Clevo M570TU Cooling Overhaul

Clevo M570TU Cooling Overhaul



Version 1.02

Status: Complete.
Last Updated: 22:30, 07 February 2009.
Author's Note: Please read post #2 before reading this article. Thank you. http://forum.notebookreview.com/showpost.php?p=5295253&postcount=2

The Goal

This thread is dedicated to the culmination of several months of effort to effectively improve the M570TU's cooling system. My initial goal was to make the M570TU self reliant on it's own cooling. But as time went by, I constantly found ways to improve the cooling of this machine to the point where I created a new goal to make this machine the coolest running notebook in it's class.

The Notebook

The M570TU is the latest generation of the M570 lineage that began with the momentus unveiling of the M570A by Clevo back in winter 2004 and it's public release in spring 2005. Over time, the M570A slowly evolved into the M570U in early 2006, the M570RU in 2007, and the M570TU in late 2008. Compared to other notebooks released by Clevo, the M570 is their longest serving chassis, along with the D900, seeing over five years of innovation, evolution, and success.

I bought the M570TU back in early February from RJTech. This is the third Clevo I have bought from them in the past six years; the first one being the D400V, and the second being the M38AW. The large size and modularity of the M570TU allowed great versatility and flexibility in OEM and aftermarket customization and modding. The M570TU also holds the title of the fastest mobile CPU/single GPU notebook on the market as of the last mobile Penryn/G92b generation.

Here are the major specs of my notebook ripped from my Steam page. http://steamcommunity.com/profiles/76561197979343959

Clevo M570TU
17.1" LG LP171WU1 TLA2 Glossy WUXGA (1920x1200)
Intel QX9300 2.93GHz 12MB L2 1066MHz FSB
Elpida 4GB DDR3 1066MHz 5-5-5-15
Seagate 7200.4 250GB 7200RPM 16MB Cache G-Force Protection
Nvidia GTX 280M 1GB GDDR3 640/1600/950

The M570TU in it's factory state has impressive cooling for all of the major components. The CPU has a very large radiator, heatpipe, and a dedicated fan. The GPU is encased in a coffin of copper cold plates with a huge radiator, twin primary heatpipes, a single secondary heatpipe, and a dedicated fan. The northbridge uses a modest aluminium heatsink. And the HDD uses the aluminium mounting bracket as a heatsink as well. With all of this cooling, the M570TU can handle all CPU and GPU intensive tasks without breaking a sweat.

The Overhaul

One of the major constraints that Clevo takes into account when engineering a performance notebook is mass. Originally, the M570TU weighs 4kg (8.819lbs) as listed on Clevo's website. 4kg is a bit on the heavy side for most users, and other notebooks in it's size class. Essentially, Clevo, like all other companies, made compromises to offer just enough cooling while retaining portability. However, I decided to forego mass and make cooling number one on the list. After completing the cooling overhaul, the notebook weighs precisely 4.863kg (10.721lbs). Copper adds quite a bit of mass, so if anybody is planning to incorporate the cooling overhaul into their notebooks and would like to avoid increasing mass, just follow the backpanel segment of this article.

My initial idea for the cooling overhaul was to increase mass and surface area to better utilize CFM circulating inside of the chassis. Increasing mass will allow more heat to saturate off the component and slow the degree of temperature change. Increasing surface area will allow better passive cooling when the fan is off, and dissipate more heat directly from the component and being vacuumed up by the fan instead of having heat travel across a long heatpipe to the radiator. This would divert some thermal load from the radiator and the heatpipes to the copper blocks mounted on the component.

The M570TU has plenty of ceiling clearance in the chassis for mounting several tall copper blocks onto the heatsinks. Each component has undergone a major facelift. To gain maximum effectiveness from the copper blocks, I used a hybrid method of application. Each copper block has been lapped perfectly flat on high grit sandpaper. I used Sekisui #5760 thermal tape on the outer border of the blocks while leaving the center free of tape. In the center of the block, I used a very, very minute amount of X23 to fill in any microgaps on the face of the block and on the component heatsink. Sekisui #5760 is the best thermal adhesive on the market bar none and will not fail to keep the copper blocks stuck where they are supposed to be. Even if half the the border was lined with the thermal tape, the block will not fall off unless one really tries to rip it off. Some of the blocks that are not on a flat surface or are "hanging" too far off the heatsinks do not have the hybrid method applied and are instead outfitted with thermal tape entirely.

