Case Study: TA790GX 128M MOSFET & Choke Temperatures w/ MOS-C1

[streetfighter 2]

Disclaimer: This is an experiment I performed and am providing a write-up in the hopes that useful information can be gleaned from it by other users. (It sure did help me!) Please do your own research before making potentially dangerous decisions.

This article assumes readers have prior familiarity with certain technical terms. If you aren't comfortable with these terms or would like to freshen up I recommend the following as a primer:
http://www.hardwaresecrets.com/arti...e-Motherboard-Voltage-Regulator-Circuit/616/1

Purpose:
The Biostar TA790GX series boards are notorious for problems with the CPU’s PWM circuit due to the lack of any stock cooling apparatus. In most cases when the board is run with stock clocks and typical ambient temperatures (0C) there should be no serious stability issues. However, if the board is pushed in the presence of an already high TDP CPU there is a serious risk of critical system failure such as in Figure 1.

Figure 1: MOSFET which caught fire as a result of poor cooling and high stress. Sorry for the picture quality. (Source: zif33rs, rebelshavenforum.com)

The TA790GX 128M uses a 3+1 phase PWM circuit with three MOSFETs per phase which dictates two important facts. Firstly the PWM circuit is not the most sophisticated or the best in providing stable power to the CPU. Secondly the CPU’s voltage regulation is accomplished by only a small handful of components which translates to large amounts of heat in a concentrated area. The heat is not merely relegated to the MOSFETs, chokes can also be a substantial contributor to heat as shown in Figure 2, but are less likely to explode or catch fire (citation needed).

Figure 2: Coil type inductors shown under load exceeding the temperatures of the MOSFETs. (Source: bing, overclockers.com)

The common solution to this issue is aftermarket cooling such as the Thermalright HR-09S, Enzotech MOS-C1 and custom modified heatsinks. Several of these methods are shown in Figure 3. Although clearly better than nothing at all, the efficiency of the various aftermarket sinks does not appear to be well studied. Unfortunately I do not have access to many of the solutions so I will only be testing one of them, the Enzotech MOS-C1.

Figure 3: A variety of MOSFET heatsink designs for TA790GX series boards. From upper left going clockwise: official Biostar heatsink only available in China (source needed), modified Biostar Cooler Harbor (source), Thermalright HR-09S (source), modified aluminum heatsink (source)

Test System:
Motherboard: Biostar TA790GX 128M with Enzotech MOS-C1 MOSFET heatsinks
CPU: AMD Phenom II X2 550BE unlocked and running as an X4 B50 at 3.7GHz 1.41V (1.39V effective)
CPU MOSFETs: NTD4863N made in week 43 of 2008 (datasheet)

Procedure:
The test was divided into two stages, the first stage involved letting the computer heat up while recording the MOSFET temperature. The second stage involved measuring the temperatures of all the MOSFETs and chokes while at max load with all components fully heated.

To measure the temperature of the MOSFETs and chokes on the board a K type exposed junction thermocouple was placed in metal-to-metal contact with the top center of the MOS-C1 heatsink on the MOSFET closest to the rear I/O panel’s ethernet port as shown in Figure 5. (This particular MOSFET was chosen for the warm-up stage of the test because it is the same as the one which exploded in Figure 1.) Prime95 x64 was then run with 4 threads and in-place large FFTs for 40 minutes with the temperature recorded each minute which is shown in Figure 6. An Extech 22-816 multimeter collected temperatures from the MOSFET’s thermocouple while another thermal probe placed just outside of the case provided actual ambient temperature measurements as shown in Figure 4. Since the door of the case was open during the test the air conditioning was shut off in an attempt to replicate the environment of the case if the door was closed.

Figure 4: Test setup. The ambient temperature sensor is visible as the grey wire just in front of the GPU towards the right. Yes, I noticed my desk is dusty, I’m getting a new one anyway.

Figure 5: Exposed junction thermocouple wedged into MOS-C1 making direct contact with the heatsink surface.

Figure 6: Recording data into a spreadsheet during the test.

