(pseudo-) Diary of a newb [lengthy]

[killem]

Hi

It all really began with an idea that my best mate planted a few weeks ago.

But let's start with the background...

I had an XP 2100+ in my box, which had average temps (under load) in the region of 50C-55C, with a stock AMD HSF and some AS3. I knew this wasn't ideal, but wasn't ready to make any changes.

3 weeks ago I bought a 2800+ along with a Thermaltake "Silent Boost" (the dual ball-bearing version, as the supplier could not source the hydrowave variety). So I plonked my 2100+ into my girlfriend's PC along with the copper-base HSF from the 2800+ (even using the regular TIM pad). I was most interested to see that the 2100+ temp remained below 40C. It seems that even a copper insert helps, although sometime I'll replace the TIM with some decent thermal compound.

(note: please bear in mind that our climate causes average summer maximums of between 25C and 32C)

Having read around the Net that Shin Etsu compound is better than AS3, I got some Cooler Master Premium. It's just as awkward to apply as AS3, but for a different reason. Whereas the AS3 is a messy affair, the Shin Etsu stuff has a thicker consistency (almost resembling putty). I found that even with the decent compound and an all-copper heatsink + 80mm fan, my CPU still ran up to about 55C.

The next step was to improve my case's airflow. I have an Acer QF50 mid-tower, a fairly standard case. I added a rounded IDE cable for the HD and tidied up the PSU cables as best I could (bundled with cable ties). Then I added 2 Coolermaster 80mm fans: front intake and rear exhaust.

With all of this, I was impressed to see the case's temperature drop safely to a max of 35C, even on hot days with many hours of uptime.

This is where my friend's idea came to play. He mentioned have ducts taking the air from intake fans to where it was needed most. Something grabbed me, about having a neat duct from an intake fan. I'm not interested in cutting holes into my case, BTW.

Lots and lots of reading up on viable materials, brought me to this thread and clocker2's idea of using Sintra. I was despondent about the chances of finding such useful stuff locally. More research, a few phone calls later and I found a plastics supplier which could sell some Komatex to me, which is another brand of the same PVC foam-board.

I happily spent a whole Friday evening getting to grips with moulding this at home. I was able to fashion an experimental duct (a test tube, if you will). I realised that having a flattened 'S'-shaped duct which ended in a sharp bend over the CPU would be quite tricky. Still, I was satisfied that I had a material that could possibly do the trick.

My friend had suggested that before finally committing to building the duct, I might also try a mockup "side-panel duct" from cardboard, to see if ducting external air would be of much benefit. So I carried on to build one, with a slightly tapered duct which enclosed the CPU fan. As the Silent Boost's fan draws air in from its sides and top, I made sure that there was room around the fan housing and added a little collar to enclose its base.

Initially I ran the mockup as-is, for 45 minutes. With the CPU idle for 10 minutes, the case temp was 36C and the CPU temp was close at 35C. I then ran a few apps for 30 minutes, including Prime95's torture test. The case and CPU temps matched each other at 39C.

I installed an 80mm Tornado fan at the opening of the duct (almost 10cm away from the CPU fan. At the 15 minute 'idle' point, the case was 38C and the CPU was 33C. After 45min of playing Unreal Tournament 2003, the case temp was 39C and the CPU was 35C! It stayed at that temperature after leaving the PC running Prime95 overnight. I suspect that the lack of sound and video demands in Prime95, meant that one was not seeing maximum temps. Also, the duct did seem to hamper the airflow in the case, as I saw the temperature exceeding the values I have with my usual setup, by about 5C.

I think that seeing the CPU temp drop by about 15C proves how effective a side-panel duct is. Still, this is not what I'm interested in doing to my case. This lead me to PM the same 'clocker2', asking for suggestions for working with Sintra.

A wonderful suggestion was made, to consider having a duct from a rear intake fan, which is far simpler to build. I will need to angle the current (warm) exhausted air away, but I'm highly optimistic.

So, I'll be implementing this idea ASAP and updating the results here.

