Small cooling surfaces going away?

winewood · 06-28-2003, 07:49 AM

I was looking at THIS I was left with a question of if the chip development is leading away from small core area cooling, vs the integrated heat spreader.

If this is true, then it leaves all the non-tec blocks with some disadvantages. Will most of the blocks designed in here be useless pretty soon, unless they find a way to cool a larger area?

I have seen some comments, but don't recall seeing a thread adressing this movement in chip design and the resulting fallout in most concepts of design that many here have come to swear by. What blocks will be rendered useless and what designs will not be able to be used for larger area cooling?

TheDanMan · 06-28-2003, 09:36 AM

The AMD opteron comes with an aluminum heat spreader integrated into it. Yes unfortunatly this is the way we are headed, however, I expect there will always be those l33t enough to come here and post how they have removed the ISH from their precious proccesors.

winewood · 06-28-2003, 10:31 AM

The l33t..

true, but the real genius of a good fix is making it accessable to as many others, with minimal effort. Ill take time to point out #Rotor whose simple design is genius. Cathars channel blocks are another one that is easy and accessable.
It's like every block that only supports a small area will have to be ported. Those heat spreaders on the core usually are very efficiently placed on (so Ive heard) and alternate meathods usually don't really achieve the same joint efficiency.
Alot of people are going to have to start over..

Cathar · 06-28-2003, 06:49 PM

I think you're slightly exaggerating the problem.

The heat-spreader doesn't result in a spreading of the heat far and wide. The bulk of the heat still gets concentrated directly above the CPU core. The added thickness of the heat-spreader means that the heat spreads by an extra mm or two in each direction. The strong centrally focussed designs hold up very well in these scenarios, especially if they are cooling ~20x20mm areas. All the blocks I make focus on cooling a ~20-30mm x 16-17mm area, and this is still the right thing to do for both heat-spreaders and without heat-spreaders.

winewood · 06-28-2003, 08:16 PM

The right thing to do, but it in many cases negates the benifits of mm from base cooling that delivers such good results because they are so focused on their specific job. Im not here to start a flame post in any way, but point out that a heat spreader will do just that and the core heat area will now be bigger.

I am thinking the proper course is not to defend what already is but to learn and adjust for this. If not learn from the newer environment that we are working then assume the arguement of keeping your older block and ask how many degrees different is it now, and is this enough to change peoples mindsets toward focusing on a couple of types of non-tec blocks.

I feel that this movement narrows the gap between the non directed blocks as the chip manufactures seek to find ways to keep aircooling an option.

Is this a logical observation? Once the larger heat spread area is accounted for, it will become a crutch that manufactures can use to spread out the heat or make the cores larger. Then we must address this topic later if not now.

jaydee · 06-28-2003, 09:48 PM

Don't know. here is the next "Duron" which is called ther Thorton. Basically it is a Barton with half the ondie memory. Kinda stoked up about these! Hope they hit the market.
http://www.hexus.net/
I bet all the Hammers/Opterons will have these IHS though. Already trying to figure a way to Direct IHS water cool it. There might be enough room on the Opteron IHS to seal it on the IHS itself. Hard to tell though without a scaled pic.

Cathar · 06-28-2003, 10:02 PM

Quote:

Originally posted by winewood
The right thing to do, but it in many cases negates the benifits of mm from base cooling that delivers such good results because they are so focused on their specific job. Im not here to start a flame post in any way, but point out that a heat spreader will do just that and the core heat area will now be bigger.

I am thinking the proper course is not to defend what already is but to learn and adjust for this. If not learn from the newer environment that we are working then assume the arguement of keeping your older block and ask how many degrees different is it now, and is this enough to change peoples mindsets toward focusing on a couple of types of non-tec blocks.

I feel that this movement narrows the gap between the non directed blocks as the chip manufactures seek to find ways to keep aircooling an option.

