CPU cooling might not be the most important thing to review

Cooling Processors quietly

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
marcus_helsinki
Posts: 15
Joined: Mon Aug 09, 2010 1:23 pm
Location: Finland

CPU cooling might not be the most important thing to review

Post by marcus_helsinki » Mon Aug 09, 2010 2:30 pm

Nowadays, CPUs can stand rather high temperatures and can protect themselves by throttling.
Instead, motherboard components, e.g. VRMs and Northbridge, may not stand as high temperatures and can't throttle if they are overheated!

So, when we see reviews where only CPU temps are measured, and coolers are ranked based on that, perhaps, more important temps are left unmeasured!

If the VRMs or Northbridge get overheated, you lose your motherboard an perhaps some other components, and may get some sort of fire in the worst case, but if your CPU gets overheated, you notice the throttling and can just add cooling without a prior damage to your equipment.

The present trend to have large tower coolers for CPU may in practice, lead to very low temperatures of CPU but much higher temperatures of VRMs and Northbridge. With tower coolers, additional fans directed at motherboard components may be needed to ensure a long lifetime of the components.

faugusztin
Posts: 450
Joined: Mon Mar 29, 2010 2:47 am
Location: Bratislava, Slovak Republic

Post by faugusztin » Mon Aug 09, 2010 3:17 pm

It was long time since i seen 60C or above on VRM or chipset, mostly they are at 50C range; even with uncooled VRM's temperatures around 70C are a rarity.

So yes, it is a good thing to measure VRM and chipset temperatures, but it is nowhere near the seriousness of good old nForce4 problems.

marcus_helsinki
Posts: 15
Joined: Mon Aug 09, 2010 1:23 pm
Location: Finland

Post by marcus_helsinki » Mon Aug 09, 2010 10:49 pm

Thanks for your comment.

Why I wrote that text was my experiences; the CPU of my new system runs mostly between 33 and 34 C while VRM:s are now at 47 C; without added cooling, VRM:s reached 61 C, while the yellow "warning range" of the temperature monitoring software by Intel for the motherboard Intel DQ57TM is 50C...60C and "red" zone >60.
The yellow zone for the CPU, instead was as high as 72C...97C, and "red" >97.
So in the system, CPU should be over 60 C warmer to reach the red zone, while the VRMs already were there!

But, I am not sure whether the temperature readings and measurements of the board are accurate. What do you think? Were the VRMs really at 60 C or was there some problems with the readings?

The cooler was Prolimatech Megahalems, now I have also added a Nexus 120 mm blowing at an angle of 45 degrees unhindered at the VRM area. Adding a Nexus to the cooler tower had no or almost no effect on CPU or VRM temperature (readings).

tim851
Posts: 543
Joined: Wed Aug 13, 2008 11:45 am
Location: 128.0.0.1

Post by tim851 » Tue Aug 10, 2010 5:56 am

Are you sure then sensor that measured the 60°C is the VRMs?

Speedfan finds three sensors on my MSI H55M-ED55 board that I cannot identify with certainty.
One (temp1) is most likely the socket diode for the CPU, staying close to the Core-temps, trailing it slowly.
I suspected temp3 to be the temp of the integrated graphics, since it seems to be affected by the speed of the cooler fan, but when I ran furmark, it didn't even blink, although the whole system almost came to a hold.
And temp2 always hovers between 49°C and 51°C, under prime95 it went slowly up. That could be the VRMs, but when I pointed a 60mm at the VRM coolers (both of them), it had no effect on temp2. Also, no matter what kind of airflow I produce, it clings to those 50°C like nothing.

faugusztin
Posts: 450
Joined: Mon Mar 29, 2010 2:47 am
Location: Bratislava, Slovak Republic

Post by faugusztin » Tue Aug 10, 2010 6:01 am

marcus_helsinki wrote:...
What do i think ? Well, you shouldn't overestimate the value of bundled software and their warning ranges.

VRM's are a lot simpler things than CPU, if they cannot withstand temperatures around 100C then they are electronic waste and not something which should be in your PC.