After the 40 minute warm up period elapsed and while Prime95 was still running, the thermocouple was placed on most of the MOSFETs and left until an accurate reading could be taken. Finally the thermocouple was placed on the four chokes and measurements were taken.

Results:

Figure 7: The warm-up stage of the experiment. The measured MOSFET reached a maximum temperature of 72C.

The MOSFET measured during the warm-up stage, labeled by a red circle in Figure 8, took around 8 minutes to reach a relatively stable temperature whereas the CPU took approximately 6 minutes and the NB (not shown) crawled from 41C to 47C and apparently would have kept increasing even after the test time elapsed. The case’s ambient temperature stayed fairly constant during the test and averaged around 27C, which, with any luck, is hotter than most users’ computer cases.

Figure 8: PWM section of the TA790GX 128M where the MOSFET measured in stage one is marked by the red circle; the MOSFETs with the white circles were the hottest; the MOSFETs labeled 7, 8 and 9 correspond to data points 7, 8, 9 in Figure 9; the chokes are labeled as 1, 2, 3 and 4.

The average temperature of the MOSFET in Figure 7 was 68C after the first seven minutes which, while high, is not enough to cause alarm. Prime95 is not considered a good measure of typical processor loads so this particular MOSFET lead me to the early conclusion that MOSFET temperatures were more than acceptable.

Unfortunately during the second stage of the test when I examined the temperatures of the other MOSFETs I found that the MOSFET measured during the first stage was relatively cool. A random sampling of the other MOSFET temperatures is shown in Figure 9.

Figure 9: Random sampling of CPU MOSFET temperatures at full load after 40 minutes.

One of the MOSFETs reached a horrifying 86C. Though the maximum junction temperature of the MOSFETs is listed as 175C, the performance degrades long before that and the temperature measured during the experiment was only the temperature of the MOSFET heatsink, which is sure to be lower than the actual junction temperature (source). Several of the high temperatures recorded in Figure 9 are cause for alarm, but only in regards to stability. I sincerely doubt that these temperatures could cause a MOSFET to explode or catch fire.

Even more interesting was the discovery that the phases were unevenly loaded. The MOSFET temperatures corresponding to the phases of chokes 1 and 4 in Figure 8 were up to 36C cooler than the hottest recorded MOSFET temperature. This unexpected result can be partly attributed to the fact that one of the phases is the sole provider of a separate voltage for the CPUs memory controller, which at the time, was not fully loaded. From the data recorded I would assume that the phase corresponding to choke 1 provided this extra voltage because it had the lowest average temperature of all the other phases. The data in Figure 9 correlates well with the choke temperatures shown in Figure 10, where cooler MOSFETs aligned with cooler chokes. Oddly though the hottest MOSFET temperature (86C) was recorded on one of choke 3’s MOSFETs, not choke 2 as would be expected from the data in Figure 10. I couldn’t retrieve a datasheet for the chokes that were tested; however I feel that 70C is still a safe temperature.

Figure 10: Choke temperatures measured after 40 minutes at full load. The choke number corresponds to the labeling in Figure 8.

From this data it’s obvious that the PWM circuit on the TA790GX 128M is hot, very hot, probably even dangerously hot. On the other hand the MOS-C1 heatsinks appear to be doing their job. A comparison of results with and without the MOSFET heatsinks would be preferential but I was unable to remove the MOS-C1s (which are stuck on there really good). I’m going to place a fan nearby blowing directly on them and rerun the test to see how much that improves the temperatures.

Special thanks to TPU users Velvet Wafer and Wile E for inspiring the test.

Notes: If I made any errors (logic, textual or otherwise) please feel free to mention them so they can be corrected. Thanks

 

[Velvet Wafer]

Disclaimer: This is an experiment I performed and am providing a write-up in the hopes that useful information can be gleaned from it by other users. (It sure did help me!) Please do your own research before making potentially dangerous decisions.