The CPU cold plate has been lapped flat for maximum contact with the twin monolothic dies on the QX9300, and has been mounted with 11 copper blocks, and an additional cold plate on the radiator, which also blocks dust from entering underneath the radiator. The blocks on the left and right side of the CPU heatpipe have been arranged so that incoming CFM from the vent above will divert to the sides allowing more time for air to come in contact with the fins on the blocks. The blocks ontop of the heatpipe have been arranged parallel to the CPU fan so that all the CFM isn't blocked off entirely. A direct middle channel and two diverting channels for airflow works well here.



The RAM has been given copper heatspreaders along with five copper blocks each.



The HDD has been given a massive copper plate that keeps temperatures low even in extreme thrashing and high room temperatures.



The thermal pad on the northbridge has been replaced with a copper shim to ensure that the most heat saturates into the aluminium heatsink.



The wireless card has also been given four copper blocks to help remove heat away from the card. Adding copper to the wireless card fortunately has no adverse effects on wireless signal.



Finally, the GPU has been completely revamped. The entire front of the GPU heatcage and heatpipes have been lined with 37 copper and black aluminium blocks. All of the blocks have been arranged parallel to the GPU fan to allow maximum direct CFM from the vents above the blocks to the fan. Inside of the heatcage, five more copper blocks can be found to act as heat retainers. The aluminium lining bordering the GPU core has been given two copper beams to assist with transfering more heat into the heatcage. The heatplate has been lapped flat for maximum contact with the GPU die. All of the original thermal pads have been replaced with high quality Bergquist thermal pads. Also, the underside of the GPU core and memory have been lined with thermal pads to transfer heat into the bottom half of the heatcage to better utilize the third heatpipe. And to top it off, a huge copper cold plate has been added to the radiator, which also blocks dust from entering from underneath the radiator.



Further into my mission to improve the cooling system, I decided to drill holes into the backpanel to increase ventilation. A lot of users have been discussing that messing with the factory vents will disrupt the direction and amount of CFM entering the intakes scattered all across the backpanel over critical components. In this case, those users are correct. Creating more openings for air to enter and exit the chassis, especially in poor locations, will negatively affect internal airflow. However, there is an upper limit to this. My idea was to drill out the backpanel entirely to overcome poor placement and emulate the cooling benefits of having the backpanel removed. A combination of increased mass and surface area with increased ventilation allows for very effective cooling, especially at load temperatures. This allows heat dissipating from the added copper blocks and cooler external ambient air to easily enter and exit the notebook through the holes to more efficiently influence the notebook to a cooler equilibrium temperature. Air will still be entering the chassis from the holes if the fan is on, but the lowered displacement in CFM is completely outweighed by pure passive cooling, and the little CFM that is left will only benefit efficiency even further.

I used a non-electric hand drill and a 1/32in cobalt drill bit to drill holes into the backpanel. Each hole is spaced 5mm apart from the next hole. I decided to drill small holes instead of large holes because I wanted to retain structural integrity in the backpanel and to use a hole small enough to be effective without the need of a dust grill or filter. I have drilled holes over the entire GPU, HDD, RAM, wireless card, Intel Turbo Memory card, and northbridge. The GPU vent has also been given additional ports for exhaust to reduce turbulence.



The Test

I have conducted three stress tests for the CPU, GPU, and HDD. The CPU was tested on OCCT CPU Linpack. The GPU was tested on OCCT GPU OCCT. And the HDD was tested on HDTune Benchmark and Error Scan.

All of these tests were executed under very tight conditions. The notebook was idling at desktop for 15 minutes after boot prior to each test. Room temperature was at a constant 22*C. The notebook was sitting ontop of the Edova Xpad. The CPU, GPU, and northbridge have been mounted with a fresh application of MX-3 thermal paste on 08 September 2009. No other applications were running during each test.

CPU Test



The QX9300 is clocked at 2.93GHz @ 1.163v. I began the test when the ACPI temperature was idling at 36*C. OCCT CPU Linpack is an adaptation of IntelBurnTest. OCCT monitors the system for one minute prior to stressing the CPU, and monitors the system for four minutes after stressing the CPU. This CPU test stresses the CPU in 150 second sessions with a 30 second rest period inbetween each interval. 64 bit instructions is enabled.









