IHS removal and reattachment on Opteron 939 / Athlon 64

lemonlimelemonlime Canada Member
edited December 2006 in Hardware
Just thought I'd say this right off the bat: Do this at your own risk. I take no responsibility for the brave souls who follow my example :sawed:

Well, after reading success story after success story, I finally decided to void the warranty on my 2-week old chip, and pop the IHS. This chip always ran VERY hot, even considering it's high OC and overvolting. At 1.55V it exceeded 60C on aircooling. Under the vapo, it struggles to maintain 5 degrees at full load, which seems too high considering an evaporator temp of -48. I know these sensors are likely not terribly accurate at below zero, but there is no doubt that this bloody opteron runs too hot. Anyhow, Here is the verdict:

Vapochill LS, 3.2GHz, 1.58V full load (FAH Gromacs) with IHS = +3C
Vapochill LS, 3.2GHz, 1.58V full load (FAH Gromacs) without IHS = -9C

My Evaporator temperature is now a few degrees higher, which is a good thing, because it is clearly pumping more of the heat away from this chip now.

I'm still not convinced that I have good contact between the evaporator and the core, but there was an immediate difference! This is with higher ambient temperatures too. 12 degree delta is more than I expected, but as I suspected, the IHS contact must have been poor on this chip. The thermal compound that AMD uses is flaky and hard. The IHS is much thicker than it looks. It is 90% of the weight of the chip. Without it, it feels very flimsy :)

I had to modify the evaporator by placing some home-made plastic spacers behind the copper head to make it protrude an additional 3mm from the clamshell. The chip is significantly shorter without the thick IHS on it.

Just some history: I butchered an old 2800+ newcastle doing this with the wrong type of blade, and I did not want to make the same mistake again. I saw someone on XS mention that the 'Mach 3' or other disposable razor blades work wonders. I wish I remember who it was so that I can give credit, but boy was he right. The blade is SUPER thin, and makes very quick work of the IHS. It took me longer to rip the blade out of the shaver than to cut the IHS epoxy. Anyhow, be very careful.. they are razor sharp (no pun intended) :D. I wrapped the bottom 3/4 of the blade in thick cardboard to protect my hands.

Without further ado, here are some pics:

attachment.php?attachmentid=18529&stc=1&d=1143491673
attachment.php?attachmentid=18530&stc=1&d=1143491673
«134

Comments

  • profdlpprofdlp The Holy City Of Westlake, Ohio
    edited December 2005
    That's quite a job, lemonlime, especially on a two week old CPU. You've either got a lot more guts than I do or you're crazier. :D
  • GrayFoxGrayFox /dev/urandom Member
    edited December 2005
    :) Very nice.
  • lemonlimelemonlime Canada Member
    edited December 2005
    profdlp wrote:
    That's quite a job, lemonlime, especially on a two week old CPU. You've either got a lot more guts than I do or you're crazier. :D

    Crazy for sure.. overclocking is a strange sickness.. Almost as obsessive as folding.. but not quite :D

    It is degenerative --> :cool::):(:eek::hair::eek3::crazy::buck:
  • profdlpprofdlp The Holy City Of Westlake, Ohio
    edited December 2005
    lemonlime wrote:
    ...It is degenerative --> :cool::):(:eek::hair::eek3::crazy::buck:
    ;D;D;D
  • sfleurietsfleuriet Texas New
    edited December 2005
    Nice work - I know I would've cut myself multiple times, and then ruined the cpu somehow.
  • csimoncsimon Acadiana Icrontian
    edited December 2005
    why didn't you just use the exacto?
  • lemonlimelemonlime Canada Member
    edited December 2005
    csimon wrote:
    why didn't you just use the exacto?

    The blade needs to be VERY thin. Thats how I hacked up that old 2800+. Those razor blades seem to be the only thing I've seen thin enough to get under the IHS. The only thing I used the exacto for was to pry the IHS off once I finished cutting :)
  • csimoncsimon Acadiana Icrontian
    edited December 2005
    lemonlime wrote:
    The blade needs to be VERY thin. Thats how I hacked up that old 2800+. Those razor blades seem to be the only thing I've seen thin enough to get under the IHS. The only thing I used the exacto for was to pry the IHS off once I finished cutting :)

    Thanks ...I certainly would have used an exacto. I almost did it a few times right out of the box but I talked myself out of it and now I'm glad I did.

