Celeron Tualatin 1GHz@1,7GHz treatment?

[Celemine1Gig]

Hi guys and gals,

well as you probably remember I was lucky enough to get a new Celeron 1GHz CPU, that was produced in the also new TB1-stepping and got it upto 1700 MHz(170 MHz FSB).

Well, as you can see in my sig, the chip now refuses to operate stable even at 1500 MHz!!!:eek:

But I think I know the reason why it behaves so strange;)

Well, I removed the IHS, because I thought it would be a god thing and would give me about 2-3 °C less on-die temp using my Thermalright SLK-800 Heatsink with Delta EHE 80CFM fan.
As it seems, in reality this wasn't such a good idea. The IHS seems to be there for a reason --> simply to spread the enormous heat away from the small die into copper of the IHS!

Now, I think at clock speeds of 1500 MHz and more, this small die of the TB1-stepping chip seems to get to hot woithout the IHS, as there's too much heat in a too small area --> crash!

Now, my thought is to reattach the IHS, but:

1) I plan to lap its downside, to make it fit perfectly on the CPU core
2) I plan to lap its upside to make the best possible conntact with my already lapped Thermalright SLK-800 heatsink
3) I plan to attach the IHS using Arctic Silver Epoxy and to fill the whole IHS's inside with it, so that there won't be any air between the CPU's ceramic and the IHS's metal, but thermal-conductive Arctic Silver Adhesive!

Do you think that this could work out fine???

And did you understand what I'm going to do?(I ask because I know that unfortunately my English isn't perfect)

[Rio71]

hi ingo. :)
is the tb1 die smaller as the ta1 ?
i think, this reattach-idea is not so good.
make sure the slk-800 sitting perfect plan on the die.

[Celemine1Gig]

@Rio,

I posted the same question in the tualatin.de-forums! Would be easier for us to discuss in German, wouldn't it?

Regards

Ingo

@all others:

Any suggestions are highly appreciated!:)

[Rio71]

yes, ich habe fertig. :D
link (http://www.tualatin.de/forum/index.php?mode=viewthread&forum_id=1&thread=266&z=last&KONTENTSID=722a2230fe934bc817359e15c51c5b00)

[toastedzergling]

Water cooling? It's about time:D

[jdmcnudgent]

i personally would not remove the IHS without watercooling. there will be a gap, i believe between the core and the HS. with watercooling, you can get the block settled nicely on the core.

[doer]

Now you must have a HS with copperbase. Do you?

[Celemine1Gig]

@jdmcnudgent:

Well, there is no gap, as the Tualatin without IHS is only negliably smaller in height, than the Coppermine!

AND, I could hardly attach the Thermalright SLK-800 to the socket, when the IHS was still ontop of the CPU-die. So removing the IHS seemed to be a great idea!

Another idea would be, NOT to reattach the IHS, but to attach a copper spacer to the CPU and to lap the CPU-die and the copper spacer alltogether to nearly perfect flatness! How about that???:confused:

@doer:

Yes I think it would be better to have at least a heatsink with a copper base or as you said a full copper heatsink.
But that is definately not my problem, because I use the best Socket aircooler around at the moment, the Thermalright SLK-800 paired with a Delta 80x80x38mm EHE 80CFM fan.

[@crilicM@n]

Hy Celemine1Gig! Shurely it´s not your cooler fault... It´s known that the IHS helps stability in other forms... People told about tualatins that would become instable whithout the IHS even when you touch the heatsinkfan, making it freeze with the touch !!!! :O Reataching it made it work normally again.... The temperatura gain of remooving the IHS is being very controverse.. Some say a positive feedback, others a negative. I´d rather not to do it, just lap the heatsink and the IHS using that very thin layer of our good AS... Use the thermal epoxy, ut do not make the IHS sit on the slug with a too much big amount of epoxy as you will increase the distance heat must pass to reach the IHS and the Heatsink, thus, making it forever hotter :( The surrounded gaps shall give a marginall efect or nothing compared with the already small bennefit you´ll have with the lapping.... (OOOOh those old 1st celerons... that was a must lapp!!! he he) My 2 cent´s is leave the IHS alone! You probably may yeld another 1c trying to improove your case cooling.... Time for the 5th or 6th silent case fan....

