It's hot in San Diego
 

I modded my A64 3700 San Diego for reading with my MAX6658. Holy Bejeebers! Not wishing to retype the post over at OCforums, here's a link to it:

http://www.ocforums.com/showthread.php?t=388922

Got to be an improperly implemented IHS I suspect.

Hoot

Is it possible they have finally positioned the diode in a spot where it actually measures useful reality?

While that's entirely possible, I'm not so sure that's the case here. If the pins used are any indication, it's off on one of the outside corners just like the Winchester. I'm gonna bring some fresh single-edge razor blades home from work this afternoon and we'll see what's up. Pictures will be posted of what I find "under the hood".

Hoot

Gonna "String Theory" it hoot? I can only hope you don't damage your new chip.

No, while that approach sounds plausible, I've got good hand and reflex control. I'll patiently work the perimeter with the razor blade. I was very cautious when I did my Winchester and that went fine. No reason why it should go any different with the SD. I do want to give the AS5 a chance to settle. As viscous as it is and with the large area of an IHS, it comes as no surprise that it would take longer to exude than a bare core. We'll see how patient I feel when I get home from work this afternoon. It's priming at home as I write this.

Hoot

I noticed that the readings from the 6655 seemed off in the high direction when I calibrated Hoot. I had never seen an AthlonXP vary by more than 1C from the water temp measured in calibration, but the Winchester was closer to 7-8C. Check the revision changes and maybe some of the diode properties have changed as part of the San Diego rev?

I can't imagine them moving the diode's location but who knows I guess.

OMG What a difference! :drool:

I have posted images and opinions in the thread over at ocforums.

For those not wishing to view the thread, here's the skinny:

With IHS

11x200 @ 1.41Vcore Idle 34C / Load 51C / Inlet 22C
9x300 @ 1.55Vcore Idle 39C / Load 65-66C / Inlet 22C

Without IHS

11x200 @ 1.41Vcore Idle 31C / Load 35C / Inlet 21C
9x300 @ 1.50Vcore Idle 32C / Load 39C / Inlet 22C

As nerve wracking as it was cutting on a $334.00 investment, I must say it was worth it. Already, the 9x300 was stable at a lower vcore. Who knows how much lower I can go, since I have not had a chance to do much testing this evening.

Heading to bed with a smile on my face and a song in my heart...

Hoot

how did you remove the ihs?

With a single-edged razor blade. The adhesive that holds it on is akin to silicone rubber. Starting with the corners, you ease the blade in a little bit and rock it, feeling your way along. Then you work you way down each side of the ihs. You only use a pushing action, the blade will do the work. Eventually it will push through the glue and you'll feel the resistance slacken. Stop pushing and move a little further, working your way entirely around the perimeter. Once you can insert the blade between the base of the ihs and the substrate all around, you pry lightly to separate the TIM, which is the consistency of bees wax and it pops off. I can't stress enough that you go slow and gentle. Patience will reward you with a clean separation. The greatest challenge is keeping your grip restricted to the edge of the CPU and not touching the pins as they bend very easily. Even being careful, you may bend one or two slightly in the row along the outside. Not to worry, just straighten them with a tweezer or fine needle-nose. Once you're done, insert the CPU gently into the zif socket feeling for resistance. If it wont go, re-check to make sure you got all the pins straight.

While I have done this twice with no problems, you have to go in with your eyes open, realizing that you're taking a risk. In my opinion, it is worth it. My Vcore is now down to 1.45V and prime stable at 9x300. I haven't tried going the other way to see what the reduced temp has gained me, meaning higher frequency. I hope to do some of that this evening.

Good Luck

Hoot

Sweet, i'm in for an IHS removal this weekend!

When I removed the ihs off my 3200 winnie, the load temps dropped 7*c. I believe Amd has some issues with their ihs's being cupped. Mine was not flat at all.

Don't forget this: http://www.overclockers.com/tips1197/

FWIW, before removing the IHS, it took 1.55Vcore to run at 9x300 stable under Prime95 Torture. Now that it is gone and the AS5 has for the most part, settled in, I can run 9x300 at 1.42Vcore. That's measured at the socket, needless to say. I have managed to run either 9.5x310 or 10x300 at 1.65Vcore, again stable under P95 Torture. At that voltage and those speeds, the core temp rises and stabilizes at 42C.

I can't wait for my G5 to ship (ahem!) so I can see how much it gains me.

Hoot

Heh, that was the first thing I checked. I wonder if the various mobo makers use different cam boxes or such what not. I thought for sure after reading about Hoot's problem that mine would need sanding too, but not so.

Nice work, Hoot. Do you guys think there is much or a benefit in removing the IHS from a Prescott? Should I try just for the hell of it?

Be careful on that one. I have seen pics of cores pulled off with the IHS on recent intel chips. At least some are epoxied down.

You should check out these links first:
http://www.xtremesystems.org/forums/...ad.php?t=44854
And to let you know the risks:
http://www.vr-zone.com/?i=308&s=1

IHS - what is 'IHS' ?
 

Oh you must mean that thing I get off as soon as I am sure teh chip works :evilaugh:

Must be why I am running 3700 SD @ 11x247 (BH5) 1.5-2-2-5, etc @ 1.49 / 31-32c load (OCCT - stopped after 26 hrs @11x246 - boring :D ) on a BIX WITH an x800xl on a maze4 gpu. Or teh 9x300 on my winnie 3200 @ same vcore /TCCD and 37-38c load on just a hyper6 w/ 80mm panaflo m1a :cool:

I do the razor thing like everyone, corners first, then the straight sides and it comes off easy (no epoxy so far).

NOTE: the chips are slightly convex. So of course the NEXT step is to flatten the top of the core :evilaugh:

2000grit max sandpaper on a FLAT glass plate - SLOW strokes (did I mention I ground telescope optics as a teenager heheh :D ) or better yet some 2 micron abrasives and less.

Thermalright XP-120/90 - BAD for this as I have found them to nibble off the edge of the core.
However my cascade WB and hyper6 have never given me any wear. I attribute this to the fact that these coolers are mounted w/o sliding over the core, unlike the XP-xxx (and give better clamping pressure, damn XP springs :mad: )

I am getting a TT Big Typhoon (Screw down, not springs :D )tomorrow, and after I lap it will try on my winnie and then on the venice that came with. I am guessing that the mounting method will give it the proper cooling ability that I never found on the Thermalrights. Will report back in a thread over the weekend.

Dunno if it would help, but when I removed the IHS from a few AMD k6-2 cpus it did REALLY help to freeze them first.
Freezing them shoudl make any glue/epoxy holding it on brittle, and make it easyer to come off.

I would be very interested to know if AMD 64's have bridges under the IHS - the k6-2's did, no idea as to their purpose though. If the AMd64's do it might mean there is a way to unlock higher multiplyers, which has not yet been tryed?

No, they just have pull-up resistor packs:

http://www.frontiernet.net/~hill195/sd/corereg.jpg

Hoot