MOSFET/CPU cooling

Cooling Processors quietly

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
Zhentar
Patron of SPCR
Posts: 918
Joined: Thu Mar 27, 2003 4:00 pm
Location: Madison, WI, USA

MOSFET/CPU cooling

Post by Zhentar » Tue Aug 12, 2003 7:52 am

Okay, first a bit about my set up. I'm running a water cooled athlon XP on an Epox 8rda+; its overclocked to 2.4 ghz at 1.85 vcore right now. One thing this overclock means is that the MOSFETs, which there are 6 in a line near the CPU, get very hot, probably past 65C at least, though I haven't checked. So, out of curiosity I balanced an NMB fan over the mosfets, so they were getting decent airflow. Not surpisingly, this considerably dropped their temperature. But was suprising was that they weren't the only thing that dropped- the CPU temp reading dropped 6C. even if you aren't watercooling thats a pretty considerable drop; but it leaves the question whether or not this drop is really the CPU or just the socket thermistor; since assuming even an awful .5 C/W (and I'm probably getting at least twice that) from my cooling, the MOSFETs would have to be dumping 12W of heat into a source an inch away- not likely from something running without even a heatsink.

So what do you guys think; is it just affecting the reading and not the actual temp, and if so, is it making the sensor more or less accurate? a drop from 42C to 36C is quite a difference.

halcyon
Patron of SPCR
Posts: 1115
Joined: Wed Mar 26, 2003 3:52 am
Location: EU

Post by halcyon » Tue Aug 12, 2003 9:43 am

With XP 1800+ (thoroughbred core, I assume) you should be measuring using the on-chip sensor, not the socket sensor which can be so much off to be practically useless :)

However I still wouldn't be too surprised if at least half of that drop was a real drop.

regards,
Halcyon

dukla2000
*Lifetime Patron*
Posts: 1465
Joined: Sun Mar 09, 2003 12:27 pm
Location: Reading.England.EU

Post by dukla2000 » Tue Aug 12, 2003 9:51 am

I would also suspect there is room for a genuine temp drop. Without the fan you only have the water block and convection removing heat from the CPU. With the fan you most likely have more forced air around the socket but especially in the gap between the waterblock and the CPU substrate (and in particular the CPU die sides). OK, pretty low airflow but conceivable it is better than the original convection to cause a measurable drop.

(I remember way back someone getting a big drop by aiming a fan at the back of the motherboard. Plenty of debate about the prime beneficiary being the thermistor rather than the CPU itself, but fact some heat comes out the bottom of the CPU substrate and gets trapped in the socket, so if you cool the socket there should be a genuine effect on the die itself.)
[size=75]2009/Oct: Jetway JNC81-LF * 4850e naked under fanless Xigmatek Apache * Antec mini Skeleton w/Nexus 120mm PWM fan * Delta 90W brick w/Skeleton DC-DC board * WD2500BEVT 250Gb blue[/size]

Zhentar
Patron of SPCR
Posts: 918
Joined: Thu Mar 27, 2003 4:00 pm
Location: Madison, WI, USA

Post by Zhentar » Tue Aug 12, 2003 10:43 am

Its a t-bred b. according to the manual the 8rda+ uses the CPU diode for the thermal shutdown stuff, but uses an in socket thermistor for the temp readings. Maybe sometime soon I'll get bored and set up a diode reader, then I'll be able to test for sure.

I stuck a length of u-channel aluminum on the MOSFETs but its still very hot, I'm trying to figure out what I can do to improve the 'heatsink'.

One thing for sure though is it can't be hurting things.

SometimesWarrior
Patron of SPCR
Posts: 700
Joined: Thu Mar 13, 2003 2:38 pm
Location: California, US
Contact:

Post by SometimesWarrior » Tue Aug 12, 2003 11:38 am

Ever since I flipped the fan on my two CPUs and ducted their exhausts (I'll try to do a write-up by the end of the month, I'm proud of all the mods I've done :P), my VRM temperatures have shot through the roof. My motherboard monitors VRM temps, and now the VRM's for one of my processors never go below 70C. If I forget to turn up the case fans before gaming, they go up to a scorching 87C. I believe it--I've touched them when they were up around 65C, and they were definitely hot.

I've got an old FOP32 heatsink (150g of aluminum, I'd guess) which I lapped and am prepared to hacksaw. It should fit on top of the VRM's after a bit of cutting. That should do a lot more good than those dinky U-channels... I tried peeling the ram sinks off my vidcard and sticking them on the VRM's, but that did absolutely nothing to help.

Edit: fixed some of my grammar mistakes. ;)

Jan Kivar
Friend of SPCR
Posts: 1310
Joined: Mon Apr 28, 2003 4:37 am
Location: Finland

Post by Jan Kivar » Tue Aug 12, 2003 9:01 pm

I've heard about this before. Some socket thermistors are affected by the airflow of the CPU fan. I doubt that the fan would actually lower the CPU temp anymore.

If You want to play more (and void your warranty) You could buy some RAM sinks and glue them to the VRM chips.

Cheers,

Jan
[size=75]E6600 (2GHz@1,163V, 3GHz@1,237V), Ninja rev. B, AB9 Pro, 4x1 GB Corsair (CM2X1024-6400), MSI 7600GS, Samsung HD501LJ/HD401LJ/SP2504C, Plextor PX760SA, Seasonic S12-430, Nexus Fans, Antec Solo, ViewSonic VP201b[/size]

halcyon
Patron of SPCR
Posts: 1115
Joined: Wed Mar 26, 2003 3:52 am
Location: EU

Post by halcyon » Tue Aug 12, 2003 11:00 pm

Zhentar,

if you are using a w32 platform, try MBM5 with Winbond2 chip. AFAIK, that should give you a temp reading off the on-chip die of the CPU and be more accurate than the socket sensor (W83697HF).

http://mbm.livewiredev.com/comp/epox.html

However, I have not used this board myself, so if you see negative values or way too high values, then the sensor selection is probably incorrect.

friendly regards,
Halcyon

bluehat
Posts: 84
Joined: Thu Sep 19, 2002 3:47 am

Post by bluehat » Fri Aug 15, 2003 4:27 am

Zhentar wrote: I stuck a length of u-channel aluminum on the MOSFETs but its still very hot, I'm trying to figure out what I can do to improve the 'heatsink'.
Since you are water cooling, why not build a mosfet block.

I glued a 1x4x3 cm heatsink to each pair of mosfets, and still they are 65 C (measured from the heatsink) with 1,65V vcore, and much less after undervolting.

nbac
Patron of SPCR
Posts: 142
Joined: Sat Jul 05, 2003 11:27 am
Location: Sweden

Post by nbac » Sun Aug 17, 2003 1:29 am

IMO, cooling parts that don't need extra cooling (like the VRM) is usually a waste of
time (and noise). If your Vcore is stable then everything is ok! Unlike your CPU,
MOSFET’s are designed to operate at high temperatures.
Maximum junction temp: 120-150'C!

It's just a piece of silicon and plastic and it doesn't care about what
is hot for the human hand or what is not. (Sorry, english isn't my native language...)

Post Reply