Purpose:
The Biostar TA790GX series boards are notorious for problems with the CPU’s PWM circuit due to the lack of any stock cooling apparatus. In most cases when the board is run with stock clocks and typical ambient temperatures (0C) there should be no serious stability issues. However, if the board is pushed in the presence of an already high TDP CPU there is a serious risk of critical system failure such as in Figure 1.

http://img.techpowerup.org/100801/100_0747-1.jpg
Figure 1: MOSFET which caught fire as a result of poor cooling and high stress. Sorry for the picture quality. (Source: zif33rs, rebelshavenforum.com)

The TA790GX 128M uses a 3+1 phase PWM circuit with three MOSFETs per phase which dictates two important facts. Firstly the PWM circuit is not the most sophisticated or the best in providing stable power to the CPU. Secondly the CPU’s voltage regulation is accomplished by only a small handful of components which translates to large amounts of heat in a concentrated area. The heat is not merely relegated to the MOSFETs, chokes can also be a substantial contributor to heat as shown in Figure 2, but are less likely to explode or catch fire (citation needed).

http://img.techpowerup.org/100801/PWM IR.jpg
Figure 2: Coil type inductors shown under load exceeding the temperatures of the MOSFETs. (Source: bing, overclockers.com)

The common solution to this issue is aftermarket cooling such as the Thermalright HR-09S, Enzotech MOS-C1 and custom modified heatsinks. Several of these methods are shown in Figure 3. Although clearly better than nothing at all, the efficiency of the various aftermarket sinks does not appear to be well studied. Unfortunately I do not have access to many of the solutions so I will only be testing one of them, the Enzotech MOS-C1.

http://img.techpowerup.org/100801/aftermarket_fet_sinks.png
Figure 3: A variety of MOSFET heatsink designs for TA790GX series boards. From upper left going clockwise: official Biostar heatsink only available in China (source needed), modified Biostar Cooler Harbor (source), Thermalright HR-09S (source), modified aluminum heatsink (source)

Test System:
Motherboard: Biostar TA790GX 128M with Enzotech MOS-C1 MOSFET heatsinks
CPU: AMD Phenom II X2 550BE unlocked and running as an X4 B50 at 3.7GHz 1.41V (1.39V effective)
CPU MOSFETs: NTD4863N made in week 43 of 2008 (datasheet)

Procedure:
The test was divided into two stages, the first stage involved letting the computer heat up while recording the MOSFET temperature. The second stage involved measuring the temperatures of all the MOSFETs and chokes while at max load with all components fully heated.

To measure the temperature of the MOSFETs and chokes on the board a K type exposed junction thermocouple was placed in metal-to-metal contact with the top center of the MOS-C1 heatsink on the MOSFET closest to the rear I/O panel’s ethernet port as shown in Figure 5. (This particular MOSFET was chosen for the warm-up stage of the test because it is the same as the one which exploded in Figure 1.) Prime95 x64 was then run with 4 threads and in-place large FFTs for 40 minutes with the temperature recorded each minute which is shown in Figure 6. An Extech 22-816 multimeter collected temperatures from the MOSFET’s thermocouple while another thermal probe placed just outside of the case provided actual ambient temperature measurements as shown in Figure 4. Since the door of the case was open during the test the air conditioning was shut off in an attempt to replicate the environment of the case if the door was closed.

http://img.techpowerup.org/100801/test_setup.jpg
Figure 4: Test setup. The ambient temperature sensor is visible as the grey wire just in front of the GPU towards the right. Yes, I noticed my desk is dusty, I’m getting a new one anyway.

http://img.techpowerup.org/100801/thermocouple.jpg
Figure 5: Exposed junction thermocouple wedged into MOS-C1 making direct contact with the heatsink surface.

http://img.techpowerup.org/100801/test_setup_desktop.jpg
Figure 6: Recording data into a spreadsheet during the test.

After the 40 minute warm up period elapsed and while Prime95 was still running, the thermocouple was placed on most of the MOSFETs and left until an accurate reading could be taken. Finally the thermocouple was placed on the four chokes and measurements were taken.

