A bit of testing on the DFI, PWM IC Temps, and How to Lower Them.

[Kasparz]

I will make heatsinks for mosfets today and see, if i get some FSB more.

[Mastakilla]

i even noticed a stability difference as soon as pwmic got above 50 degrees celsius

i now have a 120mm fan at low rpm blowin at it

it sticks below 39 degrees celsius now

[tsuehpsyde]

Thanks for the guide, Kunaak. Very good read.

[nealh]

hmm..I have a p180 and run opty 170@2700 with cpu load @ 47/49C with 1.5v
my PWMIC runs 59/60C(I have had it as high as 64C with no issues when running 15hrs of dual prime 95,odd this was when my cpu was pulling air off the HS not blowing)

I have a SI 120 with fan blowing down on cpu
rear fan is exhaust, top case fan blowing in(this previous reduce PWM temps on my venice setup), front intake fan and fan blowing on chipset..all 120mm YL fans

I have Artic cooler on my X800XL exhausting heat out

I have tried to place a 120mm fan over the PWM adjacent to my cpu fan and I saw no change in my temps

Just can not seem to get any more airflow to this area

[racinjimy]

any chance of an "after" pic, I like the before pic.............

[IvanAndreevich]

cause the PWM does seem to affect stability.

Have you noticed any tangible increases? I've had PWM up to 82C with no effect on stability

[FZ1]

How about an "after mod" pic? These spots show up in the thermal pics HERE

[Hell-Fire]

Thanks for the time and effort spent Kunnak.

This is a good guide for people who either dont want to use fans, or case doesnt permit mounting fans around the PWM area.

:thumbsup:

[Hugo van Dijk]

Nice find Kunaak

Can you post some pictures of the modded board.

[ArcticOC]

Here`s how I did it..

Linkt to the project: High MB temps? Cool your MOSFET`s and SFF R60 chokes


BTW.. there is many thermal images of the board on the i-net, so if you look for hotspots I recommend this link: DFI Lanparty nF4 Ultra-D & SLI-DR Thermal Images

[ben805]

I ordered some of these copper ram sink to cool those hot spots, can't wait to get them

[fareastgq]

I want to do that to my board just for the hell of it, hehe. Good link to thermal pics. For me, looks like the pwm temps do affect stability. Fan cooling the PWM = I can dual prime at 3 gigs, pwm hotter means it just locks up or fails or becomes unstable. I suspect that mod would help me out.

[Delirious]

Great info!

I had a bunch of these laying around and i had no clue what i was goin to use them on now i know. http://www.sidewindercomputers.com/michcokit.html

[Trice]

I saw my PWM at 60C for the first time ever today while primimg my rig for 270 (RAM Stress Test) on 2 instances of P95.

[punisher69]

This is very very good info.... Thankyou soooooooooooooo much. I'm building another rig at the moment. i bought a dfi nf4 ultra-d and it has'nt even been powered up yet. now i can sink everything before it gets its first taste of power. I just hope the board is fine. I can't get the right ammount of as5 to thermal as5.(Right Ratio so they stay on the chips but are still removable) So if the board is dead it will make it a bit harder to rma with all the little heatsinks thermal attached to the mobo.. all though i don;t think manufactuarers should refuse to rma over heatsinks that can only help lower temps.

[swaaye]

Dual core brings the PWM temps way higher than single core. I run 60C+ PWM whenever I stress the CPU significantly. 70C+ is possible with Prime95.

I'm not convinced cooling them is worth the effort. Why? Cuz we've all been running boards with PWMs without even heatsinks for years and years. In fact, this DFI board is my first board to have heatsinks on them. I ran an overclocked AXP with 1.9v 2.5GHz for years thru a NF2 board with no heatsinks on the PWMs. Imagine the heat there.

How many of you never even considered that heat source until they added a temp sensor there?

I feel cooling all those chips will be about as useful as people cooling their sound cards and southbridges. In other words, no benefit at all.

[fareastgq]

Dual core brings the PWM temps way higher than single core. I run 60C+ PWM whenever I stress the CPU significantly. 70C+ is possible with Prime95.

