Discussion in 'Project Logs' started by jokkos, 20 Mar 2006.
i love this case...
bow the block perhaps? try getting a lil' thicker diameter o'ring perhaps?
That's how we know you're really dedicated. I've had quite the same problem with the headspreader on my pentium d .
The story continues – internal works part two
Part one of this update.
Since there are plenty of guides to find about lapping a cpu or heatsink, I wasn’t planning on showing pics of it. But I just couldn’t resist
Here goes. In the previous update I told you that both the CPU and the waterblock weren’t flat. I was wrong, as you can see in the following pics.
- a flat working surface (I used a glass tablet of an old furniture piece)
- sanding paper: (180grit), 300 grit, 600 grit and 1200 grit
- Q-tips and alcohol 96%
- CPU: about 45 minutes
- Waterblock: about 8 minutes
First: the CPU, or better its HSI.
I started with 300 grit paper. After a first pass (about 1 minute), a bit of copper starts to show through. Mostly in the corners and on the edges, which is logical since the HSI is slightly concave.
A couple of minutes later. In the middle of the sink, the copper starts to show. BUT: there are clearly two zones where there’s nothing happening yet. And these are the zones where the contact between the sink and the waterblock was bad. The HSI was the reason for the bad contact after all!
Here I decided it would take very long with 300 grit paper, so I took some 180 grit for a few minutes to speed up the process. After 8 minutes of sanding, this is the result. Still two zones where the copper isn’t really showing yet.
Back to the 300 grit. This is after 15 minutes of sanding. Compare it with the previous pic. Not much has changed right?
After 20 minutes of sanding, it starts to get a bit better.
Finally, after about 28 minutes, it’s all copper. You probably noticed that on the sides, there are places where it’s still not sanded. Note the bottom edge of the sink. But these are really ‘deep’ so I stopped here.
Then I switched to 600 grit. Because of the thorough job with the 300grit, the rest of the sanding was easy. A good 5 minutes later, the surface was a lot finer.
Then on to the 1200 grit. I didn’t want to achieve a perfect result (=mirror), I just wanted a flat sink that would make good contact. The scratches you see are tiny and exagerrated because of the direct light on it. This took about 10 minutes.
The sanding was done. Time to clean the whole CPU. This is best done with ‘pure’ alcohol. You can buy it at a pharmacy. The Q-tips are handy, all the dirt and dust came off very quickly and easily. By the way, it takes a real man to use pink q-tips!
At the start of this post I said I was wrong thinking that both the CPU and waterblock weren’t flat. Why? Simple: I shouldn’t have doubted the Aqua-Computer waterblock….see for yourself:
I started with the 300 grit paper on the waterblock. I immediately felt that it had almost no resistance to the sanding paper. This is how it looked after one pass (I timed it, only 24 secondes of sanding)! Apart from the extreme sides, there’s scratches from the paper everywhere….
I kept on timing…after 3 minutes and 21 seconds, it was already so flat even my camera couldn’t focus anymore (or maybe I just focused badly lol)
7 minutes and 49 seconds after I started on the block, it was all done. Ignore that marking beneath the top left hole, I banged it hard against something right after I bought it. It is far away from the cpu so it doesn’t matter.
After cleaning the waterblock, I let the CPU and block alone for a while. To make sure the CPU was really 100% dry.
Conclusion: lapping the cpu was well worth the time. As for the waterblock, sorry Aqua-Computer for doubting you, this wasn’t necessary.
Test pic in the next post…
The story continues – internal works part two
(2/2, see previous post for the start of this update)
As I needed to go somewhere and eat something, I left the cpu and waterblock for a while. This way I was also certain the cpu would also be completely dry before I put it back on the mainboard.
Yesterday was a typical autumn day in Belgium: lots of rain, dark clouds. So the light wasn’t going to last very long. Therefore I decided to finish the holes for the pass-through fittings first.
