Removing the A64 IHS

[=[PULSAR]=]

bldegle2 did you overclock increase at all? Yes be carefull with those IC's I cut a couple of mine, my friends tell me I'm impatient.

[bldegle2]

"bldegle2 did you overclock increase at all? Yes be carefull with those IC's I cut a couple of mine, my friends tell me I'm impatient."

>>>>>>>>>>>>>>>>>>>>>>>>.

that.

i am off to do an errand right now, will play when i get back. it would be nice to get an even 2600mhz out of this 2800+, only 4/5 more clicks on the clock and i am there.

i am thinking this chip would prolly do 2.9 gig on the prommie, but then i would have to get into heavy mem dividers, max multi is 9, so that would be a 322. i have had her running 7.5x320 FSB on air, but it was not so stable, maybe the extra chillin' from the prommie would make it possible.

baldy

[bldegle2]

i am running 9x285, 286 balked. so far so good, but i am going to run some tests for stability.

i am not convinced this xp-120 as good as a 948U unit, but i have to say the 100cfm delta 120mm is very quiet, comparatively speaking to either a 92mm or 80mm of the HO variety.

upped the vre to 1.7 for the 285. i don't trust the cpu temp of 58*c reported by ITGuardian, this is with the 1.7, @1.6 it reads around 55*c.

it just keeps running and running.......................

i have plenty of case fans and a very clean box.

update, no go @285, back to 284. so it didn't help with the ultimate final overclock, but the cooler temps are always nice.........



back in awhile.

baldy

[=[PULSAR]=]

BTW has anyone tried or know of someone that uses a TEC to directly cool the core no coldplate or IHS?

[=[PULSAR]=]

So I might as well just leave the IHS on or would I see that much improvement switching to copper cold plate?

[Dagalidis]

I'm thinking to remove IHS but ......
I use EXOS watercooling with 300G waterblock.
If IHS removed will waterblock make good contact with core ???

Any of you with EXOS @ IHS removed ???

[mad mikee]

saw many post and my own experience that you need to spend some quality time (and 220 Grit sandpaper for starters ) to get flat base, usually comcave. (Maybe taking flatness lessons from nvidia?)
Follow w/ 400, 600, 1000 for better contact.

[eagle101]

Kill it with static

you mean if you kill it with static and you have the retail,they give you a new cpu?

[reject]

my new 3000+ seems to be really gay. it wont run 2.4ghz stable with 1.65v
my last one ran 2.5ghz 1.625v
is it the IHS making poor contact or does it need to be burnt in?
i will get another 3000+ when i rma the 2.5ghz one (which my kv8 pro fried)
so should i take it off or sell it if the new one is good?
in memtest i get 2 errors after 25 passes at 270x9, and windows lockups so im pretty sure its the cpu casue it is fine at 2385mhz (265x9)
any thoughts?

okay it crashed at 265x9 now im sitting on 270x8 1.55v

[Liquid3D]

I think it's something else. I traded away my Athlon64 3400 and it's now beating me at my own game. I'm running a P4 530 at 260FSB 3.9GHz, running Hyper-X5400, on an Abit AA8, with a ATI X800 XT (watercooled CPU) and my friend JNav whom I traded away my Chaintech ZNF3, A64 3400 beat me last night with that system and a X800 Pro! I scored 24k and he scored 27k in 3DMark2001.

Bottom line, the on-die MCH is somewhat finicky, and the right memory is needed to get the right OC's. There can be so many different factors, but I'd be inclined to say it's your memory, memory timmings, or voltages. Cooling is a big issue, what are your temps, and your cooling method?

[reject]

zalman 7000 alcu, temps are by probe on side of ihs: 37.5 folding and 30 idle. have seen up to 40 load in the afternoon
the hw monitor thingy says:53-58 load. but i dont trust that as it displays my voltages and psu rails lower than they are.
psu rails are 3.44, 5.17 and 12.24 measured by DMM.

[Liquid3D]

I'd gaurantee in this case your temps are much closer to the internal thermal diode, then a diode placed on the sureface. For several reasons.

1st. Any thermistor even when placed optimally, has approximately 80% of it's sruface area exposed to air. Therefore your temp from that thermistor is taking the average from 10% of it's surface area contacting the CPU's IHS, 10% of it's surface area contacting the Zalman heatsink base, and the remainder of it's surface area exposed to air circulating around it. Another example of the inaccuracy of temp measurement comes from the graph below;



source Arctic Silver
AMD and Intel have made strides to improve the accuracy of their internal thermal diode. While it's true the BIOS formula may be inaccurate, I would always be inclined to trust the higher, rather then lower temp of the BIOS reading over a thermistor. In this way your much safer.

