Using a CPU without thermal goop

[KevinF]

What would happen if I took a HSF and processor and just rubber banded them together without thermal goop? Would it fry? I know that metal->metal would be best for a HSF and processor, but that they aren't flat enough. I just ordered some Arctic Silver, but I want to put together my new system now. So anyways, would there me enough heat transfer if I just rubber banded the too together, or would it fry?

EDIT: I've been advised by someone I trust to use white toothpaste, not green but white. Thanks anyway.

 

[Mikewarrior2]

what kind of cpu do you want to try without grease?


Mike

 

[GaryTcs]

toothpaste? rubber band? yikes!! Isn't there a Radio shack in your town? Or a pad on the bottom of the heatsink? I hope it's not an AMD socket processor!

 

[lane42]

Radio Shack sells a small 6.5 gram tube of silicone based Heat Sink Compound for a few bucks. Part #276-1372

 

[paulip88]

I don't know how you can mount a heatsink onto a socket CPU with a rubberband, but I can see how you can do it with a slot CPU.

However, slot based CPUs almost always run too hot for this. Pentiums for the most part run fine without thermal paste, but PIIs, PIIIs, Celerons, and especially Athlon TBirds and Durons need something there.

I also would not use the toothpaste. Toothpaste will dry out and and shrink, which are both bad properties of thermal paste.

As for the radio shack stuff, its ok, but its not recommended for stressful environments like the new coppermines or socket a chips.

 

[lane42]

Yes, its just a temp fix until you get the good stuff

 

[KR]

Don't use toothpaste of any color. Though the most common color for heatsink compound is white it is not the same as toothpaste.

Rubber band? Your heatsink should have some sort of clip mechanism to attach it to the processor (if it's a slot 1 (Intel) or slot A (AMD)) or a clip to attach it to the socket if it's a socket 370 (Intel) or socketA (AMD). Why would you use a rubberband? Use the clip that is intended for mounting the Heatsink.

DO NOT do a bare metal to metal joint. The imperfections in the processor and heatsink surfaces will leave an air layer between the pieces. Air is a pretty good insulator so you will be only getting a small portion of the heat transferred away from the processor without heatsink compound to fill the gaps with a heat conductive material.

The only reason to use the toothpaste w/rubberband method you've described would be to destroy a perfectly good processor. If that is your desire and wish, go for it. It's your money.

 

[paulip88]

Forgot to mention this, but a rubberband just isn't a good idea. Superglue would be better if you want to be real cheap, but almost all heatsinks some with clips (CPU heatsinks at least).

The problem with rubberbands is that they don't handle heat well at all. They crack and eventually break. When this happens, you'll lose all the pressure on the heatsink, and depending on how the CPU is oriented, the heatsink might just fall off. Also, the rubberband probably does not provide adequate pressure in the first place.

 

[EvilDonnyboy]

if it has a thermal pad of somekind, it will be sufficient for low end coppermines at defualt speeds. I've tried this quite a few times, and many OEM system sellers do this. It should probably work fine in higher-end coppermines and athlons, but i don't have the persomal experience. Radio Shack goo is adequet for defualt speeds of all consumer processors.

but if ur overclocking, invest in some quality thermal goo. Name brand white goo is still good compared to expensive, online only arctic silver.

 

[longhorn]

The difference between a thermal pad and Radio Shack thermal
paste is about 2C. The difference between Radio Shack and Artic
Silver is about 1C. Maybe that degree will allow you to run
your CPU a little faster, but it's pretty doubtful. Running
with no pad or paste, however, is a very poor idea.

I work on Supercomputers for a living, some with fans that
could literally knock you over(at least if you'd had a few
beers). We use thermal pads, because of the manufacturing concerns
with getting a nice thin layer of paste. Anything thicker than
a few sheets of paper, and the paste performance is inferior
to a pad.

Processor manufacturers guarantee that the processor will work
at a certain temperture and a certain speed. We test the daylights
out of systems, well beyond their temperature ranges, to ensure
that we have adequate guardband in our design. The percent of
processor that crap out with a step up of a degree or 2 is far
less than 1%.

 

[R0b0tN1k]

Don't listen to these kids. I've stuck heatsinks on with all sorts of things and it really doesn't matter. Elastics are great, but get nice strong ones.

