High PWM temps with Abit IP35 Pro or other mobo?

[Hauk]

You may recall the thread reporting surprisingly high PWM temps using Quads on some Abit P35 boards. The question was posed, should we consider PWM temps into the cooling equation? We monitor NB and CPU temps and spend big bucks on exotic cooling solutions, but we typically leave mofset cooling stock. Many motherboard failures are attributed to power components failure.

Prior to my build, I removed the heatsink from my Abit IP35 Pro and found very poor thermal contact due to a slightly bent heatpipe and inadequate retention pressure from the plastic push-pins. I straightened and reapplied the assembly using the stock mofset thermal pad and push pins. Testing a Q6600 @ 334 x 9 with dual instances of Orthos caused PWM temps to spike to 75-76 degrees, and that was in a well ventilated case.

My goal was to reduce PWM temps using better thermal material, a slight increase in retention pressure, and active cooling at the Mofset. I replaced the push-pins with bolts, used Arcitc Silver Ceramique on the mofsets, and installed a 40mm fan.

I made a list for anyone interested in the bolt mod. It's not terribly difficult, but it can be a pain. There's very little clearance on the nut side. I intended on using a nut driver to hold the nuts in place but couldn't. I had to use a pair of needle nose pliers to hold them as I tightened the scews.

Bolts were necessary because the one-piece heatsink assembly wouldn't "lay flat" with push-pin retention strength. Bolts may not be necessary in your case. Simply using better thermal paste and a fan may help. Bolts this small are hard to find locally. I found them online:


PARTS DETAIL:

6ea Part # 5313, 4-40 x 5/8 18-8 Stainless Machine Screw
Bolt Alternative, Part # 9623, 4-40 x 9/16 is slightly shorter
6ea Part # 4107, 4-40 18-8 Stainless Hex Lock Nut Nylon Insert
6ea #4 x 1/4" fiber washers from Home Depot
Arctic Silver Creamique or other non-conductive paste
40mm fan



PICS DETAIL:

Bolts at mofset
Bolts at the NB
Bolts at the SB
Underside- note washers are just the right size
Bolted down Mofset
Bolted down NB
Bolted down SB
40mm fan- Note zip tie, one is all I used, but fan is snug against Ultra 120x base and is secure

I didn't get scientific on torque; I just knew not to overtighten and made sure I had the same number of threads for each respective screw. Place mobo on a flat surface with scews installed, apply thermal interface, apply heatsink, apply nuts and hand tighten. Use best method to hold nuts while screwing in screws. Don't overtighten! After you see the screw head has tightened onto the washer and the nut is flush, you're just about there.



RESULTS:
Q6600 @ 334 x 9, 1.362 in bios, NB @ 1.29, CPU VTT @ 1.23. Orthos dual instance small FFT's for 30 minutes. Latest version of uGuru.

Stock motherboard idle:
CPU = 36
System = 36
PWM = 52

Stock motherboard load:
CPU = 55
System = 37
PWM = 75


Modded motherboard idle:
CPU = 36
System = 42
PWM = 43

Modded motherboard load:
CPU = 55
System = 43
PWM = 60

PWM idle temp dropped 9 degees. PWM load temp dropped 15 degrees! Note however my northbridge temps increased 6 degrees. I attribute this to improved thermal exchage at the mofset. Heatsinks are connected with a heatpipe. The heatpipe is either carrying more heat away from the mofset heatsink, or allowing less heat to come from the northbridge. In any case, northbridge temps are still quite reasonable. Was it worth it? Depends on who you ask. I didn't hope for a higher overclock and am happy where I'm at. I do like to run components as cool as possible however. So for me it was worth it. Wish I'd done it at build time however. No one likes pulling a board to put it right back in!


*EDIT*
Orthos was crashing at 340x9 with same voltage prior to mod but now runs flawlessly. I initially attributed crash to lack of voltage and lowered FSB because I didn't want to raise voltage. Lower PWM temp appears to have helped stability as well. Sticking to 3.06GHz until I see a need for more.


NOTES
This is not necessarily an Abit issue. Abit's uGuru reports PWM temps, the only reason high PWM temps caught the eye. Abit P35 boards have proven to be quite stable and overclockable.
Regarding testing method, yes there are better temp tools out there. uGuru worked well enough for this test.

 

[SerpentRoyal]

You only need to make sure that the heat sink is flat and making contact with all the MOSFETs. A small amount of thermal paste and the stock spring pins should do the job. My maximum temp under load is 52C with 85C ambient. Quad will probably raise the temperature to 62C.

