Removing the A64 IHS

[tomati]

bios 5/14 HERE

hope it will help you

[Zeus]

Has anyone any ideas on how to glue the spreader back on?
Just in case the CPU has got to be RMA'd?

Reading this thread it seems an easy mod for a few free MHz.

[jonspd]

I would assume it isnt possible unless you can get some of the same stuff they used @ amd to mount it with or something that look's the same.

[=[PULSAR]=]

Yea...definitely being able to put the IHS back on to RMA it would be nice. I have a current max clock of 2.67ghz on air benchmark stable and will be taking off my IHS also get that new nanotherm PCM+ paste so hopefully better the temps even a bit more from this thread I heard a 60-80mhz increase in OC oh I would love to hit 2.75ghz on air. I will post back with results.

[Zeus]

Argh!
Just removed the spreader, cleaned the core and removed all that black rubber-alike glue and reinstalled with AS5, my RBX on top and the nuts tighted up as firm as i dared. (almost as tight as with heatspreader. )

Guess what?

Just a lousy 2C gain idle and next to nothing under load!
Reseated the waterblock, no difference...

Maybe there is a small gain with more volts to the CPU, only tested with default voltage.

[jonspd]

I'am not sure if you can do this but Y not take some pad's from a old xp cpu put then on the A64 and leave the IHS off just be carefull when mounting the waterblock and dont overtighten it.

[Zeus]

Well, the core is quit big, i think twice the size of a Barton core and looks pretty much the same material.

I've never cracked a Barton, not even with these very powerfull retentionclips of the SLK800.

This waterblock slides on and by tightening down the nuts about the same amount it should be ok, not too worried about cracking the core.

For some reason i don't think the contact is 100% so i will glue the heatspreader back on again as soon as i found the right type of glue.

To make matters worse, my max OC got down about 30MHz!!

[pkrew]

You probably didn't make good contact with the core. Esp if you lost overclock. It can be tricky with the IHS removed and its very possible to get only a part of the core contacted. As far as glueing back on, I've heard that rubber cement works well, but haven't tested it.

[saaya]

and dont forget, some people seem to not get a big improvement at all, some cpus seem to make good contact with the ihs and unless you are pumping some high volts through it you wont notice a big difference with or without the ihs

removing the ihs is only worth it if you going for 1.85v+ or have very high temps/a bad oc with the ihs.

[=[PULSAR]=]

WARNING: Do not use the XP-120 with shim removed there isnt enought tension and since its using a P4 bracket there is no way to tighten it. Turned PC on and cpu was at 90c. Also accidently cut off part of those mini IC's near the core and cpu is dead now, trying to RMA it.

[pkrew]

Originally posted by saaya
and dont forget, some people seem to not get a big improvement at all, some cpus seem to make good contact with the ihs and unless you are pumping some high volts through it you wont notice a big difference with or without the ihs

removing the ihs is only worth it if you going for 1.85v+ or have very high temps/a bad oc with the ihs.

This would probably depend on cooling. People on Phase change are most likely going to see the better results, but regardless of cooling you should never see a drop in oc unless you have a bad mount.

I'm not just talking about making contact or getting in tightly snuged down. You can still have a bad mount regardless. Its very easy with the IHS removed to have more pressure on one area over another on the core. With my LS the difference between a bad mount and a good one is barely stable at 2.7G to running 3d near 3G. With a bad mount the highest area of pressure is always off to one side.

This is also the best way to not crush the core. Its going to be much more difficult to crush a core when the pressure is evenly distributed then off to one side.

As far as benifit, I think its a benefit even if you have a good overclock. I started with 2.93 on my FX and took the IHS off an ended up with 2.98.

[Zeus]

Glued the spreader back on, removing it did bugger all for me.

Now everything seems to be back to normal.
I used AS5 between the core and spreader but doesn't make any difference over the stuff AMD used even though it was quit hard and dried out.

Perhaps the AS5 needs some time to break in?

[CrazyXP1700]

i would try and lap both sides of the IHS... down to a mirror finish... see how that works out

and AS5 has got to burn in
i've done 5 IHS's soo far... mostly on P4's... been getting good results from all of them...

next rig may have a A-64 FX or an Intel EE... i dont know yet
but the cooling will be a lapped to copper and mirror finish XP-120 with a 120mm delta and a Vapochill LS sometime afterwards

[taco]

well.. i took mine off yesterday, and i must say, it was definatly worth the time, temperatures dropped DRAMATICALLY. may i note i have an xp120 w/ 90cfm pana.. you'd think it'd be "dangerous" to mount this heatsink on such a fragile core, plz.. mounted NO problem at all!! this is a 3000+ by the way. heres the results

heatspreader on w/ stock : 32/41c
then heatspreader ON w/ XP120 : 28/35c
THEN heatspreader OFF w/ XP120 : 21/27c !!! woo

[taco]

thx spork

[=[PULSAR]=]

Taco how did you get the XP-120 to mount properly with the shim removed I had really bad contact.