 

[DaddyG]

Longhorn,

I must disagree with your Radio Shack statement. The high clip pressure and high heat of a BIRD (60watts plus when overclocked) is just too much for RS grease. Over a relatively short period of time, the grease will dry out and the thermal interface will be degraded.
PCTC has all the properties required to prevent this. Its thixotropic, it fills small gaps when it 'changes' and it is not silicone based which causes dry out over time. Also PCTC is the ONLY compound approved and recommended by AMD.

 

[Mikewarrior2]

The difference between RS grease and Arctic silver, when properly applied, is NOT 1C. On socket A platforms, it may appear to be 1c, because of very faulty temp reading conditions, but there is no way, that when both are properly applied, that AS is not at least 3C(CORE TEMP) lower than RS grease.

Yes, some thermal pads, particularly hte Chometrics (can't remember part number) socket A pads do perform very well, they are still not better than Arctic Silver.


Mike

 

[lane42]

Longhorn, any chance getting one of those supercomputers to run seti for us over at Team AnAndTech.

 

[longhorn]

My experience is with million dollar supercomputers, put
into thermal chambers, with many, many thermalcouples
attached, including the thermistor-in-the-hole to get
the core temperature. These machines have processor
that use more power than any Duron or Tbird.

Of course, we don't buy thermal paste at Radio Shack, but
all of the inexpensive zinc oxide pastes are about the same.
We also haven't used Artic Silver, but we certainly have used
high-end copper-based paste. Maybe Artic Silver is something
even better. I am a design engineer, and cooling is only
a small part of my job. I could believe that you could get
another degree or 2 with better technology. We gave up on paste
a few years ago when we found that the card assembly technicians
found it too difficult to get a consistently thin layer of paste.
The CORE temperature difference between a good thermal pad and
paste was only a couple of degrees, so that was the way we went.

My major point was that even 5 degrees C will rarely be the
difference between success and failure at overclocking. When
we test computers, we stairstep up the temperature until it breaks.
We take the computer to the maximum temperature, and bring
down the voltage until something fails or raise the frequency
until something fails.

You may occasionally find that you could get another 50Mhz
out of a processor by bring the temperature down 5C.
I would be surprised if you ever got more than that. In most
cases, the 5C would make no difference at all. Temperature is
just one factor.

I have never seen a case of thermal paste drying out and
negatively affecting its cooling abilities. I have run
systems where the procesor core temperature went over 100C,
and the thermal properties of the paste remained stable.
You have to scrape off the old paste and apply new when
the heatsink is removed. As long as the heatsink stays
in place, the gaps are filled, and you're OK.

You guys obviously take this real seriously, and have some
strong opinions. My opinion, however, is that all these
exotic heatsinks and expensive pastes that bring the
temperature down a few degrees are a poor investment.
I'd rather have a $20 heatsink, a $3 tube of thermal paste,
and $40 in my pocket, then a $50 heatsink,a $13 tube of thermal,
paste and a procesor that runs 5C cooler or maybe 50Mhz faster.

Of course, being an engineer, I also complete fail to grasp
why anyone would care about the physical appearance of a
motherboard or IDE cable, so I am obviously out of step
with some people here.

 

[Mikewarrior2]

It isn't about appearence when it comes to modifying ide cables, its about increasing case airflow. Guess you didn't see that benefit of having rounded ide-cables.

Also, when you comment on something, you should have some experience with it first. For you to make bold assumptions about the performance of arctic silver without ever even trying it is very idiotic. You can't compare RS Grease or any grease to something you have never tried.



Mike

 

[EvilDonnyboy]

In my experience, the difference between a graphte thermal pad and RS goo was 20 degress celcius. that was from 65 average under load to 45 degress celcius. Quite a large difference. The difference between some high-quality white goo and RS goo was 5 degress celcius, although this difference could be attributed to better application with the quality stuff. But the quality stuff won't dry out as quickly as the RS stuff.

but the RS stuff is still good. It has lasted 6 months of heavey use on an OC'd computer, when i took it off to apply the quality goo, i was as gooey as it waz when it can out of the tube. And if it dries out, the zinc-oxid is still left behind to do all the heat-transfering. the silicon stuff is just there to help with the application, it itself is not so great at heat conduction.

 

[longhorn]

Hey Mike--take one of your Valiums, quick. Geez.
If it's that important to you, you can be right, OK?
Why don't you tell us all your real-world experience,
as opposed to what you read on some web page?

PS I did not say airflow wasn't important, I
said appearance wasn't important TO ME.