Cheers

[Bonka]

If you use a towercase, that would explain why you'd be getting higher temps. The IHS ALSO acts as a pressure plate, for MAXIMUM contact pressure from chip to heatsink base. Now that the IHS is gone, the heatsink base isn't exactly flat with the core now, causing your higher temps. Have you tried tilting your tower horizontally? If your temperatures go down when you do that, you know that you're core is not getting complete heatsink surface.

Just a thought though...

[looktall]

i don't recommend filling the empty space under the IHS with thermal epoxy. or any thermal compund for that matter.

those products are designed to be used in as thin a layer as possible. when used in large quantities, it's ability to transfer heat is greatly reduced, sometimes it works worse than air.

if you're going to put the IHS back on use a very thin layer of epoxy between the core and the IHS only.
lapping the top of the IHS should help a bit with temps.


i also agree with comments that the heatsink is no longer making good contact after the removal of the IHS. is it possible to adjust the clip for a tighter contact?

[Celemine1Gig]

Well the odd thing is that the heatsink is indeed making good contact to the CPU die. The inprint in the Arctic Silver III thermal paste was almost perfect when I removed my SLK-800 because of the problems I was getting! The core touched the heatsink nearly 100%, thus the Arctic Silver was nearly gone where the CPU core contacted the heatsink(it was a nearly invisible thin layer and the rest got squeezed out on the sides).:confused:

[Darkseid]

wich mobo let you go all the way to 170 fsb???
because i know my tualeron can go much higher!!

[looktall]

there's a gigabyte mobo, i forget which one, that goes up to 200fsb. :D

[Darkseid]

it has /5 divider????

[looktall]

no idea, but you'd imagine it would going that high.

i'll have a look and see if i can find the model number.

[looktall]

it wasn't gigabyte, it was aopen.

i can't find the board i was thinking of, but i did find this one instead.
http://english.aopen.com.tw/products/mb/ax3spro-u.htm

if you look under the clockgen heading, you'll see max overclocking = 248mhz. :eek:
i was going to download the manual and check it out a bit more, but it's 5.5mb and i'm on dial up so f**k that. :p


EDIT: there's two other board i found that also use DDR. lol
http://english.aopen.com.tw/products/mb/DDRSK370.htm

[muddocktor]

If I remember right, Celemine1G is using a Gigabyte GA-6OXET board like I have and it indeed does have higher multipliers than 1/4. If you can find one, they are great boards to play with Cel-T and P3 Tullies on, along with the ST6.

Celemine1Gig, I don't see where you should have any problem reattaching the IHS back on the proc, as long as you didn't warp it when you took it off. Instead of filling it with AS adhesive, I would use AS3 and attach it with a little superglue around the edges of the IHS. I would think that the thermal conductivity of straight AS3 would be higher than the adhesive and would be easier to work with when reattaching the IHS with superglue. After the IHS sets up good, then go ahead and lap that sucker smooth.

[Edward2]

The gigabyte board that Celemine has (the GA-6OXET) will go up to at least 200 MHz FSB. Yes, it does have 1/5 and 1/6 dividers, but I have heard that some people have had problems getting them (the dividers) to work. I don't know the details however. BTW, that particular board is no longer in production, so you would have to find a used one somewhere.

[asw7576]

I plan to attach the IHS using Arctic Silver Epoxy and to fill the whole IHS's inside with it, so that there won't be any air between the CPU's ceramic and the IHS's metal, but thermal-conductive Arctic Silver Adhesive!

Yeah.... fill in the whole IHS to make it solid!!! yeahh.... I know it is Over Dosage idea, but I love this crazy idea!!! My added suggestion, You should give the center of area of CPU die with a lot of pure AS3, and for its surrounding area with AS3 Epoxy mixed little bit with AS3. And WHAM!!!! slap it together for overnight curing process. Definetely I support this idea!!!

Tell me the results on MHZ and Temperatures.

