New AMD heat spreaders

[Lumen]

Hey there

Has anyone tested the new Heat spreaders amd is putting on their processors? Is there any difference in temps? Are people applying as3 and whatnot to the entire surface?

Thanks

[eXCeSS]

huh? link please?

[SK8]

i think hes talking about the 64bit chips and yes they come with ihs installed like p4's , i dont think amd is now putting ihs on there 32bit procs b/c spacing issues.

[Lumen]

Originally posted by SK8
i think hes talking about the 64bit chips and yes they come with ihs installed like p4's , i dont think amd is now putting ihs on there 32bit procs b/c spacing issues.
yesOriginally posted by eXCeSS
huh? link please? http://www20.tomshardware.com/cpu/20040106/athlon64_3400-01.html
heres the link

[ayoung]

Yep, thats a A64

[xgman]

AS5 and a light lapping seem to work fine. They will run cooler witout the top, but more prone to cracked edges.

[FizzledFiend]

Originally posted by xgman
AS5 and a light lapping seem to work fine.

this from personal experience or based on Intel results?

Originally posted by xgman
They will run cooler witout the top, but more prone to cracked edges.

more prone to cracked edges than what? one with a spreader on it or more prone than the 32 bit version?



correct me if I am wrong here, but it looked like this was just a plate (afixed) to the core? is there an article that explains it's removal? is it totally covering the core like the intels do?

seems the more layers betweenyour core and your cooling is a bad thing. I just don't see how a spreader is going to favilitate in cooling.

anywhere I can chew on some facts?

[SK8]

ihs's prevent core crackage and also save amd/intel cash for not having cracked cores sent back to them and it voids ppls warrenty in search of a few degreese lower by removing spreder.
So my advidce is if its over $300 dont f'it up!

[#Rotor]

It sounds like AMD got the power management under raps... so the Heat-spreader might actually be a blessing in disguise.

one thing I don't get with these high and mighty benchers..... way don't they pitt the Intel 64bit CPUs up against the AMD 64bit CPUs... crunching on some simple 32bit apps.... I ask you WAY????? Is there something Intel don't want us to know, or what????

[johan851]

The Intel 64-bit Itaniums are 64 bit only - they won't run 32-bit applications, I don't think. So it wouldn't really work to benchmark the systems against each other.

[#Rotor]

so let's get this straight..... AMD 64bit CPU's run 32bit code, almost better than Intel's 32bit CPU's ??? but Intel's 64bit CPU's choke completely on 32 bit code...

that kinda sux.... doesn't it?

[squeakygeek]

I wouldn't even call it a heat spreader. I mean, who do they think they're fooling?

[Diggrr]

Well that's rude. AMD stole my idea for how to make an AMD heatspreader!
I posted it here some time ago......we're being watched.

Maybe I should demand a dually system with scsi raid and 4 gigs of ram as design tribute. :D

[squeakygeek]

Originally posted by Diggrr
Well that's rude. AMD stole my idea for how to make an AMD heatspreader!
I posted it here some time ago......we're being watched.

Maybe I should demand a dually system with scsi raid and 4 gigs of ram as design tribute. :D

...And the ***holes stole my schematic for a 64 bit processor with an on-die memory controller!

[Diggrr]

No, really!

Here's the earliest post I could find, but I'd made one earlier.>LINK (http://www.ocforums.com/showthread.php?s=&threadid=122037&highlight=heatspreader)
Scroll down to my post.

One of my bud's has a TBird 1400 that I did it to to, using a 1" square and arctic alumina for the goop, with silicone around the edge. I could get a pic prolly tomorrow if you'd like. He's a bit hamfisted and was taking little chips off all the core's edges.

[#Rotor]

Well heat-spreaders have been around from long before the Athlon. I have a k6-2 233 with a heat-spreader still stuck to it....

If I'm not mistaken, I think those k6's where the first to employ such device

[Diggrr]

Yeah, I know they've been around for a while. It was kind of tounge-in-cheek when I complained about it, but I did come up with it first done in that fashion. And my AMD 400 has a heatspreader, but it's a stamped one like Intel's IHS.
Since processors can vary slightly in fab with die height, the way I (well, AMD and I) came up with varies along with it. Intel's version can leave a gap that's filled by paste, which we all know is bad....the thinner the better.