60C for VRM's without heatsinks is pretty normal.

marcus_helsinki
Posts: 15
Joined: Mon Aug 09, 2010 1:23 pm
Location: Finland

Post by marcus_helsinki » Tue Aug 10, 2010 7:16 am

I don't know about VRMs in other motherboards but in the very good and precise technical manual, it is said that the temperature of VRMs should not exceed 70 C in DC57TM. Because there may be only one temperature sensor and this may not be located at the hottest point, I think it is wise to follow the guidance that the reading at the point of the sensor should not exceed 60 C, and try to get it out of the yellow warning zone border (50 C).
In the motherboard, the components aren't of highest quality (chokes) but not lowest, and it is moderately 5 stage VR system, not the best, not the worst. The VRMs do not have any heatsinks on them. This surely raises the temperature. Maybe with heatsinks or more stages, temperatures of VRMs are lower.

Cerb
Posts: 391
Joined: Tue Apr 13, 2004 6:36 pm
Location: GA (US)

Post by Cerb » Tue Aug 10, 2010 7:19 am

I haven't seen any nVidia chipset boards out in the wild, lately. You can still find cheap ones, but then you know you're sacrificing for other reasons, too.

If VRM temps concern you, get a board that cools them. They aren't expensive. Without hot NBs, and low-efficiency CPU voltage regulation, any of them will be fine (nForce boards, FI, could sometimes get damaged by the heatpipe coolers keeping everything hot).

Shamgar
Posts: 454
Joined: Wed Oct 22, 2008 8:49 am
Location: Where I Am

Post by Shamgar » Tue Aug 10, 2010 10:08 am

We've had past discussions here at various times concerning motherboard cooling. Most people don't seem to be concerned with it, while others are. Here are some of those threads:

viewtopic.php?t=50714
viewtopic.php?t=54682
viewtopic.php?t=36445

I personally prefer a setup with some sort of cooling of the motherboard components. Some better featured "brandname" boards have heatsinks on the VRMs and MOSFETS. MSI's motherboard cooling and efficiency is one of the best though, so you might want to have a look at some of their boards.

bonestonne
Posts: 1839
Joined: Mon Feb 05, 2007 2:10 pm
Location: Northern New Jersey
Contact:

Post by bonestonne » Tue Aug 10, 2010 10:35 am

I use larger tower style heatsinks for just about everything, but VRM cooling never hit me as a total necessity, and my rig runs at full tilt when it's on, I don't have throttling enabled, because it causes problems with my sound cards.

That said, temps never go beyond 50C anywhere on my computer, even when I'm running 16 tracks and multiple VSTs on each one.

When I run brutal stress tests, that's when times start to climb, but most brand name motherboards are moving over to solid state and ferrite core components, so temps aren't going to shoot up anywhere near they would on say a Northwood P4.

I mean, sure, lots of users here have silent PCs, a whole lot more have quiet PCs, but that doesn't mean cooling is sacrificed for it. When people will take one fan out to drop noise, they usually duct another one to make up for that lack of airflow. If you're really worried about your temps, get better heatsinks for the right components. Thermalright makes VRM heatsinks, but not all boards have the mounting holes for that. Even using small RAM heatsinks would do the trick on a VRM, a little airflow goes a long way.

Shamgar
Posts: 454
Joined: Wed Oct 22, 2008 8:49 am
Location: Where I Am

Post by Shamgar » Tue Aug 10, 2010 11:06 am

Aftermarket m'board heatsinks are usually expensive. Would like to see more m'board manufacturers implement better stock cooling on their boards. Same for their graphics cards.

danimal
Posts: 734
Joined: Mon Jun 08, 2009 2:41 pm
Location: the ether

Re: CPU cooling might not be the most important thing to rev

Post by danimal » Wed Aug 11, 2010 8:31 pm

marcus_helsinki wrote:If the VRMs or Northbridge get overheated, you lose your motherboard an perhaps some other components, and may get some sort of fire in the worst case, but if your CPU gets overheated, you notice the throttling and can just add cooling without a prior damage to your equipment.
the last time i looked at intel cpus, they were designed to self-throttle at 100 degrees centigrade(?), which is very very hot, and actually does little harm to the chip itself... read up on the installation procedure for indigo tim, it'll freak you out.