 

[warlock110]

i still don't get how u would do it, some pix would be good . great project though, me and other would probably do it too.

 

[clocker2]

So, I'll be implementing this idea ASAP and updating the results here.

Killem,
I think your latest plan for a duct will not only be successful, but also far simpler to implement.
That said, I'm almost sorry to see the original idea go untried as it would have demanded fabrication skills far in excess of mine...I was hoping you could teach me how you did it.

 

[killem]

i still don't get how u would do it, some pix would be good . great project though, me and other would probably do it too.

I will make an effort to get hold of a friend's digicam and snap a few pix of the finished project.

(with apologies to clocker2, who's read all this before!)

(*) My original idea was to have the duct start with a round opening at the front intake fan. It would flatten and narrow a bit, going into a lazy 'S' shape. This was so that it would curve between my gfx card and the corner of the hard-disk cage. It would then curve back toward the CPU/HSF, where I would've made it round again, after a 90-degree bend down to the CPU fan.

In truth, a maniacally complicated construction that I imagined would involve a wooden board with nails to guide the warmed, softened plastic tube into the initial shape. Clocker2's suggestion was excellent too, of making a 4-sided duct with the sides simply cut to the required shape.

Weeks ago, when I started out, I seriously considered making a "front-to-rear" duct, which was also a winding duct very close to the above description (*), but then continued on to a rear exhaust fan. The idea was to have a wide 'pocket' section enclosing the HSF, with air flowing from the intake and being sucked along by the exhaust fan. The 2 main hurdles were:

a) securing the beast safely, so that the CPU is not under extra pressure

b) whether the air would really help to cool the CPU.

This whole 'front intake' approach is just not as effective as a simple, short duct with possibly one bend at the CPU.

 

[TheWesson]

Or flip the CPU heatsink fan and take the CPU exhaust to the rear case exhaust.

How well that works depends on how well the CPU HS fan works in the flipped position. At its best, this would work about as well as ducting fresh air on top of the CPU/HSF, and it would also keep your case very cool with minimal fannage.

You can see my results in my signature below ... I flipped a Vantec Aeroflow fan, and I'm doing pretty well considering the overclock and the low HSF speed.

I got a rise of about 2c from flipping the fan, and I believe I make that back from the motherboard staying cooler (and taking heat away from the CPU through the socket.) Also, no PS 'chimney' necessary ...

By the way, I think it would also be very interesting to have a "pocket" enclosing the base of the CPU HSF (where it exhausts) and leading that to the rear case exhaust fan. That would be hard to engineer in the limited space, though.

the wesson

 

[killem]

Hi

It's been almost a month since I started out, but I finally have something to report.

After multiple attempts at building a rear-intake CPU duct, I've finally completed a working model. As usual, blind ambition and enthusiasm wound up being tamed by some practical experiences and a few painful lessons. Through trial-and-error I modified my designs until I eventually made an almost overly-simple version.

Most difficulties came from getting the Syntra/Komatex PVC board to the form I had mentally pictured. It turns out that it's not easy to build a duct so that it's a single sheet with bent pieces that join up (so as to minimise the number of glued bits). It was just not feasible, whilst incorporating a slightly angled section from the intake fan to the box enclosing the CPU fan.

The best compromise I could achieve was to have one length, folded back on itself to form the top and bottom pieces, with 2 side sections. I had to angle only one of the side pieces, to bring a 100mm-wide section at the CPU down to an 80mm intake-fan. Yep, I have a diffuser effect now. shrug

Thankfully, after adding a gutted fan's housing, as a shroud between the heatsink and its fan, the need for the angle was reduced, since the CPU fan was (incredibly) raised to within a few millimetres of the intake fan's level.

(note: the 25mm-high shroud made no difference to CPU temps)

Also, on the topic of glueing: the construction was not satisfactorily sturdy either. So I chose to cut 'tight' slots in the duct's top and bottom pieces, for the sides. This works nicely.