Here they are after drilling with my largest drill (13mm – I really need a stepping bit fast).
A bit of filing later (the fittings’ diameter is 14mm), they fit perfectly.
After this I also installed the extra wiring that goes to the external rad.
But: back to the lapped parts.
After putting the cpu back on the mainboard I did a quick power-up so I could see that it was still alright. Bios showing dual core E6600, so that’s ok.
Then I put a VERY small amount of thermal paste on the sink. A lot less than I put on last time with the bad contact. I didn’t spread it all the way to the far edges, I wanted to see how well the block would do that. I fastened it (not even all the way, say 80%), took it back off, and well, I’ll let the pic speak for itself:
Needless to say I’m quite happy with this result!
I also installed the chipset blocks. Here’ a pic with the three blocks…
I’ll get some better pics later, the weather gods weren’t kind for this sort of photography.
Second to last job of the day: I installed the Microcool sinks on the mosfets:
And another angle. They look good, don’t they?
After that, I did the last job of the day: I put the mainboard back in the case. But the light was so bad by then, I didn’t take pics. I’ll try to get a good one today of the mainboard with the new blocks.
grat job with the cpu
nice to see you working on the case again
Expecting better temps too Jokkos ?
yup...that's the idea. According to stuff I read, it could range from 2-5 degrees in idle and 5-15 degrees at 100% load. I'll of course test it and post here once I get the system up and running again.
The story continues – internal works part three
Here’s a better pic of the mainboard, it’s back in the case.
And another one with the rest of the system installed. Plenty of room between the vid card and the southbridge.
Now, it doesn’t take a genius to see where the tubing of the watercooling needs to make its tightest move: between the cpu block and the nb block. Of course this has to do with the lay-out I chose for my loop and the fact that the blocks are ‘horizontally’ oriented (fittings left and right).
I was pretty confident I could make the connection with the PUR tubing, but that turned out to be quite impossible. It kinks, I tried 5 different pieces with different length and unless I put in a huge piece of tubing, the bend can’t be made. And then it doesn’t look good at all. Soooo…I went for another solution: pvc tubing. That’s also better for the nb block in my opinion, the PUR tubing puts a lot of stress on it and since the nb is such a small chip, I wouldn’t be comfortable with that.
Using a piece of pvc tubing is one thing, but the ‘problem’ is that I only had blue pvc tubing lying around. But since the case needs to be up and running again by the end of next week, with a full windows reinstall and everything, I had no time to order transparant tubing somewhere and wait for it. So I went with the blue tubing. And, I think it doesn’t like too bad actually. But that one part seemed so lost in there, so I also used the blue tubing for the nb-sb connection. Oh yes, I also discovered the Swiftech smartcoils and I absolutely love them.
Here are some pics (there’s one part of tubing not in yet: from sb to the back panel):
I like the combined look, because the blue tubing is more in the background. Also the tubing coming from the pump to the graphic card goes nicely ‘between’ the two parts of blue tubing. But that’s just my opinion.
What do you think? Is the blue tubing okay together with the transparant one or would you replace it with transparant tubing later on?
I think it looks fine the way it is. Did you consider using the clear coils? The blue coils seem to distract from the rest of the case quite a bit.
The blue coils are also UV-sensitive and I want to see the effect with my UV leds.
You make a good suggestion though, I actually bought these blue ones before I found out there are also transparant ones. The shop where I bought them doesn't have the transparant ones.
If the UV-effect turns out to be too fugly, I will consider the transparant coils!
Why didn't you do pump-cpu-gpu-nb-sb? It would be a lot less stress on the tubing, IMHO.
Thanks for your reply, but take a look at how close the gpu block is to the nb block. Would be even worse than going from cpu to nb (imho)
Anyway, the loop is doing and I just filled it. Now worries so far, the rest of today no power for the pump, just basically checking a few times for leaks. Tomorrow morning the pump goes on until late in the evening. If everything goes well, I can comemence the windows install on tuesday. Not too bad.