I'm not convinced cooling them is worth the effort. Why? Cuz we've all been running boards with PWMs without even heatsinks for years and years. In fact, this DFI board is my first board to have heatsinks on them. I ran an overclocked AXP with 1.9v 2.5GHz for years thru a NF2 board with no heatsinks on the PWMs. Imagine the heat there.

How many of you never even considered that heat source until they added a temp sensor there?

I feel cooling all those chips will be about as useful as people cooling their sound cards and southbridges. In other words, no benefit at all.

You need to run your opty at 3+ gigs and more than 1.6 volts, you'll see instability really quick. AXP's are different, can't compare apples to oranges. This was the thread I was looking for again btw, hehe. Thanks Ryder.

[swaaye]

Yeah I only run 1.46v. The power demands scale non-linearly ya know. At 1.6v you're sucking some serious power thru those circuits. I watched the power load with different voltages thru my AC ammeter "Killawatt" and you are probably pulling over 300W total dual Primeing. So yah, I guess if they get hot enough they could start to fluctuate. But those ICs are designed for huge temps.

Athlon XP is only different in that it doesn't suck as much power as A64 usually. It is powered by the same components. A Pentium is powered by the same components. It's just that the demands have gone up, but so has the resiliency of the power regulation circuitry. We've gone from simple single-phase switching regulators to triple-phase and higher to reduce loads in the circuit. But if you think the temps of the PWM ICs without heatsinks on my Athlon XP board (running 2.5Ghz 1.9v) were insignificant, you are in error. A Barton at those settings draws about 110+ Watts. That's the CPU all by itself. And that board used only 2-phase power circuitry.

I'm sure I could never hit 3GHz. Heck 2.7 would probably require like 1.6v for my chip. I don't play with that part of the overclocking chart, where voltage and cooling demands spike way up to get a few MHz. I run my comp 24/7, and use CnQ, because I don't want to waste electricity. With CnQ @ 1.0v my comp sucks about 125W at the outlet. Without CnQ it would run about 170W. Primeing it sits around 250W. Prime + ATITool 3D Torture = 280W or so.

[fhpchris]

Yeah I only run 1.46v. The power demands scale non-linearly ya know. At 1.6v you're sucking some serious power thru those circuits. I watched the power load with different voltages thru my AC ammeter "Killawatt" and you are probably pulling over 300W total dual Primeing. So yah, I guess if they get hot enough they could start to fluctuate. But those ICs are designed for huge temps.

Athlon XP is only different in that it doesn't suck as much power as A64 usually. It is powered by the same components. A Pentium is powered by the same components. It's just that the demands have gone up, but so has the resiliency of the power regulation circuitry. We've gone from simple single-phase switching regulators to triple-phase and higher to reduce loads in the circuit. But if you think the temps of the PWM ICs without heatsinks on my Athlon XP board (running 2.5Ghz 1.9v) were insignificant, you are in error. A Barton at those settings draws about 110+ Watts. That's the CPU all by itself. And that board used only 2-phase power circuitry.

I'm sure I could never hit 3GHz. Heck 2.7 would probably require like 1.6v for my chip. I don't play with that part of the overclocking chart, where voltage and cooling demands spike way up to get a few MHz. I run my comp 24/7, and use CnQ, because I don't want to waste electricity. With CnQ @ 1.0v my comp sucks about 125W at the outlet. Without CnQ it would run about 170W. Primeing it sits around 250W. Prime + ATITool 3D Torture = 280W or so.

if you think 1.4V is alot, AFI runs 1.9V on his Dual core stuff!

I feel sorry for his PWMs!

[OC Detective]

My take on why PWMICs may cause instability is that when they become really hot their ability to regulate the vcore supplied to the cpu (one of their primary functions) is diminished and the standard deviation of the actual vcore supplied increases (for a given vcore the deviation moves from +/-0.02v to say +/-0.05v or more).
In the case of heavily overclocked systems with vcore at an optimised level when the vcore level fluctuates on the low side this may account for the instability? Just a theory.

[Kaiser_Sose]

I loosly placed a Delta SHE fan behind a hard drive bay blasting cold air on the board ... it lowered temps ny 25 * F

[wwwys]

Good to know these, thanks a lot