The issue with IP35 Pro is that the NB, SB, and MOSFET coolers are joined together, which makes it difficult to achieve a flat plane.

I lapped each heat sink on the IP35-E and use a small low speed 80mm Panaflo to cool the RAM, NB and SB. MOSFET heat sink runs cool and therefore, does not need active cooling.

 

[Sylvanas]

Good read SteelSix! 15 Degrees is quite a feat, if not improve your OC it'll at least add a little longevity to the board

 

[Hauk]

Originally posted by: SerpentRoyal
You only need to make sure that the heat sink is flat and making contact with all the MOSFETs. A small amount of thermal paste and the stock spring pins should do the job. My maximum temp under load is 52C with 85C ambient. Quad will probably raise the temperature to 62C.

The issue with IP35 Pro is that the NB, SB, and MOSFET coolers are joined together, which makes it difficult to achieve a flat plane.

I lapped each heat sink on the IP35-E and use a small low speed 80mm Panaflo to cool the RAM, NB and SB. MOSFET heat sink runs cool and therefore, does not need active cooling.

You're right on regarding the flat plane issue with this particular one-piece heatsink. Even when I straightened out the heatpipe, the mofset heatsink still wasn't making effective contact. Rather than lapping, it needed additional retention pressure, hence the bolts.

But when you say "Quad will probably raise the temp to 62C," I'm not the only IP35 Pro user to report 75C+ PWM temps under load. Apples to apples please.

Serpent conveys an important point, the bolts may not be necessary for your application. Better TIM and active cooling may be enough. Or, you may not have PWM temp issues at all. But if you do, improving contact, interface, and cooling may help.

Serpent, that's a scary ambient you have there.

 

[SerpentRoyal]

I'm sure a lapped heat sink and paste would yield PWM <62C with 85C ambient and any Quad. Why? After lapping all IP35-E's heat sinks and applying thermal paste, I put a TC at the base of the NB heat sink. Peak NB temp under Orthos large mode is 47C (85F room). NB heat sink is very warm to the touch. The SB heat sink is luke-warm. The MOSFET heat sink is barely-warm. This tells me that the MOSFET heat sink has a lot more cooling "head-room". Look at the back of the IP35 board. The MOSFET/CPU region has a massive thermal pad inside the board to dissipate heat. A decent portion of the MOSFETs' heat will go into the board, with the remaining up into the MOSFET heat sink. A gentle stream of air flow to this region will go a long way to cool the output devices.

The problem with IP35 Pro (including many fancy boards) is that it uses one interconnected heat pipe to cool the NB, SB, and MOSFETs. This solution moves heat from the hottest NB chip to the cooler MOSFET area (evaporator/condensor), which ultimately increases the load temperature at the PWM.

Based on my load/unload profile data of my E4300 (3.46GHz), I would expect to see 59C PWM with an overclocked Quad and this IP35-E rig.

I bench all stuffs in the lab next to the garage.

 

[Hauk]

* Deleted *

I hope an IP35 Pro user or someone else wanting to improve their PWM temps finds this information useful...

 

[SerpentRoyal]

You believe what you want to believe. I stand behind my post with concrete scientific data.

At this time, I have no CPU to confirm that this board will run any Quad on air @ your 3.0GHz speed with less than 62C PWM. If you cannot handle an alternate solution backed with scientific data, then perhaps you shouldn't post on a public forum.

 

[cmdrdredd]

Originally posted by: SerpentRoyal
You believe what you want to believe. I stand behind my post with concrete scientific data.

At this time, I have no CPU to confirm that this board will run any Quad on air @ your 3.0GHz speed with less than 62C PWM. If you cannot handle an alternate solution backed with scientific data, then perhaps you shouldn't post on a public forum.

I didn't see you post any "scientific data"

 

[SerpentRoyal]

Originally posted by: cmdrdredd

Originally posted by: SerpentRoyal
You believe what you want to believe. I stand behind my post with concrete scientific data.

At this time, I have no CPU to confirm that this board will run any Quad on air @ your 3.0GHz speed with less than 62C PWM. If you cannot handle an alternate solution backed with scientific data, then perhaps you shouldn't post on a public forum.

I didn't see you post any "scientific data"

Buy your own IP35-E and mount a K-type TC per my previous post. Run the Orthos Large test between 80 and 85F ambient.

 

[JohnnyTT]

SteelSix. Great post!!!

I just got my IP35-pro last week and set it up this weekend with a Q6600 running 3,330 for now. I can go farther.. but haven't adjusted voltages or tweaked it much.

