Air Cooling Project: 101. Lowered CPU load by 9-10C

[Boldeagle]

Well this was my first ever build and it began in August 2006. Since then I have upgraded CPU from 805D (sick OC beast) to Q6600, Vid Card from x1900xtx to 4850 and added 2x1Gb of Crucial to give a total of 4x1Gb of Crucial. All is listed in my sig. Current settings have my CPU at 3.01GHz (9x334MHz), RAM at (334MHz@4-4-4-12, underclocked from 400MHz), Vid Card at stock. I’m an Aussie so all measurements will be in “metric” (the logical choice!).

After acquiring a 4850, I became one of the many that suffered in game instability and the only way I could resolve it was to back of my OC from Q6600@3.15GHz to 3.01GHz and then I could finally play Crysis for hour upon hour (yes I had the Fan Profile mod and had tried 8.6, 8.6Hotfix, 8.7Beta to 8.7 and tried many variants on voltage). Frankly this frustrated me to lower OC but I had previously been hitting the wall at 3.15GHz anyway. So with this in mind and the warranty expired on my MOBO it was time to go a modding. The objective being to improve CPU, NB, RAM and overall system cooling, to increase the inflow while maintaining or lowering noise.

The project entailed 4 components;

1. Modding the P5WDH by removing the "impairing" cosmetic NB and SB covers and refreshing with a quality TIM (Artic SIlver 5);
2. Lapping the CPU, HS and SB HS;
3. Making a Memory Cooler; and
4. Adding an extra 120mm inflow Fan

Pre-modding CPU was hitting 66C on OCCT for 1 Hour, after-modding has my CPU at 56C after 12mins. Although the 2nd run is shorter the CPU was fully loaded and previous run show’s an oscillating temp range with peaks of 66C@10mins. So with the TIM still bedding in I feel it is reasonable to see at least a 10C improvement, results below. Further testing will occur over the next few days and then onto the OC run.

[Boldeagle]

Alright it’s time for the fun stuff.

1. Modding the P5WDH:
I have found 2 very good guides for modding the P5WDH to prepare for the C2D range and followed them. Basically, they indicate to remove the NB and SB HS and dump the crap TIM and vanity covers to improve the chipset cooling, then you can crank it more. Arguably the best link is (you will have to go to the end of the thread to find a post for the pics);

Making P5WDH ready for Conroe: An Illustrative Guide - XtremeSystems Forums

After removing the HS I found something that resembles cement and although the guide indicated you could use “Goo-off” I contacted my local PC shop and they gave me a “huh” when I said goo-off and he suggested Methylated Spirits, to soak for 5 mins if you can. So after scrubbing at the cement with my Isopropyl Alcohol and Q-tips for 20mins I tried what the rep said. Damn even after 15mins they gunk was very stubborn so I finally used a razor to scrap the crap of the SB IHS and HS. The NB was another story so I had to scrub a “slightly” larger (than chip) section with my Q-tip on the HS and tried for another 5mins with my Q-tip on the NB chip before tentatively using the razor to scrape the exposed chip to finally get it clean. TIM application is a curious one and after reading a “solid” assessment article, I adopted their approach (modding slightly);

http://benchmarkreviews.com/index.ph...d=170&Itemid=1

Basically I applied a tiny amount of TIM to the surface, plastic wrapped my finger and spread around a very thin layer. After all TIM is primarily there to fill in the micro voids, assisting heat transfer is secondary, and if you follow the norm and give the HS a twist before screwing down you should expel most voids.

Because I could remove the fasteners from the SB HS and its’ size, I lapped it.