in light of that, any drama you read about vrm temps and such is highly overblown, especially if the chips have heat sinks on 'em... without knowing the factory temp specs for the chips in question, concerns about life expectancy and such are idle speculation.

some of the mb chips on my msi mb don't even have heat sinks on 'em.

marcus_helsinki
Posts: 15
Joined: Mon Aug 09, 2010 1:23 pm
Location: Finland

Post by marcus_helsinki » Thu Aug 12, 2010 10:36 am

Thanks for the comment, I think You understood me wrong, because what You wrote is approximately the same than that I wrote about CPU; I meant a situation, where the CPU is safe but the VRMs and/or the northbridge get overheated; in this case, the throttling of the CPU will save nothing, because VRMs and northbridges don't throttle.

I mean that because some CPUs can do with even 100 C at least some time, and if they get ovet 100 C, then they throttle, so no damage.

But on some boards, VRMs are not so high quality production and can be without good heatsinks, and don't tolerate that much heat. If they get overheated, their effectivity in converting the voltages decreases and they produce even more heat, and so on... You will build a new computer.

steve_lee_01
Posts: 9
Joined: Tue Oct 27, 2009 4:55 pm
Location: California

my experience

Post by steve_lee_01 » Thu Aug 12, 2010 6:07 pm

I've experienced a similar issue on my intel motherboard. It's an older Pentium 4, Socket 478 motherboard (intel D865PERL). According to the intel monitoring software this board has 3 temperature zones--Zone 1, Zone 2, and CPU Zone. According to the intel website Zone 2 corresponds to your VRM zone location, which is the area between the I/O shield and the CPU. Zone 1 is the "cold" area upstream of the cooler where the RAM is located.

The CPU zone has an alarm temperature of 68C while Zones 1 and 2 have alarm temps at 50C. The intel website said that I could increase the alarm temps in Zones 1 and 2 to 60C.

When I had a top down cooler the CPU zone regularly ran much hotter than Zones 1 and 2. When I upgraded to a tower style cooler (Scythe Mugen 2) I noticed that Zone 2 increased in temperature and was within 1 or 2 degrees of the CPU temp and sometimes 1 or 2 degrees higher. Zone 1 stayed relatively cool.

This wasn't a problem in the wintertime, but I easily approach the 60C alarm temp on Zone 2 in the summer. I'm upgrading to an i7 system later this year and I always wondered if the issue that I'm experiencing would be a problem on a more modern system. It sounds like from your experience that it still is.

I tried removing the motherboard I/O shield around the rear connectors to dispel the pocket of heat that gets trapped underneath the exhaust fan, but this didn't help at all.

As a second experiment I left the I/O shield off, but blocked off 75% of the area with tape. This left a small portion around the rear connectors open. I have a positive pressure system and tried to channel more exhaust air through this smaller opening. This bought me a 3-5C improvement, but it hasn't solved the problem.

marcus_helsinki
Posts: 15
Joined: Mon Aug 09, 2010 1:23 pm
Location: Finland

Post by marcus_helsinki » Fri Aug 13, 2010 12:32 pm

I also experimented with some combinations of chassis airflow, but I noticed that the only thing very effective was a relatively direct, strong airflow to the VRM area. An other option could be adding heatsinks, but that wouldn't cool all components and would be riskier, and in the board I have, the different components are mixed on the PCB so that assembling of heatsinks would require a large number of small heatsinks, and there always is the risk of electrostatic charges when installing. So I took the easy and very effective option of adding a very silent fan blowing especially at the VRM area. There are available 80 mm and 90 mm fans too, if space is sparse, and there are also flow guiders for fans and special fans that concentrate the flow so that the fan would be effective from a larger distance, too. Still, it is safe to plan having the fans as near and possible and blowing as directly against the VRM area as possible.
In my board, there also the IG chip is located in the VRM area heating it more.