Is this a logical observation? Once the larger heat spread area is accounted for, it will become a crutch that manufactures can use to spread out the heat or make the cores larger. Then we must address this topic later if not now.

The gap is "narrowed", but what is important here is how much it is narrowed. At present IHS thicknesses (~1-1.5mm) the gap really isn't narrowed to any major degree that I can see. The IHS doesn't laterally conduct the heat far enough to make a major impact to the high focus designs. The high focus designs that I use are already made in such a way to account for IHS use. If I would guarantee that the IHS was not there, their focus would be even tighter than at present.

What is of critical important though is test-bed setup. Too many independent testbeds create a heat load that is distributed evenly across a simulated IHS area, and this is wrong. An IHS CPU has a small die with a separate thermal junction to a thin IHS.

I fear that what may happen is bogus test-bed setups are used that drive customers away from the designs that do tackle the problem appropriately.

winewood · 06-29-2003, 07:30 AM

As stated there is no way to determine at present how much heat is distributed away from the original core area with that spreader and the new core size and voltages. Knowing this it is a fallacy imho at present to say, who is going to determine what testbed to use and THEN dermine which setup is best.
Now, we could wait for the chip to be tested by someone of repute, or we could just start to adjust our focus to a slightly larger core footprint, which in likely hood will happen soon if not in this chip. A heat spreader will ultimately do just that if they are used in future chips. How could it not?
Of course you can say, we don't have a test to determine that we need to focus our minds on a larger core heat area. But in the same logic I don't feel that anyone can assume with the same logic that we can just keep what measurements and focus that we have now on the smaller one.
Cathar: if you were referring to the WW, I do not understand how this is accounting for a larger heat area. If you feel it does, please explain. If you were referring to micro cup design I think feel that that design would be better suited to this "trend?". Your ideas and explinations are as always valued. Thanks
*Back to topic*
What are some good ways to port our existing designs that work to a larger area?

Cathar · 06-29-2003, 08:10 AM

Quote:

Originally posted by winewood
As stated there is no way to determine at present how much heat is distributed away from the original core area with that spreader and the new core size and voltages.

Not quite sure why you're saying this. All the information is available at AMD's web-site. The parent link is here:

http://www.amd.com/us-en/Processors/...9_9003,00.html

Cathar · 06-29-2003, 08:17 AM

85W peak thermal power.

The heat-spreader is 37.5 x 37.5mm in size.

The heat-spreader seems to be around 1.5mm thick.

Oh, found elsewhere that the die size is 193mm^2 - wow - that's a pretty big die.

Edit 2: Could not find an actual statement of the physical X-Y dimensions of the die size of the Opteron, but found a picture of the die here:

http://www.sandpile.org/impl/pics/am...013_1m_ovr.jpg

The die is roughly 3:2 in size for the edges, slightly favoring the longer edge. My rough estimate is the actual die is about 11x17mm in size. Given a 1.5mm thick heat-spreader, this would mean that the optimal area to focus the cooling on would be about a 14x20mm rectangle. Heat conducting outside of that region would be minimal.

winewood · 06-29-2003, 08:58 AM

QUOTE "As stated there is no way to determine at present how much heat is distributed away from the original core area with that spreader and the new core size and voltages. "