    Did you lap the IHS when you were using it? I did mine but I'm not so sure how thick the thing is and I don't want to sand it to the core.
  • lemonlimelemonlime Canada Member
    edited December 2005
    csimon wrote:
    Thanks ...I certainly would have used an exacto. I almost did it a few times right out of the box but I talked myself out of it and now I'm glad I did.

    Did you lap the IHS when you were using it? I did mine but I'm not so sure how thick the thing is and I don't want to sand it to the core.

    The IHS is much thicker than it looks. I'd guess at least 2-3mm of heavy metal. You'd have to sand for a long time before you get to the core :) I've never tried to lap an IHS, but I know people have done it and had better contact. Many of them are not terribly flat, especially around the edges.
  • csimoncsimon Acadiana Icrontian
    edited December 2005
    lemonlime wrote:
    ...especially around the edges.

    Yeah I know ...that doggon thing was so convexed it took me quite a while of lapping just to get it semi flat. Whew ...now I feel like a viagra clown with no date! :-/
  • lemonlimelemonlime Canada Member
    edited December 2005
    I'm starting to kick myself a bit. The vapo is now cooling this chip a little *TOO* well without the IHS :shakehead , as I've started to see the dreaded cold bug when the machine is left at idle for too long. Its odd. If I let it sit for 10-15 minutes with 0% CPU utilization, I get hard freeze-ups, and this odd behavior where the mouse runs very choppily across the screen and the sound is super-slow and low-pitched.

    I'm going to try a stonger dram drive strength, and see if it helps.

    There really is only one cure for the cold bug though :fold::cool:

    When at load, ~10 below zero seems to be fine. It's just when sitting idle at close to -30C that things go wonkers.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited January 2006
    Crazy for sure.. overclocking is a strange sickness.. Almost as obsessive as folding.. but not quite
    It's symbiotic. It's a vicious cycle. The Folding obsession demands raw, bone crushing CPU power. How do we achieve that? Through overclocking. Overclocking becomes a means to the end of increased Folding@Home points production rather than more megahertz simply for the sake of megahertz. In so many words, each addiction justifies the other! If only all our vices were always that justifiable! :-/
  • edited January 2006
    I just popped the hat off my old Winnie 3000 just for practice and to see what the core looks like in it. It's small, but not quite as small as a Tbred A core is. I just scanned the winnie, a Tbred A and a Sonoma Pentium M for comparison sake. Note that I did skin the pcb a little on one corner; hope the old proc will still run OK as I eventually plan to put it back to work someday in a cheap board.

    Here's the scanned pic, cut down so it doesn't take forever to load.
  • lemonlimelemonlime Canada Member
    edited January 2006
    muddocktor wrote:
    I just popped the hat off my old Winnie 3000 just for practice and to see what the core looks like in it. It's small, but not quite as small as a Tbred A core is. I just scanned the winnie, a Treb A and a Sonoma Pentium M for comparison sake. Note that I did skin the pcb a little on one corner; hope the old proc will still run OK as I eventually plan to put it back to work someday in a cheap board.

    Here's the scanned pic, cut down so it doesn't take forever to load.

    Nice mudd! glad to see I'm not the only [strike]incredably insane[/strike] brave soul on the forum :D

    What was your tool of choice to do the cutting?
  • edited January 2006
    I took apart a razorblade, just like you did. Those thin blades make a heck of a difference in removing the IHS. Now I'm all revved up to de-hat my daughter's Northwood when I move it over to be watercooled. It's a D1 stepping, so Intel didn't use that low melting point metallic TIM on the IHS like they have used since the M0 stepping Northwoods and the Pressies.
  • GargGarg Purveyor of Lincoln Nightmares Icrontian
    edited January 2006
    So, if processors cool so much better without the IHS, why the heck are IHSes even used in the first place?
  • lemonlimelemonlime Canada Member
    edited January 2006
    Gargoyle wrote:
    So, if processors cool so much better without the IHS, why the heck are IHSes even used in the first place?