danimal
Posts: 734
Joined: Mon Jun 08, 2009 2:41 pm
Location: the ether

Post by danimal » Sat Aug 14, 2010 10:43 am

you can't use throttling to compare cpu chips to vrm chips, because some vrms are designed to to run without heat sinks, so there isn't an engineering scenario where they would need to throttle back because of the heat... cpu throttling is designed to protect the cpu, if the heat sink fails.

since you probably aren't exceeding the design parameters of the vrms, by using overclocking for instance, there is no need to worry excessively about the vrm temps.

i brought up the intel cpu temp spec only to show how durable these chips are.

marcus_helsinki
Posts: 15
Joined: Mon Aug 09, 2010 1:23 pm
Location: Finland

Post by marcus_helsinki » Sun Aug 15, 2010 7:10 am

Thanks for comments,

Is it really so that if VRMs get overheated, CPU throttles? I have information that the throttling is controlled by CPU temp, but I am not expert on this field.

For the Intel board, there is a design parameter "air flow at VRM area", and was it as high as 2 m/s!
So, it may be that almost, the only cooler that can do that is the stock cooler with a tiny heatsink but the fan very near the VRM area, blowing at it with a high RPM.

danimal
Posts: 734
Joined: Mon Jun 08, 2009 2:41 pm
Location: the ether

Post by danimal » Sun Aug 15, 2010 8:17 am

i'm no engineer either, and i was worried enough about the vrm temps to add a 120mm fan to blow air across 'em... but i overclock my cpu pretty heavily.

i don't know if vrm overheating coud be designed to throttle back the cpu? interesting question.

lodestar
Posts: 1683
Joined: Fri Aug 05, 2005 3:29 am
Location: UK

Post by lodestar » Sun Aug 15, 2010 1:29 pm

It might be useful to note this in the context of Asus P7P55D motherboards

".... TProbe is another feature that’s common to all P7P55D motherboards. TProbe is a built-in thermal controller that actively monitors temps on the VRM area of the board and can automatically adjust workload from one group of power phases to cooler-running phases if one area begins to overheat. By working to keep the power phases running as cool as possible, this helps to extend the motherboard’s longevity....".

The Asus P7P55D boards all have a minimum of 12 power phases (VRMs). I assume other manufacturers may have something similar to TProbe. Asus also have an onboard EPU (Energy Processing Unit) which dynamically turns off/turns on power phases depending on demand.

marcus_helsinki
Posts: 15
Joined: Mon Aug 09, 2010 1:23 pm
Location: Finland

Post by marcus_helsinki » Sun Aug 15, 2010 2:25 pm

The Intel board we discuss has only 5 stages and only measurement of temperature, and no heatsinks on VRM.

The ASUS board VRM temperature control You mentioned seems to be very advanced compared with the Intel.

Unfortunately, the board doesn't accept i3 530 so I couldn't consider it.

I think the Intel board is designed to be used with the stock cooler of the CPU.

But it accepted i3 530, was easy to install, worked at once with no problems at all and has 2*simultaneous dual DVI ports, is reliable, consumes very little energy and boots up very fast. So, it is not a bad board but it is, possibly, designed to be used with the stock cooler cooling the VRM area effectively, and therefore there was no need for adding fine VRM thermal properties. When used with other coolers, the VRM cooling may become weaker.

faugusztin
Posts: 450
Joined: Mon Mar 29, 2010 2:47 am
Location: Bratislava, Slovak Republic

Post by faugusztin » Sun Aug 15, 2010 11:29 pm

Man, stop putting so much faith in Intel and so much distrust in other manufacturers. Each and every Socket 1156 motherboard supports each and every Socket 1156 CPU. Of course if you put a CPU with IGP into a P55 board you lose the IGP functionality, but that doesn't mean it "doesn't accept i3 530".