I did notice on mine that the Mosfet heatsink isn't sitting flat. According to uGuru, my temps keep changing. I think the Hsink is moving.

I'm going to have to do what you did this week. Great post.


As for the other guy messin up your post. Just brush that $#!^ off. Great job and thanks for the info and numbers.

 

[Hauk]

Originally posted by: JohnnyTT
SteelSix. Great post!!!

I just got my IP35-pro last week and set it up this weekend with a Q6600 running 3,330 for now. I can go farther.. but haven't adjusted voltages or tweaked it much.

I did notice on mine that the Mosfet heatsink isn't sitting flat. According to uGuru, my temps keep changing. I think the Hsink is moving.

I'm going to have to do what you did this week. Great post.


As for the other guy messin up your post. Just brush that $#!^ off. Great job and thanks for the info and numbers.

Thanks. It's a great board, but after reading on Abit forums about a guy who's board blew smoke from mofset then died, and after cmdrdredd's initial post, I decided to investigate. One solution was to cut the heatpipe off from mofset to NB, other was the bolt mod.

Even if you just shoot for better thermal grease and a fan you should see an improvement. I bet you'll see an inconsistent footprint in the mofset thermal pad when you remove heatsink. Arctic Silver Ceramique is good because of its thickness. A big bead on top of each mofset will do the trick. Don't worry about slop. You want to bridge any potential gap if you stick with push-pins.

Regarding "differing viewpoints," they're valuable, but I knew the problem lie with the heatpipe moreso than the flatness of the mofset heatsink. In any case, I've retracted my statement. Hope I didn't offend anyone. Sorry Serpent.

 

[cmdrdredd]

Originally posted by: SerpentRoyal

Originally posted by: cmdrdredd

Originally posted by: SerpentRoyal
You believe what you want to believe. I stand behind my post with concrete scientific data.

At this time, I have no CPU to confirm that this board will run any Quad on air @ your 3.0GHz speed with less than 62C PWM. If you cannot handle an alternate solution backed with scientific data, then perhaps you shouldn't post on a public forum.

I didn't see you post any "scientific data"

Buy your own IP35-E and mount a K-type TC per my previous post. Run the Orthos Large test between 80 and 85F ambient.

Where is your charts? the pics of your setup? before and after temps and results?

Also why should I buy an IP35-E and downgrade from what I have currently?

 

[SerpentRoyal]

Buy me a digical cam if you want pic of the rig. Other option is to browse the net and ask those with IP35-E/quad about their PWM temp. I don't need to prove squat to you or anyone on this forum. Informations is provided as is, for educational purpose only.

Why buy IP35-E? From a purist point-of-view, IP35-E uses a better LAN controller on PCI-E bus. The Jmicron is also on the PCI-E bus. Finally, the IP35-E has a more robust BIOS.

 

[SerpentRoyal]

Originally posted by: SteelSix

Originally posted by: JohnnyTT
SteelSix. Great post!!!

I just got my IP35-pro last week and set it up this weekend with a Q6600 running 3,330 for now. I can go farther.. but haven't adjusted voltages or tweaked it much.

I did notice on mine that the Mosfet heatsink isn't sitting flat. According to uGuru, my temps keep changing. I think the Hsink is moving.

I'm going to have to do what you did this week. Great post.


As for the other guy messin up your post. Just brush that $#!^ off. Great job and thanks for the info and numbers.

Thanks. It's a great board, but after reading on Abit forums about a guy who's board blew smoke from mofset then died, and after cmdrdredd's initial post, I decided to investigate. One solution was to cut the heatpipe off from mofset to NB, other was the bolt mod.

Even if you just shoot for better thermal grease and a fan you should see an improvement. I bet you'll see an inconsistent footprint in the mofset thermal pad when you remove heatsink. Arctic Silver Ceramique is good because of its thickness. A big bead on top of each mofset will do the trick. Don't worry about slop. You want to bridge any potential gap if you stick with push-pins.

Regarding "differing viewpoints," they're valuable, but I knew the problem lie with the heatpipe moreso than the flatness of the mofset heatsink. In any case, I've retracted my statement. Hope I didn't offend anyone. Sorry Serpent.

No big deal...there are many ways to go from A to B. If I didn't have the BT for cooling, then I would probably remove the MOSFET's heat sink and replace it with one low speed 80 mm Panaflo fan.

People often under-estimate the efficiency of that big thermal pad embedded in the MB. Each MOSFET is soldered to the MB, so heat transfer should be excellent. It's much more efficient to cool the thermal pad with cold air, rather than placing a cooler on top of the MOSFET to draw heat from the thermal pad. Warm, dead air pocket around the base of the MOSFET heat sink makes excellent thermal insulator.