I saw one gent who'd added a copper spreader on top of a copper shim using a ton of thermal goop, but that can have the same problems.

With the ever shrinking die size, I'm not surprised they had to do something, and if the chip is $700 you can bet they don't want to RMA a buttload for cracked cores and thermal problems that hsf's aren't able to cope with in it's absence. All that wattage in an ever smaller spot on the waterblock/hsf makes cooling that much tougher.

I guess I'm happy to see that just maybe someone listens to me :D

[JDXNC]

Originally posted by #Rotor
so let's get this straight..... AMD 64bit CPU's run 32bit code, almost better than Intel's 32bit CPU's ??? but Intel's 64bit CPU's choke completely on 32 bit code...

that kinda sux.... doesn't it?

Itanium's can run 32bit code in a special emulation mode, but it really slow at it, I saw somewhere that the 800Mhz Itanium ran 32bit code at about the speed of a P2 300Mhz at best. Not to mention Itaniums have a very high heat disapation, around 130w if memory serves correct.

Regarding AMD's IHS, the new 64bit cpu's now have a thermal protection similar to the P4's and its my personal though that with the throttling effect, the IHS gives the cpu a slight buffer for the heat to get away from the core, not much, but its better than nothing. I personally don't think a P4 with its thermal throttling would survive without the IHS.

[squeakygeek]

Originally posted by Diggrr
... and thermal problems that hsf's aren't able to cope with in it's absence. All that wattage in an ever smaller spot on the waterblock/hsf makes cooling that much tougher.


Dude, the so-called "heatspreader" doesn't actually spread heat... It's just too thin to do that. All it does is introduce another thermal interface, which would only hurt performance. It's there for the sole purpose of protecting the core from physical damage.

[Diggrr]

So the thin/tiny fins on a copper heatsink are there only for the support of the fan huh? Maybe just to direct airflow onto the base plate? They certainly must be too thin to conduct heat.
Fact is, heat conduction doesn't necessarily depend on thickness. A thicker material often has benefits for the interface your using, and for a greater surface area to transfer heat to another medium, but with a piece like this you're using the flat side.
Granted, if you were trying to EVENLY conduct heat to an area of say 2" x 3" then yes, a 1/4"-3/8" plate would be ideal, but the goal isn't to do that, otherwise an ihs would be 2" square. It's job is to make the heat spread at least to the size of the OLD die so that all the OLD cooling methods work still. They'd be screwed if their new cpu required some new cooler that's the size of a tic tac.

Try this once, strip a 10" piece of 24 ga. telephone wire. Pretty thin, huh?
Hold one end in your hand, and the other in the flame of a candle. Burn? Yup, thought so.
Copper's heat conduction works just fine when it's thin doesn't it?

I'd argue that a heatspreader done Intel's old way would hinder thermal transfer. But done this way helps ensure a better thermal interface, and it does in fact spread heat. Of course they are out to protect the tiny core as well, but you cannot say that it does nothing to spread heat.
I'll give Intel credit though as I understand they are soldering the ihs into place now instead of slathering thermal goop in there like gpu and NB 'sinks still suffer from. Hopefully AMD's doing that as well.

[squeakygeek]

So the thin/tiny fins on a copper heatsink are there only for the support of the fan huh? Maybe just to direct airflow onto the base plate? They certainly must be too thin to conduct heat.

Each fin on a copper heatsink conducts a small fraction of the total heatload.
Try this once, strip a 10" piece of 24 ga. telephone wire. Pretty thin, huh?
Hold one end in your hand, and the other in the flame of a candle. Burn? Yup, thought so.
Copper's heat conduction works just fine when it's thin doesn't it?

Obviously, a piece of metal is going to conduct heat, I didn't mean it wouldn't conduct any at all. What I meant was, it's not going to conduct enough to really be of any use. Sure, the end of the wire will eventually have a high enough temperature to burn your skin, but there will be very little heat conducted. Heat and temperature are not the same thing.
Fact is, heat conduction doesn't necessarily depend on thickness.
Yes, it does.

http://www.overclockers.com/articles305/spread.jpg

http://www.overclockers.com/articles305/

[Diggrr]

Well, you've completely missed my point, so I'll move along....