GPU Test



The GTX 280M is clocked at 640/1600/950 @ 1.05v. The GPU is running on 186.82 WHQL drivers with the control panel settings listed below. The shader complexity in OCCT is set to eight. I began the test when the GPU core temperature rose to 36*C minutes after the GPU fan turned off at 34*C. Just like in the CPU test, OCCT monitors the system for one minute prior to stressing the GPU, and monitors the system for four minutes after stressing the GPU. However, the actual GPU stress time is a straight 20 minute session as opposed to 150 second intervals on the CPU test.





HDD Test

The HDD was stress tested non-stop with three back to back benchmark sessions and a full error scan on HDTune.



The Result

Overall, the cooling overhaul has proven to be very effective in reducing both idle and load temps. The CPU peaks at 78*C across all four cores, the GPU peaks at 69*C, and the HDD peaks at 43*C. The combination of both the copper mod and the backpanel mod has eliminated any possibilities for overheating and has opened the doors for far more aggressive overclocking. However, do keep in mind that this is only a hardware mod and one must also optimize and streamline their system on the software level. My notebook incorporates both the cooling overhaul and streamlined software to achieve the best possible performance and cooling.

I use my M570TU with the Edova Xpad as a supplement to my notebook's cooling. I think of the cooling overhaul as integrating the notebook cooler into the notebook itself, thus eliminating the "need" to carry an external notebook cooler such as the NC2000 or the Cryo LX.

Take this for example.

M570TU stock: 4kg (8.819lbs)
M570TU modded: 4.863kg (10.721lbs)

Xpad: 0.4kg (0.882lbs)
NC2000: 1.46kg (3.219lbs)
Cryo LX: 2.27kg (5.005lbs)

M570TU modded + Xpad: 5.263kg (11.603lbs)
M570TU stock + NC2000: 5.46kg (12.0372lbs)
M570TU stock + Cryo LX: 6.27kg (13.823lbs)

Not only would one get much better cooling, but also less mass and volume to carry around compared to an unmodded M570TU paired with an external notebook cooler. Also, the amount of money spent on modding the M570TU would be less, if not the same, than buying an expensive cooler such as the NC2000 or Cryo LX. From the way I see it, it would be a very wise decision to install the cooling overhaul onto any notebook.

For those who are pure performance and cooling enthusiasts, this article will give you an idea of the compromises that Clevo made and how you can restore that compromised loss. At the cost of mobility, this notebook has been transformed into one of the coolest running mobile CPU/single GPU systems on the market. It is completely self-reliant on it's own cooling, and will benefit greatly from an active notebook cooler. One can easily overclock the CPU and GPU and never need to monitor temperatures ever again. With lower temperatures, components use less power, fans will remain off or at a lower speed to save even more power, and the overall lifespan of the notebook will be lengthened.

 

(reserved space.)

You may post comments, questions, and suggestions while I finish publishing this article. However, please refrain from quoting the whole opening post as I will be constantly updating it without notice. Also, nobody likes to see quoted posts that are long with text and pictures anyway.

Update as of late 2010. I have decided to include some of my results for the W870CU cooling overhaul in this thread. More details on page 14 and onward.

 

Your cooling mod is very very effective. I didn't do a very thurough test, but I was at 86C on the GPU when you were at 67C This was with furmark with your settings, overvolt and overclock in place, on the NC2000. 86C is certainly not a bad max temperature giving the strain put on the GPU, but your temps are much lower.

IMO, the CPU cooling is even more impressive in this notebook. Been regularly playing FC2 the past week and my top QX9300 temps are in the 59-62C range (@2.93). GPU topped 68C. NC2000, 72F ambient, 630/1000/1575.

 

While maximum temperatures are important, I'm actually more interested in stabilization temperatures. I don't have an apples to apples comparison, since I have a different computer with different components, but I have noticed that simply giving out a maximum temperature doesn't always give you the best picture.

For example, when running Furmark ( my own incomplete mod, settings, and ambient here) I hit a max of 86C. 86C is the ramp-up temperature for my fans, and once they spun up I dropped to a consistent 83C. With the Zalman actually on it was 80C.

Then again, what I've noticed with the Furmark screenshots Soviet's been posting is that his max temperature is also his stabilization temperature. So I guess it's more relevant to you guys than me.

But for a hot-running notebook like the G51, the presence of copper blocks and additional ventilation (i.e. holes) is the difference between completely uncomfortable temperatures (90+) and satisfactory temperatures (<88).

Since Clevos/Sagers/whatevers run cooler in the GPU department already, it'd be really nice to see what you guys could get with half of what Soviet's done.