I have found personally that there is no way to determine how a new chip and new heat spreader dissipates heat from a picture or paragraph. I also have found no way to guess as to how a waterblock designed for cores of X size using heat spreaders of X composition at X voltage will effect it and how.
Cathar, in your case your ww block did awsome because it was designed for a very narrow definition of tight constraints to work in a well placed way. Kudos for your good work. Take away these somewhat, and I am left guessing as to if its the appropriate without modification later down the road. This is what this topic was meant to address.
However, I also do not know how your waterblock jet or channels designed to be (just guessing 1mm from core) will behave with an additional 1.5 mm of heat spreader under it. What you have left is a WW that has a new base thickness and another joint to transfer heat. This sounds like something that could effect it... shouldn't it?
Those numbers from amd are real, but I personally do not understand how it can directly relate to thermal properties and if anyone can judge from schematics how a block will do or wont from the equations. Im not that good, and not about to guess how it benifits or can subtract from an existing block not tooled to these new factors. To dismiss or accept a block working outside of its narrow confines takes either faith, experience from similar tests, or testing when it comes out.
I was hoping this thread could beat the new ideas coming with some preconcieved ideas how we could make our "champs" concepts to deal with the new changes. Not a threat directed at Cathars block, and frankly this topic wasn't expressly to even suggest that any one block couldn't work with this new core covering. However, now that you mention it, I would think the new core may not allow the same numbers as we have come to expect from ANY block.
In fact with these new core coverings and specs, it would seem logical to test as to the efficiencies and see if the carry though. Doesn't this seem prudent to address?

Edit: Ya know.. i re-read this and it sounds confrontational, arguementitive. I don't want it to sound this way. I just wanted to discuss what this means for the future. Readers.. cathar, please take it in thie spirit. I'm not saying you are wrong cathar, just help me explain how the efficiencies of past tests could pass on to this one, or the changes wouldn't need to be accounted for.

Cathar · 06-29-2003, 09:43 AM

winewood, I understand where you're coming from.

My argument is this, and based upon my experiences with my block designs on P4's with heat spreaders, which is why I keep referring to them as examples.

Before we blindly assume that we simply MUST make the area of cooling focus larger, we must first establish that such is even needed. I'm just looking at the various aspects of the Opteron implementation, and right now I cannot see a necessity for the assumption that the focus simply definitely must be larger, based upon my experiences with the P4's and their heat spreaders, and then extrapolating from there and given what we do already know about the Opteron.

That's all I'm trying to say, and that's all the value that I'm trying to add here. If you or anyone else believes my opinion to be wrong, then so be it. It's just my opinion.

My personal opinion is that a heat-spreader really isn't what it's name implies. It's a CPU core protection device. Given adequate focussed cooling, it doesn't really spread the heat, and by heat, I mean wattage.

I'll leave the thread be now.

winewood · 06-29-2003, 10:08 AM

Now that I can understand

And frankly that makes me happier to belive that, due to alot of hard work that would have needed to be tweaked for the zillionth time. Thanks for your help Cathar.
Does anyone else have a differing opinion? I always appreciate different points of view.

bigben2k · 06-29-2003, 10:26 AM

Hum... from what I remember, Cathar designed WW with the assumption that the heat would spread no more than 2mm outside the dimensions of the core, at least not with any significance, for the purpose (from his simulation?). Knowing that the baseplate is pure copper, less than 1 mm thick, that should give you an idea of what to expect from the IHS.

What seems to be ignored is that, although there will be a performance degradation from the additional TIM joint, there will also be an increased thermal gradient right through that IHS.

In other words, just pop the darn thing off, and be done with it! We already know that the P4's IHS will add a couple of degrees.

But to answer the original question: All commercially available blocks will still function properly, for the larger die of the new AMD processor. Most blocks have a baseplate that's just too thick, so it's not going to make a difference. In the case of a Maze2 for example, the heat might start spreading into the channel wall, which would actually help performance, but because of the massive bp, it won't be measurable.

WW might suffer a bit, but I think it'll still do well. Cascade might also suffer a bit, but given that it's already the best design, a small decrease in performance will still make it perform extraordinarily well. These two blocks are minimally designed for the size of the core.

jaydee · 06-29-2003, 11:03 AM

Quote:

Originally posted by bigben2k
Hu

In other words, just pop the darn thing off, and be done with it! We already know that the P4's IHS will add a couple of degrees.