The results after one night's testing:

Ambient/room temperatures were approximately 20C
Case temperatures maxxed at 35C
CPU temperatures maxxed at 52C


Not the reduction in CPU temperatures I'd been hoping for. Although I'm pleased to see the temps remaining below 52C, it's only an overall improvement of about 5C. At least I'm relieved that the duct isn't interfering with the case temps whatsoever. (that does depend on where and what exactly the case temperature reading is measuring)

There are some issues I am going to look into:

- The air-channels that I made for the rear, to keep the air from the exhaust-fans away from the intake, may just be hampering the intake fan (similar to the effect of case grills). I might revise the intake's channel and 'thin out' the grill a bit

- It may be that the temperature-sensor controlled intake fan isn't blowing strongly enough. I attached the sensor to the northbridge's heatsink, but should probably have attached it to the CPU's heatsink instead.

- It's possible that the overall air intake needs to be seriously high at any rate. This would mean using an 80mm Tornado, which I'm really not inclined to do. (we only have CoolerMaster and Tornado ranges available locally).

I'm seriously chuffed though, since this is the most effort I've ever put into a personal project. I've learned loads about PC air-cooling, plus working with foamcore PVC, which could just mean returning to this particular one in the future.

Hopefully I'll have a few pics to show, RSN.

 

[clocker2]

It's impressive that you stuck to it long enough to come up with a finished piece, even if it doesn't meet your expectations.
As you pointed out, you learned a lot in the process and that's probably worth more in the long run than the duct itself.

So, any ideas for improvements to the duct itself?

 

[killem]

As you pointed out, you learned a lot in the process and that's probably worth more in the long run than the duct itself

Indeed. So-much-so, that I was confident enough to sort out a friend's temperature hassles quite well too. His 2200+ was regularly running over 60C, even with the sidepanel off and a housefan blowing in. The poor chap's supplier had the brilliant solution of underclocking the CPU, from 133MHz FSB to 100MHz, reducing it to a 1500+ for the past year. It didn't help that the box was sitting on a thick carpet, which blocked the lower intake and rendered a snuggly blanket of fluff on his HSF!

I sorted his airflow out and returned the settings to default, to find his CPU nbow also staying around 50C even on hot days, with hours of full operation.

So, any ideas for improvements to the duct itself?

Nothing noteworthy, whilst remaining within the domain of simplicity and practicality. I would love to reduce/taper the duct from the intake to the CPU, to raise the air-pressure, but can't think up a functional design for this idea yet.

I am keen to get the shape closer to circular, to maximised the air taken from the intake fan. This could mean making a somewhat octagonal shape.

Currently I've had to exclude about 5mm on the left and right sides. This was to retain access to the nuts+bolts I've used to secure the external channel, to the case, as well as to the intake fan. Naturally, I'm also still hunting for more suitable bolts, that are wide enough so that nuts are not required.

So many ideas, so little time!


PS Forgot to mention, for the benefit of other (wannabe) DIY'ers who are similarly faced with limited funds, or specialised equipment or facilities, the 'equipment' I used:

for measuring/cutting

- clutch pencil
- old triangular ruler with 6 measuring scales
- high-school geometry plastic triangle
- utility knife.
- various wood sections for cutting on (generally the size of a good bread-board)

I learned to start cutting by making a few cautious initial impressions to ensure that the tip has a track to cut along, then drawing the knife firmly and repeatedly along that.

I definitely recommend having a right-angled triangle from high-school geometry, for getting the edges square.

for bending/shaping

- 2 knives with straight (enough) spines
- steam iron

I heated a knife by holding the blade against the iron's base, then pressed the blade on the PVC, which made folding easier and relatively precise. The really nice part about using the iron was that it allowed me to control the temperature quite well. In some cases extra heat was necessary, but for some finer folds too much heat flattened the PVC.

Oh, for the rear air-channels I aimed for a slightly curved effect. This was achieved by pressing the pieces against the outside of a large pot heating on the stove.

All folding was done VERY, very carefully!!!