Results:
http://img.techpowerup.org/100801/graphed_data.png
Figure 7: The warm-up stage of the experiment. The measured MOSFET reached a maximum temperature of 72C.

The MOSFET measured during the warm-up stage, labeled by a red circle in Figure 8, took around 8 minutes to reach a relatively stable temperature whereas the CPU took approximately 6 minutes and the NB (not shown) seemed to slowly crawl from 41C to 47C and seemingly would have kept increasing even after the test time elapsed. The case’s ambient temperature during the experiment stayed roughly even during the test and averaged about 27C, which is hopefully warmer than most users’ computer cases.

http://img.techpowerup.org/100801/labeled_PWM.jpg
Figure 8: PWM section of the TA790GX 128M, where the MOSFET measured in stage one is marked by the red circle, the MOSFETs with the white circles were the hottest, the MOSFETs labeled 7, 8 and 9 correspond to data points 7, 8, 9 in Figure 9, and the chokes are labeled as 1, 2, 3 and 4.

The average temperature of the MOSFET after the 7 minute mark was 68C, which while high, is not enough to cause alarm. Prime95 is not usually considered a good measure of typical computer usage so this particular MOSFET lead me to the early conclusion that MOSFET temperatures were more than acceptable.

Unfortunately during the second stage of the test when I examined the temperatures of the other MOSFETs I found that the MOSFET measured during the first stage was relatively cool. A random sampling of the other MOSFET temperatures is shown in Figure 9.

http://img.techpowerup.org/100801/random_sampling_labeled2.jpg
Figure 9: Random sampling of CPU MOSFET temperatures at full load after 40 minutes.

One of the MOSFETs reached a horrifying 86C. Though the maximum junction temperature of the MOSFETs is listed as 175C, the performance degrades long before that and the temperature measured during the experiment was only the temperature of the MOSFET heatsink, which is sure to be lower than the actual junction temperature (source). Though these high temperatures are cause for some alarm, it is only in regards to stability. I sincerely doubt that these temperatures could cause a MOSFET to explode or catch fire.

Even more interesting was that the phases seemed to be unevenly loaded. The MOSFET temperatures corresponding to the phases of chokes 1 and 4 in Figure 8 were up to 36C cooler than the hottest recorded MOSFET temperature. This is partially accounted for by the fact that one of the phases is the sole provider of a separate voltage for the CPUs memory controller (probably choke 1), which at the time, was not fully loaded (though in hindsight it probably should have been). This data also corresponded with the choke temperatures shown in Figure 10, where the cooler MOSFETs aligned with the cooler chokes. Oddly though the hottest MOSFET temperature (86C) was recorded on one of choke 3’s MOSFETs, not choke 2 as would be expected from the data in Figure 10. I couldn’t retrieve a datasheet for the chokes that were tested; however I feel that 70C is still a safe temperature.

http://img.techpowerup.org/100801/choke_temps_labeled2.jpg
Figure 10: Choke temperatures measured after 40 minutes at full load. The choke number corresponds to the labeling in Figure 8.

From this data it’s obvious that the PWM circuit on the TA790GX 128M is hot, very hot, perhaps even dangerously hot. On the other hand the MOS-C1 heatsinks appear to be doing their job. A comparison of results with and without the MOSFET heatsinks would obviously be preferential but I was unable to remove the MOS-C1s which are stuck on there really good! I’m going to place a fan nearby blowing directly on them and rerun the test to see how much that improves the temperatures.

Special thanks to techPowerUp users Velvet Wafer and Wile E for inspiring the test.