 

[Zoltarc]

You CAN run a SocketA CPU with just a die to HSF contact, ie no thermal paste. I'm running like that right now. I ran out of paste and haven't got round to getting some more. Its safe but make sure that you use the HSF clip not a rubber band.

 

[Mikewarrior2]

Okay, jack@ss, you want real world experience.

P3-650@982, 1.8v. 48C SIlicon Grease, 42C Arctic Silver Grease, Chometrics PCTC Pad 45C.

P3-600@800, GORB, RS Grease 44C, AS 38C

P3-600@812, GORB, Melcor Aluminum Grease 42C, AS 39C

P3-700E @ 933, Alpha PAL6035, melcor aluminum 41C, AS 38C

P3-700E @ 966, Alpha PAL35T, Melcor Aluminum 45C, AS 40C

Duron 700 @ 1ghz, PAL6035, RS Grease 48C, Arctic Silver, 45C* Note inaccurate readings because of socket A platforms.

In addition to these, I have used it on almost 70 CPUs. Every single one has resulted in a temp drop, regardless of platform. Again, for you to claim a 1C Temp difference from Arctic Silver to Radio Shack grease is a major Assumption on your part. And because you have never tried arctic silver, you have Zero right to make that claim.

I don't take stuff from websites outright. All people are capapble of errors, even websites like Anandtech. Even people like you, who make assumptions about stuff you haven't even used. I trust real-world uses from people on a BBS much mroe than a website review.


MIke

 

[KevinF]

Hey everyone, what's up? I realize you can't attach them with a rubber band, I'm not sure what I was thinking. I'm using a Socket A Duron 600 tested at 1000. According to the Via Hardware Monitor, my CPU is running at 31/87 right now while browsing. I understand the inaccuracies of the built in monitor, so I'm gonna wait until I get some real thermal paste to go to 1 gig.

Oh, and he said that white toothpaste would work because of its high calcium content. I figured green toothpaste just had coloring, but Solace generally knows his stuff. He claimed it'll last for a couple months which seems to be a bit much.

Sorry that I've cause a few minor flames...

 

[lhampel]

Easy easy now, no need to get huffy.

From my experience there comes a point where lower temps (however obtained) don?t help. & it won't matter if the CPU is cooled down to freezing temps, there is a limit & once you've reached that limit nothing (except another CPU) will make the system run any faster.
(example: Unlocked P2 300 @ 1950Mhz won't ever happen, not ever.)

Yes, good cooling will help reach the CPU's limit.
(Arctic Silver, APLHA, High CFM Fans, etc.)

Yes, added voltage will help reach the CPU's limit.

So, a good HS & fans & goop are what every self respecting Overclocker (and that's what we are) needs in his/her attempt to find/reach/stabilize the CPUs limit.

 

[paulip88]

About that toothpaste, still not a good idea. I'd be more confortable if you followed that carck head Zoltarc and used no interface material. Reason is that toothpaste is water based, which means that it will dry and shrink quickly. This is not what you want, since when it does dry , it could make plenty of room for air to move in the insulate between your chip and the heatsink.

 

[EdipisReks]

you're the jack@ss, mike, not longhorn, and none of your overclocks are terribly impressive, especially since you are most likely just running most of those processors at the speed that the cores naturally run at because of chip yields. Arctic silver isn't that spectacular. i've been there when a duron 700 was run at 1.15 ghz with the cheap radio shack compound and a nice taisol, and when my buddy tried the arctic silver, there was absolutely no difference in the overclock, and the temp difference was like 1 degree c. he knows how to apply thremal grease, so it wasn't misapplied. before you start spouting like a spoiled brat, try running a processor at something impressive, something that requires a lot of voltage, and skill.

--jacob

 

[Mikewarrior2]

EdipisReks,

No, you obviously don't know anything about Socket A temp measurement. It has been stated many times on this forum that temps from Socket A mb's are not accurate to any degree. The fact that you assume NO temp drop with Arctic Silver versus Radio Shack grease says tons about what you truly know.

If, On a socket A mb that temps between a PAL6035 and C-Orb(ala Anandtech Heatsink review), a core difference of at least 12C is a reported 2C.

That is a compression factor in the temp readings of at least
6C. So, you're measured 1C temp difference could be anything from a 4 or 6 or Even Higher POtential Core Temp drop.

SO before you open your stupid @ss mouth and start spouting about something you don't know $hit about, why don't you read and learn from the forum first.


Mike