secondary board heat 'losses'

[BillA]

another member, Colt357tw who agreed to be quoted, was at the IDF in Taipei this year where Intel stated about 20% of (secondary) heat losses goes thru traces and passed over to PCB and surrounding components on their 875P models

Jim then asked right away if Intel can provide these type of tests for their partners(motherboard makers) since hooking up over 30 thermal couples on an ATX motherboard and 8 more on the IHS is way too much work for any motherboard makers, even 1st tier ones like ASUS, engineers would kill themselves try to gather meaningful data from run that type of testing.
the Intel response is not known.

the Asus twist can be seen here
but the addition of such a plate could boost that percentage considerably

the same old reason to use a heat die

[killernoodle]

Hmm, so having a fan on your motherboard tray IS a good thing...

[Les]

Oh woe
My reasoned guesses have been upto 40%

[BillA]

jeez Les, ancient history and still interesting 2 yrs on
(same old group)

[jaydee]

How could you test something like that? What about all the caps and mosfets surrounding the socket? Surely those add substantial heat into the traces as well. I can definatly see some significant heat loss though.

I can't see how having a fan there would be a bad thing except for throwing off onboard temps. I wonder if they calibrate the secondary loss's into the formula on their temp monitoring? I would think the more efficent the cooling the more off the temps would be being, I would expect, they test with OEM heatsinks being thats what is mainly used in the OEM and general real world.

Of course I spent the last 3 days outside baking my brains in 100+ weather here in Texas and may not be thinking strait.

[Cathar]

Would not the percentage of cooling offered by the board be relative to the cooling effect being applied to the CPU?

i.e. say the board has a C/W of 1.5.

We stick on a heatsink with a C/W of 0.5.

The total system C/W then becomes (1.5 x 0.5) / (1.5 +0.5) = 0.75 / 2 = 0.375. The motherboard now plays a 25% role in the cooling effect.

If we now stick on a waterblock with a C/W of 0.2, we have a total C/W of (1.5 x 0.2) / (1.5 + 0.2) = 0.176, and the motherboard now plays a ~12% role in the total cooling effect.

Still a significant amount of cooling, sure, but a fixed figure of 20% surely cannot be relied upon.

[AngryAlpaca]

^^^ Don't forget the fan's cooling effect. (Not sure if it's relevant or not, but fan power on CPU heatsink is directly proportional to MB cooling effect, isn't it?)

[BillA]

I think the 20% related only to that board AND the Intel reference heatsink
other boards and processors, other values
that is why Jim asked if Intel was inclined to test
and he had a joke " I think boys from Intel Malaysia (packaging/PCB thermal) lab would really have a cow if Intel decided they can provide these services with a minimium fee. (I can already see ASUS, MSI, Gigabyte, ECS, ABIT and 20 other motherboard makers with at least 8 motherboards each line up for testing.)"

agreed Cathar, the better the sink the lower the secondary losses

[Colt357tw]

20% is relative to their particular tested CPU, HFS, and motherboard. Testing method can be found on Intel Developer papers as well as various Thermal Design Guides. My guess is Intel now realized 130+W CPU combined with 150W PCI-E graphic cards is a little too much for their proposed little PC case.

Since these info was from a "Forum" (IDF = Intel Developer Forum) and most of these data/testing results are from their online library, if Intel want me "Do not reproduce" as stated on the booklet, I guess I will nont have any kids then .

[BillA]

"proposed little PC case" would be the BTX ?

it seems strange to hang all that on a 92mm fan - crossflow

[Colt357tw]

Originally Posted by AngryAlpaca

^^^ Don't forget the fan's cooling effect. (Not sure if it's relevant or not, but fan power on CPU heatsink is directly proportional to MB cooling effect, isn't it?)

Fan on heatsinks would have more cooling effects to MOSFET, coils and capacitor since these components had slightly larger surface area than rest, which would in turn improve motherboard DC(CPU) power supply effeciency thus higher wattage CPU can be use......provided the fin overhang is large enough and/or the HS fin orientation is "omni-directrional"(whatever that means :shrug: )

This was also an issue DFI RD Engineer Oskar Wu (Designer of several very famous/infamous mobo but all highly oc'able) raised when we had a little chat about LGA775 layout/thermal design last December. Our conclusion was, Intel unleashed a fire-spitting dragon and will force many common user(who purchase CPU with no aftermarket/custom cooling) switch to AMD A64/A64FX playground, and looks like the trend is indeed heading that direction.

[BillA]

are mobo mfgrs forecasting a significant shift in production %s then ?

[Les]

Originally Posted by Cathar

Would not the percentage of cooling offered by the board be relative to the cooling effect being applied to the CPU?

i.e. say the board has a C/W of 1.5.

We stick on a heatsink with a C/W of 0.5.

The total system C/W then becomes (1.5 x 0.5) / (1.5 +0.5) = 0.75 / 2 = 0.375. The motherboard now plays a 25% role in the cooling effect.