[zabomb4163]

Originally posted by Diggrr


I'd argue that a heatspreader done Intel's old way would hinder thermal transfer. But done this way helps ensure a better thermal interface, and it does in fact spread heat.

http://www.xtremesystems.org/forums/showthread.php?postid=234821#post234821

I got a huge temp drop from removing my spreader and condensation issues dropped to almost "0"...

or this thread
http://www.xtremesystems.org/forums/showthread.php?s=&threadid=15991&highlight=ihs%2A
a couple guys with systems over 4ghz comment in it.

the purpose of the IHS is to protect the core. even conventional heatsinks realise a bentifit of 4 degrees when the ihs is removed

[xgman]

Originally posted by FizzledFiend


this from personal experience or based on Intel results?

Yes, done several times on the amd one.



more prone to cracked edges than what? one with a spreader on it or more prone than the 32 bit version?

than the athlon XP was. To remove it you just slice the silicon at the edge with a straight rasor (very carefully) I did 2 and it did lower temps a bit, but to easily damages, so no more. One of the problems is that the pressure exerted on the cpu is much higher than it used to be, so when you take the spreader off, all the pressure is on the exposed die. The key is the right pressure on the "lapped" spreader to get the best results with it on.



correct me if I am wrong here, but it looked like this was just a plate (afixed) to the core? is there an article that explains it's removal? is it totally covering the core like the intels do?

seems the more layers betweenyour core and your cooling is a bad thing. I just don't see how a spreader is going to favilitate in cooling.

anywhere I can chew on some facts?

[xgman]

Originally posted by squeakygeek


Dude, the so-called "heatspreader" doesn't actually spread heat... It's just too thin to do that. All it does is introduce another thermal interface, which would only hurt performance. It's there for the sole purpose of protecting the core from physical damage.

Once you take it oiff, it is actually thicker and heavier than it looks.

[nikhsub1]

The IHS's do NOTHING to help cooling. I have to disagree with you diggrr... I took my IHS off and saw a 5C drop in temps. And, the IHS does not spread the heat well at all. It is still concentrated right above the core and does not spread more than a couple of mm's.

[Diggrr]

CONGRATULATIONS Nichsub1, THAT'S MY POINT!!!!, just a mm or two is all it needs to make it's hot zone the same size as the older t-bred core!
I'm not saying that entire heatspreader is really worth a damn, I'm saying that it spreads heat to at least the same area as the older, larger cores (as well as protecting it).
If you put the same ammount of heat in a smaller and smaller area, your old cooling solutions will fail to perform so nicely.
Put 150 watts into a 2 foot square copper plate, and you change the plate's temp very little.
Put that same 150 watts into a square millimeter, and you're welding!!!

Yes additional interfaces can hurt cooling, no arguement there. You wanna save 4 degrees by removing it?, fine! But AMD now has the means to send your RMA to the skip. Will your old hsf's and waterblocks work well on it...oh yeah!

Still amazes me that I have to yell (if only in print) to get this point across. I thought it'd be easier idea to pick up on for this crowd.

Oh, and one thing not mentioned yet, Direct die (well, spreader)anyone?

[squeakygeek]

Originally posted by Diggrr
CONGRATULATIONS Nichsub1, THAT'S MY POINT!!!!, just a mm or two is all it needs to make it's hot zone the same size as the older t-bred core!
I'm not saying that entire heatspreader is really worth a damn, I'm saying that it spreads heat to at least the same area as the older, larger cores (as well as protecting it).
If you put the same ammount of heat in a smaller and smaller area, your old cooling solutions will fail to perform so nicely.
Put 150 watts into a 2 foot square copper plate, and you change the plate's temp very little.
Put that same 150 watts into a square millimeter, and you're welding!!!

Yes additional interfaces can hurt cooling, no arguement there. You wanna save 4 degrees by removing it?, fine! But AMD now has the means to send your RMA to the skip. Will your old hsf's and waterblocks work well on it...oh yeah!

Still amazes me that I have to yell (if only in print) to get this point across. I thought it'd be easier idea to pick up on for this crowd.