......sometimes you have to give up on cooling performance to gain the overall
apperance a modder is after ............
About the blue tube .... since you asked ......... I would change over to all clear ,just my opinion .......
if you like it keep it ... as it also looks good ...........
uuuups.. sry pastet the wrong text^^
i really love your case! it`s giving me something to dream at night;-)
keep the blue tube !
The end is near – internal works part four
I’ve got a new update for you. Took some pics in the past days, so some are with lots of light and some are during the evening with yellowish light. They should give a good idea of the new look.
After the experiment with the blue pvc tubing inside the case, the need came to put it for the outside part as well. The PUR tubing isn’t flexible enough to stay clear of the wall and window tablet behind the case and the box where the external rad is in. I also put in quick-disconnect fittings.
The quick-disconnect on the tubing towards the external rad is reversed to the one on the tubing that goes back to the case. This way, I can remove the whole external rad if necessary and just put the two ends of tubing coming from and going towards the case together. Could come in handy someday.
Here’s the case and the box more or less correctly positioned. It will of course move back to the right on the desk, but this was just to check the clearance of the tubing. Looks good.
Close up of the backside. In its final place the case is not bothered by the window tablet, so there’s enough space.
I filled the loop, no problems there thanks to the handy fillport. The previous time, the loop needed about 45cl of water. Now with the extra blocks and external rad, is took about 80cl of liquid to fill it.
Then came the testing. The pump still happily moves the water around. Here’s a close-up of the Aquatube, you can see there’s still a ‘bubble’ from the water that comes up in the Aquatube from the internal rad. It’s smaller and less ‘active’ then before, but it’s still good.
From the Aquaero readout, I now have a flow of about 45 liters/hour in the loop. This is a lot less than before (100 liters/hour) but still enough. This amount will also change because I need to finetune the performance. But I like to use Aqua-Computers software for it, and for that I obviously need hard drive access. So this will happen in a few days.
Now some dusk/evening shots.
An overview of most of the inside loop.
Another one. In the right my auxiliary power supply that is connected to run only the Aquaero and pump.
Here most of the internal lighting is on. The UV effect of the coils is not very pronounced but the side panel isn’t on and that has 20 more UV leds.
A close(r)-up with some artistic orange…lol.
This is more or less how it will look seen in frontal view. The outside tubing is not very visible, but the beige/brown Noctua fans are. I will definitely need to paint those. You can also see I top-filled the reservoir.
Finally, a shot without any lights on in my livingroom. Except the ones of the case.
See you in a couple of days with some more pics of the final look with speakers/monitor/side panels, also with some details of the (let’s hope improved) performance of the watercooling.
thats really awesome dude!!! i love the blue tubing! but as you said... change the color of the noctua fans ^^
outstanding good work! the out-of-phase connection is a nice trick! will extemly facilitate (or should i use "ease?? pls tell me!) the disconnection of the external rad .
can`t wait for pics with the side panel mounted...
it's looking great.
quick question, what did you use to mount the mosfet coolers?
@500mph: those are PLL sinks from Microcool (www.microcool.it). They come with a piece of tape on them. You know, the kind of tape that usually comes with thermal probes, but sticky on both sides. However, that wasn't really sticky enough. So I put some thermal paste on each mosfet and then stuck the sinks on that. At first, it was still loose, but after the paste cured, the sinks are on pretty firm.
I'm quite happy with the new cooling power. There must have been something really wrong with the contact area between the cpu block and cpu...
Before, the cpu would be at 45° in idle, with all the fans in the case and on the internal rad at 0%. In that situation, the mainboard (NB and mosfets) became pretty warm, so the casefan would inevitably start and at some point, the water temp became higher than 29° (which was the target temp in the Aquaero) and from that point on, the fans were contantly on, going from 60% to 90% and back to keep to the temp. The cpu would then drop to about 38°.