Can you just "pop" it off an AMD CPU though. Anyone see an example of this. It looks from the pics it will be pretty hard to get that thing off without taking the Die with it....

winewood · 06-29-2003, 11:12 AM

QUOTE]Cascade might also suffer a bit, but given that it's already the best design,[/quote]
the Cascade beat the ww??

jaydee · 06-29-2003, 11:16 AM

Quote:

Originally posted by winewood

the Cascade beat the ww??

It is all in Cathars thread. By 2.5C last he stated.

bigben2k · 06-29-2003, 11:21 AM

Linky:
http://forums.procooling.com/vbb/sho...&threadid=6666

MadDogMe · 06-30-2003, 04:28 AM

GPU/NB's are the same, even though the size of the 'package' is say 30x30mm it's only the very centre (about 10mmx10mm) that heat is generated from. Concentrate on that central part and you'll cool the chip better (or at least as well) than a '30x30mm area~optimised' block...

06-28-2003, 07:49 AM	#1
winewood Cooling Savant Join Date: Nov 2002 Location: in my chair Posts: 574	Small cooling surfaces going away? I was looking at THIS I was left with a question of if the chip development is leading away from small core area cooling, vs the integrated heat spreader. If this is true, then it leaves all the non-tec blocks with some disadvantages. Will most of the blocks designed in here be useless pretty soon, unless they find a way to cool a larger area? I have seen some comments, but don't recall seeing a thread adressing this movement in chip design and the resulting fallout in most concepts of design that many here have come to swear by. What blocks will be rendered useless and what designs will not be able to be used for larger area cooling? __________________ -winewood-

06-28-2003, 09:36 AM	#2
TheDanMan Cooling Neophyte Join Date: Jun 2003 Location: Oklahoma Posts: 11	The AMD opteron comes with an aluminum heat spreader integrated into it. Yes unfortunatly this is the way we are headed, however, I expect there will always be those l33t enough to come here and post how they have removed the ISH from their precious proccesors. __________________ danny@edlangroup.com monthly LANs every 4th sat www.edlangroup.com Edmond, Oklahoma

06-28-2003, 10:31 AM	#3
winewood Cooling Savant Join Date: Nov 2002 Location: in my chair Posts: 574	The l33t.. true, but the real genius of a good fix is making it accessable to as many others, with minimal effort. Ill take time to point out #Rotor whose simple design is genius. Cathars channel blocks are another one that is easy and accessable. It's like every block that only supports a small area will have to be ported. Those heat spreaders on the core usually are very efficiently placed on (so Ive heard) and alternate meathods usually don't really achieve the same joint efficiency. Alot of people are going to have to start over.. __________________ -winewood-

06-28-2003, 08:16 PM	#5
winewood Cooling Savant Join Date: Nov 2002 Location: in my chair Posts: 574	The right thing to do, but it in many cases negates the benifits of mm from base cooling that delivers such good results because they are so focused on their specific job. Im not here to start a flame post in any way, but point out that a heat spreader will do just that and the core heat area will now be bigger. I am thinking the proper course is not to defend what already is but to learn and adjust for this. If not learn from the newer environment that we are working then assume the arguement of keeping your older block and ask how many degrees different is it now, and is this enough to change peoples mindsets toward focusing on a couple of types of non-tec blocks. I feel that this movement narrows the gap between the non directed blocks as the chip manufactures seek to find ways to keep aircooling an option. Is this a logical observation? Once the larger heat spread area is accounted for, it will become a crutch that manufactures can use to spread out the heat or make the cores larger. Then we must address this topic later if not now. __________________ -winewood-