[Boldeagle]

2. Lapping:

I’m sure there is a guide on this site and I read others, also watching 2 u-tubes on this topic. I obtained the following grades of sandpaper from my local chain hardware store “Bunnings”, but even better was my local chain automotive parts store “Super Cheap Auto”;

• 600
• 800
• 1200
• 1500
• 2000

Naturally after writing this (and thus completing the project) an article of interest appears from "freecableguy" (who is that masked man) on lapping a LGA775. Frankly he is an Electrical Engineer (I'm an Enviro Sci myself) and writes some of the most technically detailed articles I have read to date, but sadly he has ""disappeared"" and it is a great loss to the PC community;

How to lap the Intel LGA775 IHS flat - The Tech Repository Forums

Previous research efforts made me adopt a method of (whereis FCG! );
·Slightly wetting the paper with soapy water;
·Sliding back and forth, applying light pressure (a stable pressure to maintain flush surface contact, but not “force”) for 10 times then rotating 90˚ and repeat for ~4 complete revolutions.
·Polishing
·Cleaning with alcohol

The CNPS9500 had an exceptional “finish” from start and I think I may have actually degraded its’ surface in the process, albeit slightly. Whereas the CPU and SB HS had obvious lateral micro lines (grain) running across them (clearly visible to the naked eye).

After the final grade (2000) I was surprised to note obvious micro scratches still present and the guides I had read aren’t very clear on the “final finish”, with most not discussing the polishing and buffing (you could almost believe the sandpaper alone will give you the finish, with some). Thankfully I had a metal polish lying around and was able to buff out the vast majority of micro scratches and obtain a “very good” finish (next time will be even better).

[Boldeagle]

3. Making a Memory Cooler:
I had never really thought about my RAM too much until I hit the FSB wall and with the 4850 causing instability I was thinking extra cooling here might help. The true inspiration came from the following guide for making your own memory cooler;

HOW TO MAKE YOUR OWN MEMORY COOLER(56K UN-FRIENDLY) - Overclockers Australia Forums

I had some nylon washers lying around so I added one to each bolt right below the fan to assist with vibration dampening (rubber washers would be the best).

[Boldeagle]

4. Case inflow increase;

I bought my Cooler Master Centurion 520 because of the front inlet filters, didn’t have a PSU and was at a good price. My front panel is completely lined with filtering material (works incredibly well as I shook it and had a thick dust cloud) and I only have the one 5 ½” hardware that left 4 empty bays where I could mount a 120mm inlet fan. I further removed the lower HDD rack to increase the lower fans inflow potential. But with my non-modular PSU, size limitations and lack of any channels cable management is a nightmare (I spent an hour trying to sort this during the project but to no avail). ElDonko has some of the best set-ups I have seen and now I can see why. I bet using a case “significantly” decreases “performance potential” by impairing CFM in air cooled systems.

[Boldeagle]

I use OCCT because imho it generates the greatest stress for my Q6600 and is the first app to show any system instability. Furthermore I love the logging and plotting of data. Here are the initial results. That is a 10C decrease for loaded CPU with initial testing.

[freakgeek1337]

If you want to lower your temps even more you can take care of that wiring mess there. But nice job otherwise. I also think you should buy some Arctic Cooling MX-2 or some OCZ FreeZe. That will also lower your temps.

[Boldeagle]

Well I took care of the wiring mess and lapped my CPUs properly (no vanity issues here!!!!). I probably could get an extra 1-2C from other thermal pastes (very hard to actually quantify, but I will be testing!). But my new system set-up has given me another 10C on the video card load, 3-4C on the CPU load and maybe 3-4C on NB, SB and or system in general. The Video Card range has dropped by 10C and I cannot get it to go over 70C where it used to range from 50-80, now it runs at 50-70. Idle temps did not change but load has shown I have removed the thermal backwash from the case build. The hottest component in the build are the HDD's which are very warm to touch and I would estimate 45C. Previous case build had a primary inflow over the HDD with a large % taken by Vid card and the remainder by the CPU HS. These temps will scale with ambient and would have been a primary contributor to GPU and overall system temps.

[jay51]

excellent thread,so much great info,thanks

[Boldeagle]

More will come and thank you.