Yesterday the pc was on for about 5 hours, I was constantly installing, configuring, restarting, testing, etc. The cpu never went above 39°, the mainboard (mosfets) stayed below 36° the whole time. All the fans stayed off for more than 4 hours, only at the end the water temp reached 29° and then the fans came on and off for a bit. Much more silent than before. And especially: even though it had some load most of the time, still the cpu is -when wc was passive- on average about 5° cooler than before. Can't wait to test it more thoroughly, setting the wc fans at 65% for example, also at 100% load (at 100%, the cpu should be at least 10° cooler than before, judging by yesterday's experiences!).
I'll report more accurate results after those tests...also will be posting some pics with side panel on and the 21" widescreen. Oh yeah, and the usb-vga adapter for the third monitor, it works very well!
All done, second time – pics and temps
Part 1 of 2.
A bit later than expected, but as promised: here are some final pics and some rambling about the new temps. This is the final update, so enjoy! For clarity I’m putting this in two separate posts.
First some pics about the final look on my desk.
Here’s the first one, with the new Samsung 215TW. On the left above my cell phone you can see the blue usb-vga adapter.
Everything okay, microsoft user agreement of windows installation…
A few pics after the first boot into Windows, with different focus.
Hmm, 1680x1050 sweetness…
Here’s a pic of the external tubing. My window tablet came in handy for the long tubing that goes back to the case.
I installed the usb-vga adapter to use three monitors without a second vid card. It actually performs way better than expected. It plays divx hassle-free, dvd is also ok but it is a bit slow in menus and so on. So for basic desktop-like work, it’s perfect. The real beauty of it though, is that it only shows up in the screen where you put the monitors in the right position (see pic). It doesn’t show up in nvidia’s control panel or anything, it is primarily a usb device. This is good for me, since I don’t have the issues anymore that I had with the second pci vid card.
Here it is, a nice blue led that fits in with my system as well…all good.
To finish, a couple of low-light shots.
That’s it for the pics. If you’re interested in the new temps of the extended watercooling, check out the following post as well…
All done, second time – pics and temps
Part 2 of 2.
In this post I’ll try to give an overview of the new performance of the extended watercooling.
Before I start, there are a few things to take into consideration:
1) there have been pics in this thread with temps of the first wc loop in this case, but those temps aren’t completely applicable as reference values. The room temp for the new tests was about 2 degrees higher. I also flashed the bios on my mainboard and in the newer bios, there have been ‘adjustments on the temperature algorithm for the cpu’. And that really does make a difference. Before, tools like Everest, Hmonitor etc always reported a cpu temp that was higher (a couple of degrees in idle and about 5-7 degrees in load) than my bios read. After the flashing, the bios temps are now consistent with those of the monitoring tools. So I’ll adjust my ‘old’ reference temps a bit, setting the water and cpu temps a bit higher;
2) the lapped cpu made a nice difference in ‘idle’ mode (ie while installing and configuring windows and software) and the watercooling almost passive. Almost passive, because the fans did go on from time to time, for a short period. It turns out this made a big impact, as you can see below. The lapped cpu does give better temps, but the difference of about 5° in idle goes to only about 6-7° with 100% load. I hoped for a bit more. Nevertheless, these 6-7° make a big difference on required fan rpm and noise, as you’ll see below.
So let’s see the numbers, right? Right! I’ll try to keep it as clear as possible…
Basically, I ran five tests, each for about an hour (except one which I needed to stop after 30 minutes):
1) windows desktop (idle) with the 4 fans on the rads and the casefan at 0%;
2) windows desktop (idle) with the 4 fans on the rads and the casefan at 65%;
3) 100% load with the 5 fans at 0%;
4) 100% load with the 5 fans at 100%;
5) 100% load with the 5 fans at 50%.