If we now stick on a waterblock with a C/W of 0.2, we have a total C/W of (1.5 x 0.2) / (1.5 + 0.2) = 0.176, and the motherboard now plays a ~12% role in the total cooling effect.

Still a significant amount of cooling, sure, but a fixed figure of 20% surely cannot be relied upon.

Have considered various combinations in this discussion (only 5 months ago)
For example:

[Colt357tw]

Originally Posted by unregistered

are mobo mfgrs forecasting a significant shift in production %s then ?

Sorry Bill,

that's mobo mfgr PM's call, and guys in RD knows better to make any comments on that subject :P

Back to the subject on secondary loss. I had always fond to use ~90% (DC/DC effeciency) of VRM spec maximum supported wattage (VID*ICC) as max CPU thermal energy needed to take care of, though VRM 10.1 (Intel documentation 30235601) shown a very nightmarish number on page 14.......1.4V*119A*0.9~150W. Intel really playing with fire when it tells everyone its CPU only put out 115W(Table 5.1, Page 74, Intel Documentation 30235101) 150W by 20% of Intel claimed would be 30W, which had enough impact to make/break a thermal design on a °C/W curve

Bill, looks like I aint kidding when I said we might need a setup to handle 500W of dissipating heat in a very near future ah? Would someone please figure out how to make electricity from these wasted thermal energy so we can all recover some from our electricity bill?


by the way, had anyone noticed -30W from that 150W, 120W is quite close to 115W Intel claimed?

[BillA]

did not mean to lead you astray there with my question

"Would someone please figure out how to make electricity from these wasted thermal energy so we can all recover some from our electricity bill?"
- now THIS is indeed the subject of intense research worldwide
I know of 6 or so groups, and I don't know squat about all those working in silence

[Blackeagle]

Bill,

I recall reading a thread at overclockers a good while back in which a individual was interested in somehow attaching a simple finned sink to the back side of the motherboard where the socket is. He then intended to cut a hole in the back side of the case and install a fan to blow onto this sink.

Your response indicated, if I recall right, that such active cooling could indeed add a notable amount to the cooling. In fact you urged him to report back results.

Did he? Or have you yourself done any testing in this area?

Thanks

[jaydee]

Originally Posted by Blackeagle

Bill,

I recall reading a thread at overclockers a good while back in which a individual was interested in somehow attaching a simple finned sink to the back side of the motherboard where the socket is. He then intended to cut a hole in the back side of the case and install a fan to blow onto this sink.

Your response indicated, if I recall right, that such active cooling could indeed add a notable amount to the cooling. In fact you urged him to report back results.

Did he? Or have you yourself done any testing in this area?

Thanks

Speaking of such http://www.overclockers.com/tips1177/ no heatsink though. I seen this tried many times over the years and most still rely on onboard temps for measurments. Well what happens when you put a fan back there? Cools the thermister and throws off the reading.

[Blackeagle]

Originally Posted by jaydee116

Speaking of such http://www.overclockers.com/tips1177/ no heatsink though. I seen this tried many times over the years and most still rely on onboard temps for measurments. Well what happens when you put a fan back there? Cools the thermister and throws off the reading.

Yes I've read that sort of thing too.

But in this case a actual sink attached to the back of the board (I think he intended to use frag tape & the mothrboard holes around the socket, but I may be in error on this), would do more to really lower temps than just a fan blowing on the PCB.

Bill's comments were I'm sure of a positive and incouraging nature, and asked the guy to report back his results.

It was Bill's remarks that got my attention at the time. I'm hope'n he recalls the thread and any results that may have beeen reported on.

[DrMemory]

IR thermal imaging can be used to actually see the temperature differences across the MB and give you some idea of the temperatures involved. I don't know how accurately it reads temperatures but its probably good enough (a couple of degrees C ?) to do overal thermal design with.

[BillA]

no, another of those pebbles dropped in the well - without another sound

the greater the secondary path, the lesser the load on the primary - all is good
as long as the intent is to dump the heat inside the case
which is not the best idea
for the future

jd
it doesn't "throw it off", it just changes it; how good was it to begin with ?

DM
Harvard Thermal has a trace analysis program that does this and more.

[DrMemory]

I'm sure this is a very useful program for thermal design, but sometimes you just have to be there to see what's really going on. This is especially true in a complex thermal environment like a modern PC. I'd be surprised if Intel and other MB manufacturers didn't use both tools.

[Colt357tw]

ABIT used FloTherm exclusively for Thermal modeling, I had ask for IcePack and other programs, but err....price of those are pretty scary even for 5 licenses

[jaydee]

Originally Posted by unregistered

jd
it doesn't "throw it off", it just changes it; how good was it to begin with ?

True I guess. If you don't know were you started then you certainly can't tell how far off the path you have been or were you ended up.

[Blackeagle]

In a best case situation, how much of a improvement in CPU temps could be possible?
Provided efforts were made to increase secondary heat path losses with fans/sinks on the MB of course.

Estimates?

[BillA]

eh ?
its board specific