 

[cmdrdredd]

Originally posted by: SerpentRoyal
Buy me a digical cam if you want pic of the rig. Other option is to browse the net and ask those with IP35-E/quad about their PWM temp. I don't need to prove squat to you or anyone on this forum. Informations is provided as is, for educational purpose only.

Why buy IP35-E? From a purist point-of-view, IP35-E uses a better LAN controller on PCI-E bus. The Jmicron is also on the PCI-E bus. Finally, the IP35-E has a more robust BIOS.

1) the LSN is a non issue as I am not saturating the network enough to matter
2) I don't use Jmicron and if I did it would be plenty for my usage
3) more robust BIOS? how so? the features are exactly the same...overclocking ability the same.

 

[SerpentRoyal]

Higher RAM speed with four modules. I had to reset my CMOS several times and/or reflash BIOS when working with IP35 Pro because the BIOS screen would flicker during adjustment. Never had this problem with IP35-E.

 

[cmdrdredd]

Originally posted by: SerpentRoyal
Higher RAM speed with four modules. I had to reset my CMOS several times and/or reflash BIOS when working with IP35 Pro because the BIOS screen would flicker during adjustment. Never had this problem with IP35-E.

I've never had that happen. I've used all the different revisions. I have no problem running high memory speed on my Ip35 Pro but I generally like to keep my memory voltages down. For a while I was running some Crucial 10th Anniversary at DDR2-1100 5-5-4-12 but needed 2.4v. Killed the memory quickly

 

[GreyGeezer]

I still run an Abit AT8-32X Opteron 170/2 Gig GSkill DDR500... uGuru would report PWM temps all the way up to 103C at prolonged full load ! I do not have air conditioning in my computer room and deal with high ambient temps-sometimes 31c. I maxed out the warning temp, and disabled temperature shutdown. My most common program to trip the alarms was DVD Shrink. Nero Recode dosn't seem to create such high temperature... Go figure. I needed to reduce my overclock from 2.65Ghz, back to stock (2.0Ghz) until the heat wave broke. Back at 2.65, but need to deal with the hot PWMs soon.

 

[Lvcoyote]

Good tutorial except that system temp is not the northbridge temp. There is no northbridge temp monitoring on most abit boards. The system temps are from a small sensor on the face of the motherboard and is nothing more than a reading of what the temperature is inside your case. OTher than that, very nice tutorial!

 

[Hauk]

Hmm, I had an idea (after 3 Red Bulls) and changed the orentation of one of my Spot Cool fans, had it blowing onto the back of my 8800GTX card. It was likely moving hot air towards the sensor. One variable I hadn't considered as I tested.

I changed the orentation to point mostly back at the motherboard and have system temp back down to 36. Thanks...

 

[Lvcoyote]

Did this Mod today, luckily I was able to find all the parts needed locally. Ace hardware had all the #4 machine screws and bolts. I couldnt find nylock #4 bolts so I just used regular ones instead, they should hold ok. Also had to go with 3/4 inch long bolts, the didnt have the 5/8 or 9/16, but again they worked just fine. Just though I'd post this so people know they can find the stuff locally that will make it work. I had almost the same end results too, about a 15c temp drop idle and load.

 

[Hauk]

Glad it worked out. I read on another forum someone found #4 hardware at Ace. Save $$ on shipping cost. Only Ace near me is $5.00 worth of gas away. Gas, now there's a topic for debate.

 

[cytoSiN]

Any reason why the bolts/nuts need to be stainless? Local hardware store has #4, but nothing stainless. Thanks.

 

[SerpentRoyal]

Originally posted by: cytoSiN
Any reason why the bolts/nuts need to be stainless? Local hardware store has #4, but nothing stainless. Thanks.

No. The important thing is to get a non-conductive washer. Also do not overtorque the assembly. There are delicate electronic traces under the board.

 

[cytoSiN]

Ok, so the washer underneath the board needs to be nonconductive, but the nut on top doesn't? And no nonconductive washer on top? I'm a noob with fastening things to the board (other than my HSF, which has its own bracket mounting system), and I just want to be sure I don't screw anything up. My PWM temps on an IP35 (not E or Pro) idle at around 44-47 C, and get close to 70 C at high load. I know I won't be running at those loads for very long in normal circumstances, but the temps still seem like they can be lower just be increasing the contact with the stock heatsinks, so I'm interested in giving this mod a try...Thanks.