This should make sense, I wanted a minimum and maximum scenario in both idle and load. The casefan got the same % as the watercooling fans to see how hot the mainboard and mosfets get in the new setup. In test 2, I went for 65% fan power instead of 100% because I figured it would be plenty (it was the avg. % of the ‘old’ watercooling setup with comparable room temp). Because of the awesome results of test 4 –see below- I added test 5.
So, on to the temps. I’m going to list them in the following format:
Test# / water temp (W) / cpu temp (C) / gpu temp (G) / mainboard temp (measured at mosfets) (M) / new or old temps
*idle with passive cooling*
1 / W 32,6° / C 44,0° / G 51,0° / M 39,1° / new
1 / W 34,8° / C 51,0° / G 55,0° / M 47,0° / old
*idle with 65% fan power*
2 / W 26,8° / C 36,0° / G 44,0° / M 31,1° / new
2 / W 28,6° / C 40,0° / G 47,0° / M 39,5° / old
*100% load with passive cooling*
3 / W 37,6° / C 61,0° / G 63,0° / M 44,4° / new (aborted after 30 minutes)
3 / W 38,0° / C 62,0° / G 65,0° / M ??,?° / old (aborted after less than 10 minutes)
*100% load with full 100% fan power*
4 / W 26,4° / C 46,0° / G 50,0° / M 33,2° / new
4 / W 29,0° / C 53,0° / G 53,0° / M 35° / old
*100% load with 50% fan power*
5 / W 32,6° / C 54,0° / G 57,0° / M 34,8° / new
5/ wasn’t a realistic scenario in the old setup, temps would be too high for me to feel comfortable with it.
That’s it. I hope it’s clear…
A couple of observations/experiences so far:
- the biggest effect of the new watercooling setup is to be found in the addition of the NB (and SB) to the loop. Before, the casefan was running a lot of the time to cool the heatpipe on the mainboard down (I had it set to keep a temp of 39°). In the new setup, it never really needs to run in idle, but if it does, it can stay at 50% (then it is unnoticably silent) and it cools more than enough.
- before I changed the loop, I asked in a couple of forums what the effect would be of a second radiator. Most people who responded were in agreement that it would make a difference of 1-2 degrees. Looking at the numbers above, they were right.
- the cpu clearly runs cooler than before. This is partly due to the lower water temps. However, the difference between the water temp and the cpu temp is always lower in the new setup than it was in the old setup. So the lapping did have an effect, albeit smaller than I expected.
- The cooling power of the new loop is way bigger than that of the old loop. It takes a lot longer for the temps to change than with the old loop. When going from 100% load to idle, the water cools down a lot faster than before. Which is of course logical: I doubled the cooling surface and the flow rate is lower than before.
- Finally, I think test 5 says it all. I *could* set the fans at 50% and run a stable system, at least until the warm summer months. This wasn’t possible in the old setup, the cpu and mainboard temps would swing out of control under load.
The Aquacomputer control unit (Aquaero) and software allow to set ‘target temps’ that the system will try to achieve. If the actual temp is below the target temp, the fans are off. When the actual temp is close to or above the target temp, the fans come on and go faster the bigger the difference between actual and target temp is.
In my old setup, I set the target temp for the water to 29° and for the mainboard it was 39°. Look at the temps I reported above and you’ll see that the fans were on a lot of the time, especially in the warmer months. At 100% load, the fans were almost always at 100%.
For the new setup, I ‘m going to start with a target temp of 31° for the water and 36° for the mainboard. Again look at the temps: this would mean a passive cooling at idle most of the time and on average, less than 70% of fan power needed at 100% load.
And this is where I think it’s time to stop writing. I’m happy with the new watercooling setup. It clearly is a lot quieter than before with better temps, and that was the goal. The laaping of the cpu had less impact than I expected, while adding the NB to the loop was a very wise decision, looking at the new temps. It was fun. Hope you enjoyed this long read. Thanks for your comments and support!
Separate names with a comma.