    I guess for several reasons:

    1) The A64 SOI cores are apparently much more fragile than the older barton/tbred cores, and genuinely need some protection.
    2) Intel did it first.. simply put, the chips are heavier and look more expensive with spreaders, so it could be an 'image' thing as well IMO.

    All of AMD's mobile chips are spreaderless for lower weight and better cooling. Laptops are banged around quite a bit, so I don't think the core is *TOO* fragile in that regard.

    The stock chips run so cool now-a-days, that I don't think it matters too much. When overclocking and ramping up the vcore, it really makes a difference though.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited January 2006
    I think there are two primary reasons why AMD and Intel distribute their CPUs with heatspreaders: 1) assembly of computers at computer manufacturers is more efficient as fewer precautions need to be taken by the assembly line workers (less risk to damaging CPUs in a quickly moving assembly process), and 2) people like me don't crush the cores when fooling with our computers and blame Intel or AMD for the damage. OK, I've never blamed AMD or Intel for a crushed core, but I did once damage an Athlon Thunderbird when seating a heatsink. MANY of US DID DAMAGE CPUs back in the early Athlon days. Don't deny it! :banghead: ;D
  • edited January 2006
    Yeah Leo, I pretty much think you hit the nail on the head with your reasons. I personally have never damaged a core but I did break the center tab of a socket A zif socket once with a crappy TT hsf. But for assembly line operations, having an IHS is much preferrable, especially since specific heat load has gone up so much with the newer cores (compared to stuff 5 years ago) and both AMD and Intel have raised the clamping pressure required for an approved heatsink solution. It's impossible to crush a core on a proc with an IHS.

    As far as the A64 core being more fragile than a Tbred core, I don't believe it. Otherwise they would have put an IHS on the mobile proc cores too.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited January 2006
    Personally, I was relieved when AMD followed Intel's lead by putting protection on the cores. At least for me, it makes swapping heatsinks and experimenting with cooling/overclocking much less riskier. Case in point -- just last week I destroyed a motherboard by crusing a northbridge heatsink when experimenting with an unconventional heatsink. No, I was not careful enough, but that wouldn't have happened with a modern CPU.
  • edited February 2006
    You could lap one of these CPUs for a week before doing any damage. Most of the weight is in the IHS and you hit copper after getting through the external layer anyway.
  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited February 2006
    You could lap one of these CPUs for a week before doing any damage. Most of the weight is in the IHS and you hit copper after getting through the external layer anyway.
    Well, yes. You are making my point. (Maybe I didn't understand yours.) I just meant that the IHS makes it far less risky when quickly swapping heatsinks and assembling/disassembling computers. I never did ruin a CPU core from crushing or chipping, but I came very close to it several times before the protective heatspreaders became standard.
  • csimoncsimon Acadiana Icrontian
    edited February 2006
    Leonardo wrote:
    Well, yes. You are making my point. (Maybe I didn't understand yours.) I just meant that the IHS makes it far less risky when quickly swapping heatsinks and assembling/disassembling computers. I never did ruin a CPU core from crushing or chipping, but I came very close to it several times before the protective heatspreaders became standard.

    My IHS is lapped but I only see copper on the outter edges about 1/4" in from the edge. I'm considering going even farther with it to see if I can get it all copper exposed. I'm also considering removing the IHS to reapply the TIM and the replacing it back but not sure what to use to affix it with.

    Any suggestions? I've heard silicone but I'd rather be safe than sorry.
  • lemonlimelemonlime Canada Member
    edited February 2006
    csimon wrote:
    My IHS is lapped but I only see copper on the outter edges about 1/4" in from the edge. I'm considering going even farther with it to see if I can get it all copper exposed. I'm also considering removing the IHS to reapply the TIM and the replacing it back but not sure what to use to affix it with.

    Any suggestions? I've heard silicone but I'd rather be safe than sorry.

    If you are going to go to the trouble of removing the IHS, why not just run it without the IHS? If you have a waterblock or screw-down HSF, it should be pretty easy to get good contact :)
  • csimoncsimon Acadiana Icrontian
    edited February 2006
    I've considered that as well. My thinking is that with the IHS in tact I will reduce the risk of crushing the core and also the heat will be spread.
    I don't think the waterblock I use would do as good of a job on a bare core as it does on the IHS. Just something that was noted at procooling when it first came out. I'm using the apogee btw.
    I'll also be mounting a csp-MAG directly over the wb to see if that helps out. I'll probably be using 2 pumps in parallel as well.
  • JimboraeJimborae Newbury, Berks, UK New
    edited February 2006
    Has anybody actually got a better o/c by removing the ihs?