Notes: If I made any errors (logic, textual or otherwise) please feel free to mention them so they can be corrected. Thanks

your test results seem promising, and nearly 90c is very,very at the edge... but still acceptable. I should warn you again,tho, that this lonely electrolytic cap will die first, because it cant stand the temperature overtime. especially hot summers, and overheated rooms, are the 2 things, that are as dangerous to the FET, as cancer is for a Human
Thanks for mentioning me,also

 

[streetfighter 2]

your test results seem promising, and nearly 90c is very,very at the edge... but still acceptable. I should warn you again,tho, that this lonely electrolytic cap will die first, because it cant stand the temperature overtime. especially hot summers, and overheated rooms, are the 2 things, that are as dangerous to the FET, as cancer is for a Human
Thanks for mentioning me,also

Actually I'm willing to bet that the MOSFET which I measured at 86C did get even hotter, perhaps >90C. In Figure 7 there are large sudden fluctuations in the MOSFET temperature. This is certainly an eye opener for me and the reason why I'll be conducting the experiment again but with a dedicated fan blowing on the MOSFETs. I'll also be placing some old VGA ramsinks on the chokes.

I don't think the temperature will kill the the electrolytic cap even though I admittedly didn't measure it (but I will after I make the cooling improvements). I would wager a guess that as the FETs heat up and eventually malfunction a sudden surge in current or voltage could cause the cap to blow.

 

[Velvet Wafer]

Actually I'm willing to bet that the MOSFET which I measured at 86C did get even hotter, perhaps >90C. In Figure 7 there are large sudden fluctuations in the MOSFET temperature. This is certainly an eye opener for me and the reason why I'll be conducting the experiment again but with a dedicated fan blowing on the MOSFETs. I'll also be placing some old VGA ramsinks on the chokes.

I don't think the temperature will kill the the electrolytic cap even though I admittedly didn't measure it (but I will after I make the cooling improvements). I would wager a guess that as the FETs heat up and eventually malfunction a sudden surge in current or voltage could cause the cap to blow.

I said, thats the crappiest PWM i ever worked with!140w is to be seen as maximum here, not standard

can be that a voltage spike killed it, but mine blew relatively exotic
not the top broke, where its supposed to be.... somehow the sidewall of the cap has come off partly, there were even very small spurts of electrolyt, around the cap, on the board. i have just pictured it for you

 

[blkhogan]

Very interesting. Ive been a Biostar board user for quite sometime, actually had this exact board not to long ago. I rocked it pretty hard during its life with me with no problems. I always had a high flow 120mm fan mounted on a bracket that I made blowing on the area you are describing. With out air flow directly on it, it was impossible to hold your finger on it for more than a second or so without having to remove it quickly.

 

[Velvet Wafer]

it still runs with that damaged cap, but you get 15-20% worser OCs without it
also, i can totally agree with you, regarding the touching part, thats not only an ungood feeling, its totally impossible, like touching a medium "warmed" cooking plate

 

[p_o_s_pc]

i owned a board like this and put one of those long ram sinks on the mosfets and it did fine for me. But when doing high voltage OCs for bench runs i would point a high flow 80mm fan on the heatsink also. I didn't have the mosfets burn out but i have the 2 12v lines on the 24pin burn out. (turned completely black and melted the plastic on the 24pin(PSU side)

 

[Wile E]

I said, thats the crappiest PWM i ever worked with!140w is to be seen as maximum here, not standard

can be that a voltage spike killed it, but mine blew relatively exotic
not the top broke, where its supposed to be.... somehow the sidewall of the cap has come off partly, there were even very small spurts of electrolyt, around the cap, on the board. i have just pictured it for you
http://img818.imageshack.us/img818/7116/p1020455.jpg

That's not a heat related failure, that's a blow out. Recap it and see if the replacement cap blows.

 

[Velvet Wafer]

That's not a heat related failure, that's a blow out. Recap it and see if the replacement cap blows.

if i had a soldering iron with more than 16w, i would love to do so!
may i ask, why you are suggesting a "blowout"? by that, do you mean internal shorting of the board? i only know, that electrolytic caps blow, when you reverse polarity, and have tried it a few times with spare caps
im sorry, but i must say, i have never heard the term Blow-out, especially not in this conect i believe we need some enlightenment here!

Honestly, up until now, i suggested heat as the problem.... i did that, because, since i run no quad in it anymore, it started to behave MUCH better, but if i clock too high im getting my myterious BSOD 0x00000124, again, the same that always appeared when i used the 955 permanently for crunching on that board

 

[Wile E]

It exploded quickly, not slowly failed due to heat. That something more along the lines of what a short would do.