From my personal A64 3400 experience these chips run warmer then the Socket-478 Prescott's at Idle. I found this to be true again and again. Another factor to conisder is the surface area of the IHS itself. Similair to Intel, your A64 IHS occupies 70% more surface area then the actual core beneath it. It's primary job is to "spread heat" hence the name Integrated Heat Spreader (2nd is to protect the core). So if your placing a thermistor on a IHS for temps, and it;s not about 2/3rd of way to the center, your measuring the temp of the IHS, not the core. And while the IHS may be transferring heat from the core over it's surface, it's also transferring the air-temp beneath it. Ever lok at waht's beneath the IHS at it's oyuter edge? Well it's basically the chip pacakage. And you wouldn't place your thermistor on the chip package to take a measurement.

I've never cracked an Intel core and I've easily removed two IHS's one from a 2.4C and one from a 3.0C, and I've cranked down numerous (at least four different) Socket-478 waterblocks onto those Intel cores. SO you can rest assured those IHS's are there because of exponentially increasing heat issues.

Measuring the temp of thermistor (because a thermistor only measures it's own temp) at the side would be analogous to placing the thermistor almost a full CM away (about a half CM) from the center of the IHS. Not to mention the air-flow above from the fan which is removing heat as rapidly as it's generated (if it's doing it's job). My recommendation is the following: watercool the A64 if you can afford it, and my honest opinion is the small amount you'll invest in an mid-level CPU watercooling system, will give you an immediate boost in performance far beyond what air-cooling can. More importantly, H20 would be a long-term investment extending the life of every CPU you cool with it thereafter. It would only cost a few dollars to switch to an Intel 775 or 478 mounting hardware down the road. If you want help, I can get you into a high quality water cooling system for under $180 (for everything needed) that will bring your temps down to a true 31C at full load, and 26C/27C at ilde if your room ambeint temp is 18C.
liquid3d@madshrimps.be

[reject]

wow thanks!
so the gist of it is: if i keep air cooling id need to remove the ihs and then i could also put the diode on the side of the core
-or-
save some money and get watercooling. im thinking to go straight water first
maybe a swiftech 6000, silverprop fusion HL, mag 2 pump, clearfelx hosing and i wanna try evaporative instead of closed sys

[WildKard]

Holy crap Im nude and loving it!

20c drop under load! OMFG

XP 120 with Panaflo medium(everyplace was sold out of 190cfm deltaS?)

with the IHS on I was getting 43c idle 69c load, I could bench and game at over 2600mhz but prime would fail instantly(other than that it was 100% stable odd)

Right now I am priming away at 2600mhz 1.65v @ 48c LOADED! I am amazed 38c idle 48c load, and I am hoping for 45c when the AS settles, and maybe if im REALLY lucky 40-43c when I get the 190cfm delta(with over 2x the airflow of whatI have now)

I think this chip might have 2750 in her on air now no complaints here

[Dmytry]

Hey could you answer this: Is it safe to take off the IHS? I hear the cores are fragile. I don't want to kill my CPU. I'm using the venus 12 (very heavy heatsink).

[WildKard]

Like anything, its ok if youre careful

Just take your time and dont be careless and youll be ok

[aoc007]

Anybody know anything about removing the P4 IHS? I have a northwood c.

[Perc]

Anybody know anything about removing the P4 IHS? I have a northwood c.

as long as its not MO stepping you should be fine. i ruined one it was a 2.4c MO and the ICh was glued to the core and i wound up shatering the core in like 600000 difernt pieces went from the mail box to the trash in 20 mins.... good luck take your time.

peace perc,

[aoc007]

Mines a D1, anybody know anything about them, also would it crush easily when mounting a vapo ls to it?

[reject]

well i got a new 3000+ today from rma, so ill try this.
if they both suck then ill remove the ihs off the old one

[Liquid3D]

Anybody know anything about removing the P4 IHS? I have a northwood c.

I've removed two Northwood IHS's and had no problem. Northwoods manufactured after 12/03 may have been epoxied. However; if your not watercooling it's completely a waste of time, as only the smaller footprint of the waterbock removes heat by isolating impingement on the core. Also Aircoolers such as Thermalright or other aftermarket heatsinks use standoff's to ensure the heatsink base only tightens down so far. Once the IHS is removed your unable to tighten down the heatsink so it makes solid contact onto the exposed core. Once I tried booting up only to find the paste hadn't even been disturbed by the heatsink. I filed down the stand-offs removing the same amount of material to match the thickness of the IHS (about a mm). Northwood

The 2c to 4C drop in temps really isn't worth the potential damage, and there's simply more effective ways to drop temps then nit-picking at your Processor.

[aoc007]

I'll be using a vapo LS, but mine was manufactured in january of this year so I wont mess with it.

[hakme2deth]

With all of those IHS removed, has anyone thought to look at changing the multipliers with the exposed links?

(or can this be done somewhere else -excluding clockgen software)

[Liquid3D]

Great idea and welcome, but the multiplers are open in most BIOS's

[reject]

i think he means upwards