    For me the main reason to remove the ihs was to reduce the temps which was very important as I'm using a resorator to cool an Opty sli rig:eek: Full load temps have now dropped around 10c so it was well worth it.
  • lemonlimelemonlime Canada Member
    edited February 2006
    Jimborae wrote:
    Has anybody actually got a better o/c by removing the ihs?

    For me the main reason to remove the ihs was to reduce the temps which was very important as I'm using a resorator to cool an Opty sli rig:eek: Full load temps have now dropped around 10c so it was well worth it.

    I saw about the same decrease.. about 10-12'C.. It is well worth it on these E4 opterons, thats for sure.. :)

    It didn't do anything for me on phase, but I bet that would have amounted to something on air for sure. I used to hit 60'C with an XP90, I'm sure I could have maintained some additional stability at 10'C lower temps.
  • edited February 2006
    |I'm using a socket 754 3000+ with an Arctic Cooling Freezer Pro 64.

    The biggest difference I saw was a smaller difference in temps between idle and load and a nice drop in temps.

    Not really sure about before and afters, my 'testing' was far from scientific.

    If you're cooling with anything that might lead to some form of condensation I'd suggest leaving the IHS on the chip instead, its going to keep the core safe from what would otherwise be CPU frying conditions.
  • lemonlimelemonlime Canada Member
    edited February 2006
    rapture-IC wrote:
    ..If you're cooling with anything that might lead to some form of condensation I'd suggest leaving the IHS on the chip instead, its going to keep the core safe from what would otherwise be CPU frying conditions.

    I did some research into this before I removed it, because I was planning to run it on the vapochill. It actually makes little difference, and in some respects, it is actually safer to run it without (from a condensation perspective). I was careful to ensure an air-tight seal around the entire CPU socket area, so there should be little to no moisture there. Also, I put dielectric all over the CPU's bare PCB, including the little IC's etc. This area is not possible to reach when covered by the IHS. Even when the IHS is on, there is an opening on one side to allow ventelation, so regarless if it's on or off, the core and CPU IC's can get exposed to moisture..

    The biggest fear I had was getting good contact, and trying not to crush the ICs. Unlike the barton chips, the core is very short, and does not protrude from the CPU much, making it very easy to put pressure on the ICs.
  • lemonlimelemonlime Canada Member
    edited March 2006
    Well, just thought I'd share some observations now that the IHS has been off for a while.

    As I mentioned earlier, I dropped about 12'C at load, which is quite impressive. Unfortunatly, I've got a mild case of the 'cold bug' now as a result.

    I'm very limited in how high I can OC my memory now, and I can not exceed 300MHz base HTT by much. If I push the memory beyond about 230MHz, I get nasty hard-locking when the PC is at idle. At 3.3GHz, it seems to occur no matter what I set the memory to. Seems fine as long as the temperature of ths chip is above -15'C or so. F@H helped to solve that problem, but as soon as it finished a WU and downloaded the next, I got the nasty hard-lock. Thats how quick it occurs. Cranking the vapo heaters to 100% did help, but not by much when it is at 11x303MHz. It seems ok at 3.2GHz (11x291 with a 5/6 divider for 228MHz ram). I used to have no problems at 3.3GHz prior to removing the IHS, but I used to see temps in excess of +8-9'C at load.

    At idle, it drops down to -27'Cish. This is just another example proving that the cold bug stems from the integrated memory controller in the 90nm chips.

    Looks like this thread is going to turn into the 'I put my IHS back on my Opteron' thread :)

    Since I really didn't gain anything by removing the IHS (aside from better temps, and instability), it is time to figure out how to put it back on. I'm going to take advantage of the fact that it is off, and put some dielectric on the PCB before I re-cap it. I'm pretty confident that some 'cold friendly' TIM should help be beneficial under the IHS, and hopefully it'll raise my temps enough to keep the 'bug' at bay. Stay tuned for some pics..
Sign In or Register to comment.