 

Forge, I think the sole reason as to why your G51 ramps up to a peak before settling a bit lower to equilibrium is because of the temperature threshold for your fan to kick in to max speed is 90*C if I'm not mistaken. I can't remember the exact temperature, but from my observation of you guys in the Asus forum, it's pretty close to that.

You should have played with us earlier. SACH7002 and I were tearing it up on TF2.

 

It's around 86-87 for ramp up to "medium" speed. Once I hit 90 (God forbid) then the fan goes crazy fast. But the possible registry(?) fix to that is for the Asus forum. This show is all yours.

I'm interested in the heatspreaders you put onto your memory. Google seems to think that putting heatspreaders on your laptop DIMMs will reduce room--thus airflow--on the sticks and thus actually be detrimental. Plus, looking at pictures of DIMMs--not to mention my own--it seems there is no room to put them, period.

 

saw the thread that using registry hacks can set when the fans will kick in after certain degrees....
so...that thing doesnt work?

 

Google can kiss my shiny metal *ss. Heatsinks on the RAM means all the heat would saturate into the copper and the larger surface area will dissipate it much more efficiently than just having the chips exposed. Even if there is zero CFM inside of the chassis, the heatspreaders would do their job, to an extent of course. The difference in temperature may not be noticable like night and day, but it will have the benefit of having more stable RAM, especially during FSB overclocking like I mentioned in the Asus forum.

I can say the same thing about the northbridge. Which is going to run cooler? With the heatsink? Or without? Yeah, that's how I roll.

 

Were the copper plates you put onto your HDD special ordered or just picked up in-store? The local Home Depot didn't have any.

 

I got the sheets from The Home Depot and cut them myself.

If you can't find them, then pick any of the material stores on the front page of Google. Search "copper sheets."

 

Will do. It was odd, since both of the employees I asked in both stores game me a funny look when I asked for copper.

And get your pictures up already. I'm reading the GPU description and I want to see how it looks.

(By the way, you might as well insert what thermal paste you're using, just for the sake of completeness)

 

I'll have some pictures up tonight, then most of it will be done. All there would be left to finish is new CPU and GPU tests on OCCT.

Yes, that is a good point. I'm looking at some of my older tests and I have forgotten to state what thermal paste I was using. I'll make revisions to the article when I get back.

 

GREAT work Soviet

ive seen a few pics and i must say im impressed and it's given me some ideas for the m17 i will be working on.

+ rep

 

Very cool, I would love to be able to do something like that but what did you use to to cut the copper after you bought it?? Very nice work tho. What would be really cool if someone could install a water cooling system or some crazy like that. I'm just scared I'm going to mess up my laptop trying to mod it.

 

Already ordered 40 ram copper heatsink to mod the M86.

 

There isnt much space to put them. 7-8 pieces max on the GPU and 4-5 is max on the CPU. Hope you ordered the smalles ones

 

I counted should be 13 on CPU and 2X on GPU.

 

Ok, the CPU has 3 places to put them, I only thougt of one, on top. But on the GPU the space is just to small to fit a whole heatsink on it without bumping into something.

Take a look.

 

Great job Sov. When i get the time ill be sure to make some mods to my M860. +Rep

 

I will have it all done by my posted deadline, 15:00 tomorrow, but I may have to put off the CPU and HDD tests until Monday. The ever changing weather here in San Francisco is messing with my room temperature and I'm trying not to touch my thermostat as I am trying to pinch the energy bill right now. Also, I got a buttload of research to do for my classes.

The ever important GPU test is done and I will have the OCCT graph up for you guys tonight once I get home.

 

Planning is to do this.

 

Curse you Clevo people and your large clearance.

 

how you stick the copper plate to HDD?

 

thermal tape.

Just got my copper plate today so going to put it in tomorrow for testing.

 

heh the m17x doesn't have ANY clearance because of the design


@ soviet

whats your take on copper sheets stuck to the GPU heatsink via thermal tape to increase the surface area? ?

im going to be trying it on the M17 and i was just wondering. Thanks

 

What do you want me to say? If it fits, then it's good. It's as simple as that.

As long as you Alienware people don't buy the entire stock of copper blocks from the eBay sellers and jack up the price like you did the QX9300's, I'll help you with your M17, to an extent.

 

My guess (well really Soviet's guess, but I want to take all the credit) is that while it'll nominally help--any surface increase is beneficial--it'll be so minimal it wouldn't be worth the effort.

