• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

The tale of two tuallies

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

DonP

Registered
Joined
May 18, 2002
Location
Cincinnati, OH
I have two Retail 1.0A Celleries. All of these test are with a BE6 ver 1.2 upgraded to 2.0 with BIOS hacks that also allow the Tuallies to be id'ed. An Alpha PEP66 w/ Delta 38 provide cooling. 256MB of Micron memory know to hit 156 CAS2 on a KK266, and a Retail 8500 64MB for video.

One is a Philly chip, that at 1450 and 1.70v hits 41c-42c under prime 95. It's dated week 36 of 2001. It cannot hit 1500 without 1.80v, and it's really pushing it at that.

The new one is a Malay chip dated week 11 of '02. It hits around 46c at the same speed as the above chip, yet it can hit 1490 at 1.70v. It gets farther easier than the Philly chip, but it just runs hotter.

So this is what I figured. The heat shield on the Philly chip was put on better (runs cooler), but the silicon is worse because it can't hit as high with lower temps.

The Malay chip has better quality silicon, but because the IHS was put on badly at Intel, it runs hot.

So I decided to pull the IHS off of the Malay chip. This went successful, but I can't seem to get the chip to cool properly. Putting an Alpha PEP66 directly on the chip gets me to the 60's. using the IHS as a shim to level things out doesn't work either (high 50's).

So now that I've already done it, how do I cool this tiny cpu?

Some interesting tid-bits:

Running at 1500, I'm getting 8300 3dmark2001SE. My XP in the KK266 running at 153*10.5 = 1600 only gets 8000. I am really interested in keeping the AGP at 100MHz. Even at 2x, the true throughput is what counts. I'd even go for modding the BIOS on my KK266 to prevent the AGP 1/2 divider from kicking in at 133.

Running the Tually at 66*10 (667MHz) at 1.4v, using my finger at the heat sink, it gets as hot as a soldering iron within a few seconds. These chips are really hard to keep cool because of the small form factor.

How about this: If I mill a hole in the copper of the PEP66 so that the HS sits flat on top of the CPU, and the core sits flush inside the recess?

Does someone know the heat output in watts of a 1500 tuallie?

HELP! before I go crazy. I've always had good sucess overclocking (550E running at 853, 933 running at 1120, XP 1600+ running solid at 1600). I see a lot of prople hitting 1600 with these tuallie chips. I've had 3 (the other was a 1.1A for a friend with good markings), but none have been able to break 1500 without extreme measures. I'd really like to hit 1550 with a 1.0A on my BE6-II. Is this asking too much? Am I at the end of the line with these CPU's?

Thanks for the help.

Don
 
In my opinion, it's pretty hard to get the heatsink mounted properly on the Tualatins without the IHS. I tried this late last year (after killing 2 chips on my first attempts) and was not able to get any better results. I then put some Artic Silver on the core and glued/epoxied the IHS back on. With the IHS back on, it performed about the same as it did originally.
 
That sounds like a good idea (AS2 epoxy). With the black stuff scraped off, the core is a little taller than the IHS. The clamp of the HS presses down on the IHS and bows the organic sustraight inward. with this much pressure on the core, I would think I was getting a good contact.

Too bad there isn't something like a putty to fill in the air gap under the IHS and transfer the heat away from the core better.
 
I dont see why people feel the need to take the IHS off, it makes mounting a breeze, it keep the CPU cool, and Intel chips dont give off alot of heat anyway, so why go through all the trouble?
 
If you read my first post, you would have read that I have a good Malay chip (overclocks higher), yet it runs hotter that a philly chip by 4c to 5c at the same speed (1450) and voltage (1.7v).

I concluded that Intel must have put the IHS on wrong on the Maylay chip, and that if could find a better solution I might have a really good overclocker.

The IHS came off without a problem, so now I'm just trying to find a better cooling solution. One solution is packing the IHS in Artic Silver Epoxy and clamping down really hard on the IHS.

A petlier might be an idea too, using the IHS as the cold plate.
 
DonP said:
If you read my first post, you would have read that I have a good Malay chip (overclocks higher), yet it runs hotter that a philly chip by 4c to 5c at the same speed (1450) and voltage (1.7v).

I concluded that Intel must have put the IHS on wrong on the Maylay chip, and that if could find a better solution I might have a really good overclocker.

The IHS came off without a problem, so now I'm just trying to find a better cooling solution. One solution is packing the IHS in Artic Silver Epoxy and clamping down really hard on the IHS.

A petlier might be an idea too, using the IHS as the cold plate.

Try making a shim....or buying one....
 
I dont see why people feel the need to take the IHS off
Because most of the members who have done it have seen temperature drops of 2-3C under load, which makes it worth the effort. You just have to figure out how the mount the HS on there, which is the problem/challenge I see here. Have you lapped your heatsink, DonP? It may help. Also, take a look at this article. It shows how to move the fan of the PEP66 to the top of the HS for a better fit on the mobo. This would also affect the weight distribution of the HS on the core. Also, you didn't mention what kind of paste (if any) you used between the HS and the core? And a shim, like funnyperson1 recommended, would help out a lot.
 
If I grind down enough of the IHS, won't it serve as a shim?

In fact, couldn't one look at the IHS as a shim with a copper plate over the top of it?

The thing that kills is this tiny .13 core is so hard to cool. .18 P3's and T-Birds didn't have this problem. We put the HSF right against the core.