 

[Velvet Wafer]

It exploded quickly, not slowly failed due to heat. That something more along the lines of what a short would do.

wouldnt render that the entire board... bricked?
The Cap may have been exploded, when i was asleep, and the rig crunching, but even after that, not much of his behaviour changed, except this daily random 0x00000124 bsod (mostly USB related, the SB (NB?) seems to make problems here
you may are right, but because i only used the aluramsinks from an accelero to cool the Fets, that cap may get a little hot... look at its proximity, its surrounded by FETs, Chokes, and... solid caps. i would say it was a case of "saving money on the wrong end"
built a cpu PWM, just supplied by solid caps, as usual
and then fuck it up, by using one of the cheapest, crappiest OST caps, for the NB, just a few cm away from the real heat

 

[Wile E]

Heat may have contributed to the failure, but I've never seen a cap fail in that manner unless it was due to a short or surge or some similar event. You'd have to hold it to a direct flame to get it to do that with just heat.

 

[AsRock]

Heat may have contributed to the failure, but I've never seen a cap fail in that manner unless it was due to a short or surge or some similar event. You'd have to hold it to a direct flame to get it to do that with just heat.

Kinda like if one of the legs in side shorted with the casing.

 

[Mussels]

interesting read.

 

[Velvet Wafer]

Heat may have contributed to the failure, but I've never seen a cap fail in that manner unless it was due to a short or surge or some similar event. You'd have to hold it to a direct flame to get it to do that with just heat.

a surge would be more logical, i believe, with a totally overloaded PWM in the Background, am i right? especially, since its only a 6.3v cap! From what i know now, i will never overvolt a PH2 on that board again undervolting ist a must, in this case

 

[Mussels]

PLEASE, if you want to host big files use TPU.org - imageshack is slow as shit for us non americans

 

[Velvet Wafer]

PLEASE, if you want to host big files use TPU.org - imageshack is slow as shit for us non americans

im european, but okay^^ it was just that one pic, i wanted to show! sounds logical that australians wouldnt have the best connection. also thanks, i never knew there was a direct uploading page from TPU!

 

[Wile E]

a surge would be more logical, i believe, with a totally overloaded PWM in the Background, am i right? especially, since its only a 6.3v cap! From what i know now, i will never overvolt a PH2 on that board again undervolting ist a must, in this case

Considering the cpu only takes 1.xxx volts, 6.3v caps are plenty overkill in that respect.

 

[streetfighter 2]

Considering the cpu only takes 1.xxx volts, 6.3v caps are plenty overkill in that respect.

How can we even be sure that the capacitor in question is part of the CPU PWM circuit? Seeing as how it's close to two other MOSFETs which are clearly not in the CPU's PWM circuit it is just as likely that it is for the USB/DVI/ethernet which would mean it is >2V. Then again, I really have no clue . . .

im european, but okay^^ it was just that one pic, i wanted to show! sounds logical that australians wouldnt have the best connection. also thanks, i never knew there was a direct uploading page from TPU!

http://www.techpowerup.org/upload.php

They sure don't advertise it well! Nevertheless if you google "TechPowerUp image host" it's the first result.

 

[Wile E]

How can we even be sure that the capacitor in question is part of the CPU PWM circuit? Seeing as how it's close to two other MOSFETs which are clearly not in the CPU's PWM circuit it is just as likely that it is for the USB/DVI/ethernet which would mean it is >2V. Then again, I really have no clue . . .

Good point. I didn't consider that at all. lol.