EDIT: Damn it, Soviet.

 

Kaltmond, There isnt enough space to the top if put heatsinks on top of the GPU (over the die). You have to cut some holes on the backpanel then.

Ahh, wow, I really didnt see the space where the yellow sticker
are Just wondering, are you gonna remove the stickers or just put them on top of it?

You will see some good results on the CPU I`ll guess, because I have only one heatsink on it and it lowered the temp by ~4c.

CPU pic. haha

 

Your CPU must not feel loved.

Anyway, just sand down the shortest blocks you have and they should fit.

 

Expected completion date has been pushed to Monday. Time is hard to come by for me lately.

 

>.> couldn't you at least post your temps after the mod was done? Theres enough info out for these laptops to compare your temps to unmodified laptops.

 

Or I could have posted the entire article after I was completely finished. You should be grateful that I am typing this article openly so that you can see my progress.

Like I said, I have more work on my plate than you can possibly fathom. Getting all of it done while tending to my other obligations comes first before this. Don't worry, I'll get those temp graphs in. I'm trying to catch my room at my testing temperature of 22*C so I can properly execute the test. Ambient temperature has a major impact on component temperatures, even if it is only by 1*C. For example, I know that if my ambient temp is 20*C as opposed to 22*C, I will get around 3-4*C lower load temps.

 

meh to each their own...sounds like a over complication of something rather easy.

 

Precision is key, Rorschach. If I'm not satisfied with the testing conditions, I won't do the test and wait until the conditions meet the standard. If you are impatient, then don't post here. I really don't want to take down the entire article and repost it after I finish it.

 

I just don't see the point of using a controlled scenario when the testing could be done in just a few minutes....Using a controlled scenario is really just pointless, because no one sits around and waits for the weather to change so they can use their computer. All you have to do is turn on the a/c and record your findings, it seems so simple to me. At this rate I might as well just wait for kaltmond to post his findings.

 

I care deeply about accuracy. As an undergrad computer engineer, I am obligated to keep my work top notch and flawless with the least amount of variability as possible. Right now my room is 18.8*C, which is enough to skew my results by a considerable enough margin. If I did the CPU and GPU test now, I would be getting lower temps compared to the several Furmark tests I have done in the past under the same environmental conditions. If you see this as pointless, then get out. You are free to read my article after I post it on Monday. If you want to wait for kaltmond, then wait for him. Just know that you aren't going to be getting the level of precision and detail when he posts his results.

 

+1

Pulling numbers out of the air, if a stock M570 hits 70C on the GPU while the ambient is 18C, and Soviet's uber-modded M570 also hits 70C but while ambient is 23C, <strike>ignorant</strike> people will miss that key fact and think "Oh, Soviet did all that work for zero gain. lolololol @ him."

Building on that, it will also help people judge the effectiveness if they do modding of their own. People haveto realize that simple temperature graphs don't tell the whole story; you need ambient, too. I bet my hot-running G51 could beat Soviet's Clevo... if I was running it in an industrial freezer.

For testing precision is key, and I'm not even as focused on it as Soviet. I try to do my own testing (different computer, but still) within a range of 72F-74F, just so I can have fairly consistent results to show effective it is.

 

Well said Forge . Sov's drop in temperature will be more evident if ambient temp is constant and consistent, two key factors for an experiment like this.

 

Also, keep in mind that I am not being paid to write and publish this article. This is something I am doing on my own accord because I feel like it. If Rorschach wants me to hurry up and turn on the air conditioning or the heater to quickly get the room to 22*C, then he can Paypal my energy bill. Electricity isn't exactly cheap here in the city.

I have four tools used to monitor ambient temperature: a thermostat on my wall, a weatherclock on my desk, a temp gun, and a digital thermometer stick. I rely mostly on my weatherclock to determine room temperature, and then my thermostat to cross check.

 

hey SS, nice tf2 weekend, got a lot of hw now lol been playing video games all day yesterday (beat horde mode gears 2 1-50 in one sit through and played tf2 a lot).

on to the point, i was wondering how well the edova xpad would work? for ur computer is there really a difference with or without it? also, it serves as lapdesk, which i was gonna buy, so i think i will buy it. where did u get urs from? but yea i was wondering how effective it is cuz its either using this or buying an expensive nzxt which i would liek to avoid due to sheer size and price (parents think im crazy for buying a cooler *that makes 5C decrease*)

o yea how can u be in a room at 18.5C?? thats cold as heck for me lol...i usually keep it at 23.5C

 

It was foggy and raining in San Francisco yesterday and my condo doesn't exactly have good insulation. 19*C is still fairly normal here for this time of the year.