06-29-2003, 07:30 AM	#8
winewood Cooling Savant Join Date: Nov 2002 Location: in my chair Posts: 574	As stated there is no way to determine at present how much heat is distributed away from the original core area with that spreader and the new core size and voltages. Knowing this it is a fallacy imho at present to say, who is going to determine what testbed to use and THEN dermine which setup is best. Now, we could wait for the chip to be tested by someone of repute, or we could just start to adjust our focus to a slightly larger core footprint, which in likely hood will happen soon if not in this chip. A heat spreader will ultimately do just that if they are used in future chips. How could it not? Of course you can say, we don't have a test to determine that we need to focus our minds on a larger core heat area. But in the same logic I don't feel that anyone can assume with the same logic that we can just keep what measurements and focus that we have now on the smaller one. Cathar: if you were referring to the WW, I do not understand how this is accounting for a larger heat area. If you feel it does, please explain. If you were referring to micro cup design I think feel that that design would be better suited to this "trend?". Your ideas and explinations are as always valued. Thanks Back to topic What are some good ways to port our existing designs that work to a larger area? __________________ -winewood- Last edited by winewood; 06-29-2003 at 07:37 AM.

06-28-2003, 06:49 PM	#4
Cathar Thermophile Join Date: Sep 2002 Location: Melbourne, Australia Posts: 2,538	I think you're slightly exaggerating the problem. The heat-spreader doesn't result in a spreading of the heat far and wide. The bulk of the heat still gets concentrated directly above the CPU core. The added thickness of the heat-spreader means that the heat spreads by an extra mm or two in each direction. The strong centrally focussed designs hold up very well in these scenarios, especially if they are cooling ~20x20mm areas. All the blocks I make focus on cooling a ~20-30mm x 16-17mm area, and this is still the right thing to do for both heat-spreaders and without heat-spreaders.

06-29-2003, 08:17 AM	#10
Cathar Thermophile Join Date: Sep 2002 Location: Melbourne, Australia Posts: 2,538	85W peak thermal power. The heat-spreader is 37.5 x 37.5mm in size. The heat-spreader seems to be around 1.5mm thick. Oh, found elsewhere that the die size is 193mm^2 - wow - that's a pretty big die. Edit 2: Could not find an actual statement of the physical X-Y dimensions of the die size of the Opteron, but found a picture of the die here: http://www.sandpile.org/impl/pics/am...013_1m_ovr.jpg The die is roughly 3:2 in size for the edges, slightly favoring the longer edge. My rough estimate is the actual die is about 11x17mm in size. Given a 1.5mm thick heat-spreader, this would mean that the optimal area to focus the cooling on would be about a 14x20mm rectangle. Heat conducting outside of that region would be minimal. Last edited by Cathar; 06-29-2003 at 08:36 AM.

06-29-2003, 08:58 AM	#11
winewood Cooling Savant Join Date: Nov 2002 Location: in my chair Posts: 574	QUOTE "As stated there is no way to determine at present how much heat is distributed away from the original core area with that spreader and the new core size and voltages. " I have found personally that there is no way to determine how a new chip and new heat spreader dissipates heat from a picture or paragraph. I also have found no way to guess as to how a waterblock designed for cores of X size using heat spreaders of X composition at X voltage will effect it and how. Cathar, in your case your ww block did awsome because it was designed for a very narrow definition of tight constraints to work in a well placed way. Kudos for your good work. Take away these somewhat, and I am left guessing as to if its the appropriate without modification later down the road. This is what this topic was meant to address. However, I also do not know how your waterblock jet or channels designed to be (just guessing 1mm from core) will behave with an additional 1.5 mm of heat spreader under it. What you have left is a WW that has a new base thickness and another joint to transfer heat. This sounds like something that could effect it... shouldn't it? Those numbers from amd are real, but I personally do not understand how it can directly relate to thermal properties and if anyone can judge from schematics how a block will do or wont from the equations. Im not that good, and not about to guess how it benifits or can subtract from an existing block not tooled to these new factors. To dismiss or accept a block working outside of its narrow confines takes either faith, experience from similar tests, or testing when it comes out. I was hoping this thread could beat the new ideas coming with some preconcieved ideas how we could make our "champs" concepts to deal with the new changes. Not a threat directed at Cathars block, and frankly this topic wasn't expressly to even suggest that any one block couldn't work with this new core covering. However, now that you mention it, I would think the new core may not allow the same numbers as we have come to expect from ANY block. In fact with these new core coverings and specs, it would seem logical to test as to the efficiencies and see if the carry though. Doesn't this seem prudent to address? Edit: Ya know.. i re-read this and it sounds confrontational, arguementitive. I don't want it to sound this way. I just wanted to discuss what this means for the future. Readers.. cathar, please take it in thie spirit. I'm not saying you are wrong cathar, just help me explain how the efficiencies of past tests could pass on to this one, or the changes wouldn't need to be accounted for. __________________ -winewood- Last edited by winewood; 06-29-2003 at 09:08 AM.

