that made me laugh.

Well done for a very informative thread to ppl whi have not yet tried IHS removal and TEC stuff!!!

Im interested in the TEC mostly though IHS removal seems like a hell of a lot of work for minimal gain.

Im glad you like polishing... I hate it!

[MomijiTMO]

Great, one look at this thread and I'm capped. lol

Anyway, SUBBED!

[Rayce185]

Thanks guys! Your posts are what keep me motivated!

Yeah the TEC cooling will probably have the biggest cooling gain, but is also the most work to prepare (insulate). I have spent the last few hours doing so and should have everything up and running in a bit! (if it works, LOL)

[LowRun]

Like others have said, you use way too much TIM on everything there + when done lapping you should not use any polishing stuff, only clean the lapped parts with isopropyl alcohol.

Otherwise, nice build log.

[Rayce185]

Yeah I do use alcohol to clean everything off.
I studied watchmaking, where we also had classes in goldsmithing. I guess that's where I got the tendency to polish everything with compound. Surface compaction and all

And IMO the last time I applied TIM on the dies (second to last set of pictures) it's just the right amount, right?
I have a picture of it deinstalled and shows how the TIM spread. It'll come with the next log update.

[Rayce185]

By the way: If anyone has experience with TEC insulation, please throw me a PM ASAP. I have a few questions that I'm stuck at.