Did you even bothered going to Asus website and check out the CPU support list ?
The first item in list ( http://www.asus.com/product.aspx?P_ID=RBA8CzWoopUlYRFZ , CPU Suport List tab) is :

Code: Select all

Core i3 530(2.93G,L2:4M, IGPU,rev.C2,73W)   	 ALL	
1207
As i said before, desktop Intel boards are not the best boards on market, they are pretty much the lowend stuff for OEM's. Of course it uses a little less power when they don't put a aditional SATA, Firewire, IDE, secondary ethernet, USB3, SATA3,... support on their motherboard.

The only positive thing about your board is that it has dual DVI outputs, as dual digital outputs are pretty rare. But otherwise your board is not worth the problems you have with it - next time look around before shopping. The only Intel motherboards i see usefull to buy are their ITX boards, the rest is "meh, forget it".

Fëanor
Posts: 46
Joined: Sat Jun 02, 2007 12:01 pm

Re: CPU cooling might not be the most important thing to rev

Post by Fëanor » Mon Aug 16, 2010 5:53 pm

You're looking to solve a problem that doesn't really exist. There are not common problems with the temperature of these components.

Adding cooling costs money, power, and noise. Show me the benefit of those expenditures, and it better be more than theoretical component life if you want to convince me.
faugusztin wrote:As i said before, desktop Intel boards are not the best boards on market, they are pretty much the lowend stuff for OEM's.
There really is no "best brand". Intel boards are only "low-end" in that they are not feature-rich. Intel favors a bare-bones, solid reliability approach as opposed to ridiculous overclocking options in the BIOS, a whole lot of extra controllers, etc. This is why they're used by OEMs: why pay for features you aren't going to use?

marcus_helsinki
Posts: 15
Joined: Mon Aug 09, 2010 1:23 pm
Location: Finland

Post by marcus_helsinki » Tue Aug 17, 2010 9:21 am

Thanks for the good comments. I think the same. It seems that Intel makes well designed products, because they are extremely well documented and work as documented. I haven't noticed any deviation from the documentation, and installation was only to put the hardware in, put the CD in, READY!, everything works. The motherboard is mostly designed to be used with the stock cooler. A temperature of 60 C of VRM is, as You said, a question of long time durability of components. But, I want that the motherboard doesn't burn after 10 or 15 years of use ( I plan that the computer is used a long time) with any probability, because it is located at home where children also sleep. So no risks. Of course, the components are protected with bromides against fire, but if there is dust accumulation inside the cabinet because of possible lazyness of user, and anyway, I don't want any electric component burning smell.

The ASUS board mentioned here is designed to be used with other coolers than the stock coolers, too.
The ASUS board has double the amount of stages in VR, so the temperature increase of VRMs would be only 50%?, heatsinks -> a further large reduction, and sensors and algorithms used for loading VR components equally, so may be those work with tower coolers also? Any experience? Any experience on other boards?

BlackWhizz
Posts: 266
Joined: Wed Mar 31, 2010 8:19 am
Location: OV, The Netherlands

Post by BlackWhizz » Sun Aug 22, 2010 9:28 am

Its al about the airflow in your case. A little breeze over your motherboard will make a big difference.

VRM's can easily stand temps like CPU's. Also it has a link with the TDP of the CPU.

kjetil
Posts: 14
Joined: Mon Nov 09, 2009 3:05 am
Location: Oslo

Post by kjetil » Mon Aug 23, 2010 2:53 am

--Deleted my message--

djkest
Posts: 766
Joined: Mon Sep 10, 2007 1:05 pm
Location: Colorado, USA

Post by djkest » Thu Aug 26, 2010 8:17 am

faugusztin wrote:It was long time since i seen 60C or above on VRM or chipset, mostly they are at 50C range; even with uncooled VRM's temperatures around 70C are a rarity.

So yes, it is a good thing to measure VRM and chipset temperatures, but it is nowhere near the seriousness of good old nForce4 problems.
Ha! I have an nForce4 motherboard. The northbridge had this raging fan on it that make a huge noise. I bought a pure copper heatsink for $25 and it got "burn your fingers" hot even at idle. I ended up buying a $30 massive thermaltake heatpipe heatsink to tame that. It ran smoking hot without that thermaltake heatsink. Biggest pain in the butt...

Post Reply