I did a test with the Xpad here. http://forum.notebookreview.com/showpost.php?p=5280293&postcount=714

I didn't do a control test without the cooler because I never use my notebook without the Xpad. If I were to guess, I'd say that you would see an average of 2-3*C lower on load temps on the Xpad compared to a desk. I picked it up directly from the manufacturer when they still had that promotion going on. But you can pick one up from Amazon too.

 

okay SS, i see, its pretty good, but if you could choose one, the xpad or NC2000/NZXT cryo (apparently Xoxide has NC2000 it for 32USD...) which one would u choose? i will prolly get it at newegg or something tho...

 

I would stay with my Xpad for portability sake. But if I were choosing between the NC2000 and the Cryo LX, I would go with the NC2000 simply because it is smaller and isn't too different in performance compared to the Cryo LX.

 

Active cooling will always give you better temperatures if you only look at that aspect of the cooler. But things like portability come into play, so a good passive cooler would be ideal.

The way I'm thinking of doing this, is having my Zalman anchored in my room, and the XPad for me on the go. We have different computers with different cooling, but Soviet's testing showed that the XPad had better performance than an inactive Zalman.

 

Just by looking at the layout of the Cryo versus the Zalman, it looks like the Cryo will predominantly cool the upper half of the laptop, while the Zalman is the opposite. I would think then, that the Cryo is better for laptops that have their air intakes at the upper part of the bottom, and vice versa.
At first, it then seemed to me that the Cryo would be a better choice for my D900, as the D900's fans are positioned on the upper half of the bottom - but - it (the D900) also has passive vents at the bottom half, and I think therefore I would be better served with the Zalman - the logic here being that the passive vents would benefit more from having air blowed into them than the main vents would... after all, those allready have fans.
Anything wrong with my logic here?

Oh, and pardon if this question may be a little off the thread topic.

 

ok so i think i mite stick with the nc2000. its looks good. the xpad is lite but i wont take it outta my house so it deosnt matter. i could also use the NC2000 as a lapdesk....

 

That's alright. That is a perfectly valid point, ReDuNZL. The main purpose of any active notebook cooler from a performance perspective is to force cool air into the notebook through any intakes as well as keep nearby ambient air underneath the notebook cool so that the notebook fans can suck it in. I've seen a lot of tests for the NC2000 and the Cryo LX and both seem to do their jobs exceptionally. Yes, the Cryo LX will be slightly better in performance as most notebooks have the hottest components near the rear of the notebook. However, the numbers are fairly close and, in my opinion, the minimal increase in performance does not justify the sheer difference in size and weight between the NC2000 and Cryo LX. Of course, this is really entirely up to the end-user. If the user keeps the notebook at home most of the time as a desktop replacement, then the Cryo LX would be a better choice. If the user takes it along frequently, then the NC2000 or a passive cooler will be better. And so on and so on, the possibilities are endless.

 

I have a suggestion.

Buy a 5V 45mm x 45mm x 6mm DC fan, install it internally (6mm thick so it should fit nicely). Solder it directly to the a USB port for power.

You may be able to power several of these fans from 1 usb port since each fan draws less than 0.1A and USB should be able to supply 0.5A.

Here is a link to one of these fans:
http://au.mouser.com/ProductDetail/...=sGAEpiMZZMvuXzJ9NcThaj48REXHSl5/vHDUEN2Ah28=

some specs for you:
6000 RPM
32.6 dBA
9.2 CFM

 

Moral, that seems like a great idea, but not being a electrician, how would you get around it running full blast the whole time? Those little fans make a whole lot of noise when run at those speeds.

 

Yes, I have considered installing a third fan. If I can recall, I think emike09 put that fan in his M570TU.

As much as possible, I'd prefer to avoid using any additional active cooling. I'm sure it will be effective, but the only place that I can fit that fan is over the upper right section of the GPU heatsink where the GPU MOSFET's are. I cannot put it under the CPU vent as the mounting holes on the backpanel are 34mm apart, and the posts are ~4mm tall, which eliminates that possibility. Also, there is no way of controlling the speed of the fan unless I completely rewrite the DSDT table in the system BIOS, which I am not willing to waste my time on for a tiny fan, or splice in a variable resistor with the physical control for it.