06-29-2003, 09:43 AM	#12
Cathar Thermophile Join Date: Sep 2002 Location: Melbourne, Australia Posts: 2,538	winewood, I understand where you're coming from. My argument is this, and based upon my experiences with my block designs on P4's with heat spreaders, which is why I keep referring to them as examples. Before we blindly assume that we simply MUST make the area of cooling focus larger, we must first establish that such is even needed. I'm just looking at the various aspects of the Opteron implementation, and right now I cannot see a necessity for the assumption that the focus simply definitely must be larger, based upon my experiences with the P4's and their heat spreaders, and then extrapolating from there and given what we do already know about the Opteron. That's all I'm trying to say, and that's all the value that I'm trying to add here. If you or anyone else believes my opinion to be wrong, then so be it. It's just my opinion. My personal opinion is that a heat-spreader really isn't what it's name implies. It's a CPU core protection device. Given adequate focussed cooling, it doesn't really spread the heat, and by heat, I mean wattage. I'll leave the thread be now.

06-29-2003, 10:08 AM	#13
winewood Cooling Savant Join Date: Nov 2002 Location: in my chair Posts: 574	Now that I can understand And frankly that makes me happier to belive that, due to alot of hard work that would have needed to be tweaked for the zillionth time. Thanks for your help Cathar. Does anyone else have a differing opinion? I always appreciate different points of view. __________________ -winewood-

06-29-2003, 10:26 AM	#14
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Hum... from what I remember, Cathar designed WW with the assumption that the heat would spread no more than 2mm outside the dimensions of the core, at least not with any significance, for the purpose (from his simulation?). Knowing that the baseplate is pure copper, less than 1 mm thick, that should give you an idea of what to expect from the IHS. What seems to be ignored is that, although there will be a performance degradation from the additional TIM joint, there will also be an increased thermal gradient right through that IHS. In other words, just pop the darn thing off, and be done with it! We already know that the P4's IHS will add a couple of degrees. But to answer the original question: All commercially available blocks will still function properly, for the larger die of the new AMD processor. Most blocks have a baseplate that's just too thick, so it's not going to make a difference. In the case of a Maze2 for example, the heat might start spreading into the channel wall, which would actually help performance, but because of the massive bp, it won't be measurable. WW might suffer a bit, but I think it'll still do well. Cascade might also suffer a bit, but given that it's already the best design, a small decrease in performance will still make it perform extraordinarily well. These two blocks are minimally designed for the size of the core. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

06-29-2003, 11:12 AM	#16
winewood Cooling Savant Join Date: Nov 2002 Location: in my chair Posts: 574	QUOTE]Cascade might also suffer a bit, but given that it's already the best design,[/quote] the Cascade beat the ww?? __________________ -winewood-

06-29-2003, 11:21 AM	#18
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Linky: http://forums.procooling.com/vbb/sho...&threadid=6666 __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

06-30-2003, 04:28 AM	#19
MadDogMe Thermophile Join Date: Aug 2002 Location: Just shut up ;) ... Posts: 1,068	GPU/NB's are the same, even though the size of the 'package' is say 30x30mm it's only the very centre (about 10mmx10mm) that heat is generated from. Concentrate on that central part and you'll cool the chip better (or at least as well) than a '30x30mm area~optimised' block...

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)