[DrJay]

I removed the ihs from a 3000+ last night. Cpu is a C0 stepping, 512k clawhammer, week 0343. It is indeed a failed clawhammer core as it uses the full sized 1mb die. Anyway, the ihs came off quickly and with no problems. I've only had a couple hours to test it out but the cpu is already running 70 mhz faster than its previous best overclock (2.6ghz vs 2.53) at approx. 6c cooler load temps. The before and after comparison may not be valid though as I used different heatsinks. The before was with a zalman7000-cu modified, using a 69cfm, 92mm fan and the after was with an xp-90 using the same fan. Gonna try this with a CG stepping core that overclocks a bit better....I wonder how the true NC core (being smaller)will transfer heat without the ihs.

Jay

[krojczy]

Hi, i found at this forum 2 links to www with description about signs at A64, but i can't remember these adress, do you know where can i find it?
I think that it was writen by Karnivore or other advanced user.

[RacerX]

Well finally got around to remove my IHS on my 3700+. Before with it on I maxed out around 2630mhz at 1.725v with my H2O setup. With it off I see a nice 70mhz jump and 4c decrease in temp. It took a few times reseeding my waterblock just making sure I don't crush the core. Never could run 3dmark at 2.7 no matter what voltage.



Haven't ran prime yet but successfully ran looped 3dmark 10x without a single crash and SuperPI at 32. I'm assuming I'll need to drop it another 20mhz to make it prime stable.

Worth the removeable for me atleast.

[bh2k]

Hey racerX that's a pretty neat little cpu program there. Where did ya pick it up?

[RacerX]

Hey racerX that's a pretty neat little cpu program there. Where did ya pick it up?

Another member posted it here......http://xtremesystems.org/forums/show...728#post553728

scroll down towards the bottom, can't miss it.

Yes a neat little program indeed.

[bldegle2]

"WARNING: Do not use the XP-120 with shim removed there isnt enought tension and since its using a P4 bracket there is no way to tighten it. Turned PC on and cpu was at 90c. Also accidently cut off part of those mini IC's near the core and cpu is dead now, trying to RMA it."

>>>>>>>>>>>>>>>>>>>>>

i noticed something when i removed the sp-120 i am using, the heatsink had transferred some thermal paste to the bracket on one side. so, i got to looking and the heatsink definately hangs right above it, so when you removed the CPU covering, the heatsink lowered but was held askew a little bit because of the heatsink hitting the bracket. easy fixola follows.

so, i solved this and the retention problem with a little peice of sandpaper (80grit), i sanded the bracket between the stands so there was sufficient clearance, and then to help with the tension because of the lower setting, i took the same piece of sandpaper and counted twenty strokes on each bottom leg, then remounted it with no problems. i know this isn't scientific, but it works. you may want to actually sand the feet a little more, depending on your inclinations.

anyway, now you know why your temps skyrocketed to 90*c.

baldy

[taco]

i had no problem at all mounting the heatsink onto it. i actually made a shim with some thin rubber material i had laying around, JUST to be sure it didnt crush, but it actually didnt make contact, and it would shutdown automaticly. so i took those off, and just mounted it "as is" NOT one problem at all, my temps are amazing and i highly recomend the heatsink without the spreader ++ just be careful

ill get some high res pics up soon!

[LBJGH]

hmmm, I'm currently getting a 400mhz o/c out of my old (sept. '03) CO stepping 3100+ (1800-> 2200mhz) at 1.70v.

I used Tmod's biosmod to enable the Vcore setting of 1.80v on my Epox mobo but the extra voltage doesn't help at all o/c further.

Can I expect more out of this old chip or should I hold out for a new one. I don't feel like killing (and having to replace it) right now.

[bldegle2]

and reinstalled the xp-120.

got a 4/5*c drop in temps. i reduced the vre another .025, still running the 9x284. gonna keep reducing the vre until i find out what the lowest limit it will run this setting, then work from there.

hey, in the world of clocking, any temp drop is good.

one may note, i sanded the base bracket the legs again reducing the height even more than before, and then i sanded the base of the mounting brackets where is makes contact with the MB, ie screw holes, one must do this to get a flat fit, after i sanded the legs, the bracket rocked. also, by doing all this, one does slightly increase the heatsink pressure on the CPU, the heatspreader is quite thick and much tension is lost when running nude.

the heatspreader came off in about one minute, the whole operation took less than 20 minutes.

the cpu core is rather large, so 'landing' buttons are not needed.

this was very easy to do.

the only note of caution is be real carefull not to penetrate too far with the razorblade, there are little units inside around the core that can suffer damage is one is too aggressive.

clock on.

baldy