 

[Velvet Wafer]

Considering the cpu only takes 1.xxx volts, 6.3v caps are plenty overkill in that respect.

this cap had nothing to do with the cpu, it was related to some USB functions (USB related bsods), maybe it was for the NB...also, as you suggested a surge, we both dont know, how high that surge may was
also, the onboard VGA needs an extra molex for beeing able to power up, so i think this cap may also supply it partly, as well

How can we even be sure that the capacitor in question is part of the CPU PWM circuit? Seeing as how it's close to two other MOSFETs which are clearly not in the CPU's PWM circuit it is just as likely that it is for the USB/DVI/ethernet which would mean it is >2V. Then again, I really have no clue . . .



http://www.techpowerup.org/upload.php

They sure don't advertise it well! Nevertheless if you google "TechPowerUp image host" it's the first result.

damn, you were first! thanks for the linky!

 

[streetfighter 2]

I pulled a heatsink off an old Nvidia southbridge chip and I'm using a hacksaw to cut it so I can fit it on my mobo's chokes. I was looking around at thermal adhesives and I find the selection in the consumer market to be disappointing. The only choice you have is really silver epoxy (Arctic Silver ASTA-7G, MG 8331 Silver Conductive Epoxy). While that stuff is really fantastic (I've been using the MG 8331 for years and it is really amazing) it has one serious flaw when used as a pure thermal interface, it's incredibly electrically conductive. I would prefer a thermal adhesive that is thermally conductive but not electrically conductive like 3M 8815 Thermal Tape, but it would appear that no one sells that in small quantities. I'll keep looking though but if anyone knows of a material that's available in small batches for reasonable prices I'd love to hear about it.

EDIT:
ROFL, I searched for "ceramique thermal tape" in google and 2"x2" 3M thermal tape for $2.50 popped up at frozencpu.com
http://www.frozencpu.com/products/1...nsfer_Tape_-_2_x_2.html?tl=g8c127&id=6J4IsgUj
Hopefully I'm done being an idiot for the day (but probably not).

 

[streetfighter 2]

I added heatsinks to the CPU's chokes and added a fan blowing directly onto the CPU's PWM circuit. Unfortunately this weekend is going to be incredibly busy for me so I probably won't get a chance to do a thorough test but some preliminary data I got from running Prime95 for 10 minutes is absolutely astonishing.

Comparing my old data at 10 minutes with my new data at 10 minutes (with the same ambient temps) the measured MOSFET was an unbelievable 13C cooler with the addition of a fan! The chokes with the new heatsinks were all measured around 38C, which is incredibly short of the 70C max they achieved without a fan and heatsink.

This data isn't conclusive but I am very optimistic and looking forward to running the experiment again to quantify the improvement. (Unfortunately the placement of the fan is going to make it much more difficult to measure the MOSFETs the second time around...)

Here's a teaser image I just took with the computer running. The exposure time is 1/30th of a second which is why it looks as if the fan is stationary.

 

[Mussels]

I added heatsinks to the CPU's chokes and added a fan blowing directly onto the CPU's PWM circuit. Unfortunately this weekend is going to be incredibly busy for me so I probably won't get a chance to do a thorough test but some preliminary data I got from running Prime95 for 10 minutes is absolutely astonishing.

Comparing my old data at 10 minutes with my new data at 10 minutes (with the same ambient temps) the measured MOSFET was an unbelievable 13C cooler with the addition of a fan! The chokes with the new heatsinks were all measured around 38C, which is incredibly short of the 70C max they achieved without a fan and heatsink.

This data isn't conclusive but I am very optimistic and looking forward to running the experiment again to quantify the improvement. (Unfortunately the placement of the fan is going to make it much more difficult to measure the MOSFETs the second time around...)

Here's a teaser image I just took with the computer running. The exposure time is 1/30th of a second which is why it looks as if the fan is stationary.
http://img.techpowerup.org/100807/mosfet_fan.jpg

its not much of a surprise that the fan would get the temps way lower, really.

passive cooling vs active cooling is no contest, even small amounts of airflow will rapidly lower temperatures.

 

[newtekie1]

I figured I'd throw up some pictures of my eVGA board and what happens when you use a 3 Phase power system on a 775 motherboard and put a 95w Pentium D 805 in it:

And this board is supposed to be able to handle 130w Core 2 Extreme processors... Surprisingly the board still works and is totally stable. But I won't run it again until I get some mosfet coolers on it.:shadedshu