Oh, and one thing not mentioned yet, Direct die (well, spreader)anyone?

Ok, let's say it does increase the area to the size of the old core... how is this advantageous? We all know that it only increases temps.

Direct die/heatspreader is also pointless when we have today's water-blocks. In the early days of water cooling, a water-block was no better than an open chamber, in which case the base of the waterblock was simply another thermal interface, and didn't help any, and so direct die got rid of this extra interface. Nowadays waterblocks are designed to greatly increase the surface area of copper exposed to water (microchannels, pins, cascade-like holes, and the like). Going direct-die with all these options is just silly. It will have less cooling power, and your chip will die.

Still amazes me that I have to yell (if only in print) to get this point across. I thought it'd be easier idea to pick up on for this crowd.

Don't blame it on us. Perhaps you need to explain things a bit better.

[sandman001]

Originally posted by squeakygeek

Direct die/heatspreader is also pointless when we have today's water-blocks. In the early days of water cooling, a water-block was no better than an open chamber, in which case the base of the waterblock was simply another thermal interface, and didn't help any, and so direct die got rid of this extra interface. Nowadays waterblocks are designed to greatly increase the surface area of copper exposed to water (microchannels, pins, cascade-like holes, and the like). Going direct-die with all these options is just silly. It will have less cooling power, and your chip will die.

I don't quite follow this, copper is still there, no matter the design of the block. It's still another thermal barrier.

As such, Direct Die would help temps. The heat still has to transfer through the same area as the water would touch (AS3). It would also elimate two thermal barriers.

[wildfrogman]

Waterblocks tend to handle small focused heatloads much better than heatsinks do. Also they usually are not clamped down nearly as hard compared to the large and heavy heatsinks for the p4/athlon64 type cpu. I dont remember exactly what psi the intel IHS/heatsink clamp was designed for but it was around 60~80psi if not greater.

P4 boards at a time had MAJOR trouble with certain boards warping and bending because of the clamping pressure and sometimes causing the board to die because they lacked a metal bracket like server boards~thermalright/taisol had. The main feature of the IHS is to allow heatsinks to be under much greater pressure that leads to better performance. The IHS stops amd/intel worrying about chipped cores and also gives the cpu a bit of mass attached to the chip just incase sombody tries to start the system up without a heatsink. OR....with no thermal paste and have it actually still work halfway decent. I have seen that before were people didnt use thermal paste yet they operated the system for probably a few weeks before they talked to me saying they thought their cpu temp was a bit high.

[squeakygeek]

Originally posted by sandman001


I don't quite follow this, copper is still there, no matter the design of the block. It's still another thermal barrier.

As such, Direct Die would help temps. The heat still has to transfer through the same area as the water would touch (AS3). It would also elimate two thermal barriers.

I will try to clarify.

Today's waterblock designs greatly overcome the performance drop of having another thermal interface.

With a well designed block, heat is transfered from the copper to the water over a large surface area (fins or otherwise). With direct die, you don't have the benefit of the large surface area.

Direct die improved performance before we had well-designed blocks.

[fafnir]

http://www.ocforums.com/showthread.php?threadid=251653

btw, use floss

[fafnir]

http://www.theforumisdown.com/uploadfiles/1203/Top.jpg

http://www.theforumisdown.com/uploadfiles/1203/Back.jpg

http://www.theforumisdown.com/uploadfiles/1203/k6naked.jpg

http://www.theforumisdown.com/uploadfiles/1203/Bottom.jpg

http://www.theforumisdown.com/uploadfiles/1203/Top1.jpg

[fafnir]

http://www.theforumisdown.com/uploadfiles/1203/Front1.jpg

and thats my 5900xt with the nvidia heatspreader popped

i'd do some pictures of a lapped barton core and another lapped 9800, but its just a pain to take apart the things just for photos,...

but otherwise, its a 3-5 degree drop (idle less, load a tad bit more)

edit: load dropped 5, idle dropped 3

[fafnir]

http://www.theforumisdown.com/uploadfiles/1203/M0.jpg

and (thanks to SilverSinkSsam), ever wondered what happens if the thing DOESN'T WANT TO SEPERATE?



(intel M0 p4 bonded/ee edition pictured)