And for those who say these don't run hot, your wrong! It might not put out 100 watts, but 15 watts in a soldering iron is a lot harder to cool than a 60 watt lightbulb. These cores don't put out as many watts but those watts are so compact they get real hot.
 
Takiwa,

I saw that article before, but I'm using a slot 1 converter so I don't have clearance problems.

Hurray! I found the problem. The bad clamping design of the Alpha PEP and PAL series seems to be the problem.

I took an all copper HSF I got from SVC (about the same rating or a little better than the PEP) and got "normal results"

Mounted dry, at 1500 and 1.65v I was getting 46c idle, 51c under prime95 torture.

At 1520, maybe .5c higher. 1530 has problems making it into Windows.

I'm going to order a good quality all copper HS, making sure it has a good quality mounting system. Probably a Thermalright SK6, AX7 or the new one (forgot it's name).

My problems are that I'm on a slot, with a large HSF on the BX chipset, and memory clearance problems.

But I'm very glad I can get this thing to cool. I think with the PEP66 it wasn't seating properly because of the BX cooling.

Thank god this wasn't a t-bird. Yes, I lost a 1.333 birdy to bad HS mounting (interference with the socket).
 
Last edited:
You guys aren't going to believe this, but throwing a thin layer of AS2 dropped the temps down 10c!

I really need to do some lapping, but now I'm hitting 35c idle and 39c under prime95. 1500 at 1.65v

My horse lost a shoe (while trying to hit really high FSB speed hard drive corruption) but I am very happy with this.

At 1500 I'm getting 8300 3dmark2001 default settings with a retail 8500.

To those who ask about why remove the IHS, remember that I started with 46c idle and 51c prime95 at 1450 and 1.7v. This was with the PEP66, but both the PEP66 and the SVC heat sinks are using Delta38 fans. I don't think changing the HS made the 10c difference.

Once I lap the SVC to a mirror finish, I think I'll have a really good setup. I just need a shim to make sure it doesn't get bumped out of alignment.
 
I made a thread dedicated to making a shim for this chip. SImply you take 2" to 2.5" wide packaging tape and cut out squares and put about ten of them stacked on top of each other on your tualatin. I have made about 5 of these and they work very well. I used my IHS for the longest time on my cel-t but as of late temps were getting bad again so I took it off and used a shim again. The IHS was only stuck to the HS with ASII when it was used. I never keep the IHS on the cel-t ever. I have removed about 5 of them also. USing a shim is the only way to insure a proper fit with a large HS especially one like a pep.
 
'ol man,

Thanks for the support. I think I'll try the tape method to build up the support like you suggest.

Is the AK4 to AN26 bridge still the method of choice? I've seen people saying to connect AK4 to Vtt or Vcc. I've been doing this testing without the bridge, and figure that must be why I am having slight stability problems.

Let me ask one question from everyone with celly-T experience: Does it seem like the 1.2's are made with better silicon?

I have a real hard time getting higher than 1500 with two 1.0A's and one 1.1A. I see the 1.2 chips always hitting higher than this (maybe not always 1.6 but definitely higher than 1500).

I guess I'll just have to buy one to find out.
 
The 1.2 cele has a hhigher multiplier (12) so if you hit 1500 with a 1.2 then you are only at a FSB of 125 where 1500 with a 1.0 = FSB of 150. The higher FSB of the 1.0A will give better overall performance. Now if you get a 1.2A that will hit 1600+ then you start to reap the benifit of the 1.2 cele.
 
Hi,

In my experience I tend to think that *modern* silicon is equally good for 1.2, 1.1A or 1.0A, that is, all of those will *probably* have the same overclocking potential, in terms of total absolute Mhz... of course, FSB will be lower for big multipliers...

I know that the database here shows 1.2s overclocking better than 1.1s or 1.0s, but I think this is mainly because there are many entries of *initial silicon* (jul-oct/2001) where probably there was a real difference, and also because with lower multipliers you need to have 'better' overall systems, to reach high FSBs (i.e memory,...) thus causing many not to reach the real chip potential... i.e, how many people has really tried to overclock a 2002 1.0A to 160+ Mhz FSB on a system *known* to work at such high FSBs ??? (would like to be proven wrong, but probably almost noone)

...Of course I may be biased by a good 1.0A experience, but also with some common sense, it does not look like intel may be having trouble to make it do binning at such low frequencies...

...just an opinion
Regards
FTC
 
FTC,

What type of cooling are you using? That is a very good chip hitting that speed at the default voltage. Have you tried pushing it higher?

I have an Asus CUSL2 that will run a P3 933 at 1120 (CAS=3, but everything else maxed).

Using my better chip (the week 11 '02 Malay 1.0A) without the IHS, and a solid copper SVC heatsink against the core with a delta 38, it's pulling teeth trying to stay stable at 1500 and 1.65 to 1.7v.

I just don't get it. Maybe the 933 I have is a real good FSB processor.

Is it tempting fate to try and hit 1600 with a 1.2?
 
DonP,

I am using a GlobalWin CBK68-II (very good aluminium HS, but worst than almost ANY copper based one) with fan vmodded for 7volts, so it is better than retail, but not a top notch cooling solution.
I know I have been very lucky with this chip. (But it is my third one, I tested another 1.0A and also have a 1.2, so in part I have forced this luck).
I DO think that trying to get anything better than 1540-1560 Mhz with tualatin core based chips is a bit tempting...
And for last,I have NOT tried to push the chip higher. This IBM aptiva has a ICS 9250-19 clock generator, which can go 150FSB MAX.

Regards
FTC
 
Back