[funnyperson1]

i recommend reattaching the IHS, i killed my Celly 1.0A by removing it, it ran for a couple months after with me having to randomly reseat the heatsink to make it work and then it just died one day :(

[JCLW]

Remember that different materials expand at different rates when heated. When you fill that IHS with epoxy you fix everything in place and so when you load your system up there will be no place for things to expand/move.

OK. That was really badly writen. But you know what I mean (I hope).

- JW

[Darkseid]

well thanks, i see if i can find any of those mobos and celemine1g sorry for stealing your thread ;)

[6502kid]

Darkseid,

Since you know your chip goes over 133 at default core,
you might also be able to use an ASUS TUSL2-C board
with that chip to hit higher FSB's.

I havent tried a celly in mine yet, (Waiting to find a 1.2 1.3 tb1
stepping chip to play with), but my P3-s cranked up to 151+
easy when I tested it out.

Try to find the XET if you can. It's 1/4 divisor (PLL16) works
great with a Tualatin Celly in there. (Both systems in my
sig are running in XET-C boards.....)

[Celemine1Gig]

Well you don't even have to use PLL16 to get the 1/4 divider to work! Just select 133 MHz FSB as default via the onboard dipswitch and the 1/4 PCI divider and 133 MHz FSb are default!:D

[6502kid]

Celemine,

I got the XET-C versions of the board, which dont have all
the dip switches or dual bios, but other than that I think
they are the same....

[Celemine1Gig]

Hmm, on one hand, I think you would just have to solder the dipswitch in and it would work!
On the other hand, you don't really need it!

Regards

Ingo;)

BTW, I can run my board stable at 180 MHz FSB! What's your max FSB on your GA-6OXET-C???

[6502kid]

158 is the max I have had it running stable at for any
period of time. Any more than that and either my memory
or HDD flakes out. I tried playing with the PLL16 for
speeds higher than that, but it seems to work intermittantly
with the P3-s chip in there. It works great as 1/4 for the
cellys however....?

I had it up to 172 once and was able to run Win98 and get
sandra benches, but Prime95 would always find errors
at that speed.

[ol' man]

Yeah I have not even attempted to remove the IHS from my 1.4 PIII-s. I was at first one of the biggest advocates of removing it and have done so about 7 or so times on different CPU's but in the end it is not worth it. Too bad you did it too that nice celly. First off if your HS was really touching the core as good as you say you would notice slightly better cooling. Bad thing is it is very hard to get the HS set on that teeny core. I suggest a shim but better yet I suggest gluing the IHS back on. You hit 1.7GHz with it on. SHould have left it alone. etc....

Anyway, I have made many different shims for my IHSless cel-t's- PIII-t's atc...... I found packaging tape is the best in layers with the core cut out. I still have a PIII setup running like this for my mom but I regret now that I removed the IHS.

It really has alot more purpose than simply spreading the heat, alot more.

Like I said a properly aligned HS will always perform equal or slightly better than the IHS bound chip. After buying my last 1.4 PIII-s I decided against removing the IHS anymore.

[belinassu]

First i removed the cover
then i replaced it using artic silver...

But the best performance came when i removed teh cover again,
the lapped the thin core using the wishers method (seen in Martinus page)
and used a coppr based HSF (hand made, see it in my web page)

the lapped wishers are as wide as the cpu core, so the become
naturally spacers ;)

I think the best thing (and i'm plannig to do it)
is to replace the original copper cover with a Silver one
(a 10% of dissipation gain and a sure unbreakable core ;) )

[Flu!d]

Well you don't even have to use PLL16 to get the 1/4 divider to work! Just select 133 MHz FSB as default via the onboard dipswitch and the 1/4 PCI divider and 133 MHz FSb are default!

Strange....

I tried that when I first got my Tualatin Celeron using the old bios (which didn't allow a 1/4 pci divider on a 100 mhz FSB chip) and it definatly didn't work on my XET....I had to complain to Gigabyte and wait for a new, modified bios to be released.

Mabye it works now with the newer bioses as I haven't tried it since....