Ivy Bridge's heat problem is indeed caused by Intel's TIM choice

[vladware]

I did this with an old Opteron. I can't recall the model but it was a dual core 1.8Ghz that I had running at 2.8Ghz with a Zalman Orb cooler.

One of my friends still has it and is using it as a main pc.

Would it be possible to somehow 'balance' a big heatsink in contact with the de-lidded silicon? Kind of like how a GPU heatsink sits on top of the GPUs silicon without using a metal heat spreader in between the silicon and heatsink?

Like the head-on commercial, apply directly to the forehead silicon.

 

[sm625]

Better to just use very low resistance tim and put IHS back on...

Isnt that what they did? They didnt specifically say that they put the IHS back on before testing but I assume they did because if they did not then additional modifications to the cooler and / or socket would have been required.

I wonder if intel did this just to thwart overclockers and hopefully force them into IB-E when it releases? Thanks alot AMD for cornholing us! Intel couldnt do this sort of thing if they had competition.

I wonder if the mobile chips have this same problem....

 

[tigersty1e]

*tinfoil hat*

This could be Intel's unofficial way of deauthorizing the warranty for overclockers. It's fine stock, but if you want to overclock, you have to delid and reapply good tim, hence voiding the warranty.

 

[ShintaiDK]

Isnt that what they did? They didnt specifically say that they put the IHS back on before testing but I assume they did because if they did not then additional modifications to the cooler and / or socket would have been required.

I wonder if intel did this just to thwart overclockers and hopefully force them into IB-E when it releases? Thanks alot AMD for cornholing us! Intel couldnt do this sort of thing if they had competition.

I wonder if the mobile chips have this same problem....

The link goes to vr-zone that then goes to a japanese site. People tend to use google translarte for random results more or less.

Intel didnt "thwart" overclockers. And IB-E is not on any roadmaps.

AMD havent been any competition since 2006. Yet you didnt complain about SB?

Plus you are projecting an overclocking problem to the 99.9% rest of the users.

 

[sequoia464]

You could always use the approach taken by AMD on the old XP line of processors...4 rubber dots on each corner of the PCB to help stabilize the heat sink.

That arrangement was not my favorite. I never crushed anything, but I was always really careful changing out a processor or heatsink.

 

[ReefaMadness]

Of course you need to be careful, but as has been noted many times, the bare processor (no IHS) isn't really rare, and they aren't quite as "fragile" as some may think.

You probably run more of a risk bending pins in one of the newer Intel sockets (LGA775/1156/1366/ ect) than you do swapping the heat sink on an XP processor. It wouldn't be that much different on IB vs what was done with a de-lidded AMD Opteron.

 

[BTRY B 529th FA BN]

Might be from the same source but Legit Reviews is reporting the same thing: Link

 

[zephyrprime]

The problem with going block to bare die (without replacing IHS) is 1) the bare die without IHS is very flimsy, I shorted/fried a cpu/board running one in past. 2) You have to remove the socket, and also file/grind down area around the socket, otherwise fail to get adequate pressure. 3) even if you manage to do it, I wouldnt run 24/7, because again the bare die will be so flimsy will tend to bend, cause inadequate contact, short and fry.

Better to just use very low resistance tim and put IHS back on...wonder if could use indigo extreme between die and IHS.

Give me a break! Everyone used to run with bare dies in the Athlon XP days including myself and we all still run bare dies on our GPUs. It's not that risky. Inadequate mounting pressure is an issue though.

 

[frozentundra123456]

OK, sorry to sound stupid, but could someone explain exactly what these abbreviations mean and exactly what Intel and the tech site did? I am not at all familiar with the internal parts of a CPU. Dont know what TIM and IHS mean even. Sorry for my lack of knowledge, but I am a bit confused by what is going on here.

 

[MPiland]

Might be from the same source but Legit Reviews is reporting the same thing: Link

Same one. They're just reposting from the OP's link.

 

[Binky]

Give me a break! Everyone used to run with bare dies in the Athlon XP days including myself and we all still run bare dies on our GPUs. It's not that risky. Inadequate mounting pressure is an issue though.

Do you really not recall the posts about cracked cores? I broke at least one exposed Athlon core myself.

 

[Ferzerp]

I chipped the die on the one open die proc I had (coppermine celeron 2). It still worked, but I torqued the golden orb I was installing poorly and chipped it.

 

[SickBeast]

I'm quite surprised it took people this long to figure this out, and I'm even more surprised that this got past Intel's quality control people.

Hopefully Intel will revise the chips so people don't have to void their warranty.

 

[Ferzerp]

It's still only a single anecdote. I'm considering testing it myself though, only with after results though. I won't have before results.

 

[SickBeast]

It's still only a single anecdote. I'm considering testing it myself though, only with after results though. I won't have before results.

Yeah be careful still. Someone else did it and they had no drop in temps whatsoever.

 

[john3850]

I chipped the die on the one open die proc I had (coppermine celeron 2). It still worked, but I torqued the golden orb I was installing poorly and chipped it.

I chipped the die on the one open die proc I had (coppermine p-3 800) also.
It very tiny chip but at times it refused to start after the chip.

 

[Ferzerp]

The only reason mine refused to start was that the Abit board I had at the time had a pretty big flaw.

Custom FSB was set on power on, but custom voltage wasn't set until after post. I ended up pin modding the thing for higher voltage, but since I developed my own method, I wrapped wire (from an old toy DC motor) around the pins instead of just jumpering the correct socket holes. It worked, but it was a ton more work to wrap that tiny wire around the tiny pins and then tie it off without shorting any of the other ones.

 

[GroktheCube]

This doesn't really surprise me based on the behavior I've seen, but I'm not willing to do anything too ballsy personally.

I'm running an XSPC Raystorm block and an XSPC 240EX radiator on my new 3570k. In my IBT stress testing, temperatures rises quickly to a certain level depending on voltage where it holds stead after 60 seconds or so. In the case of my best IBT Max Stress stable 4500 OC, ~75-78C for the warmest core. When load drops, core temps are back down to 27C (about 3-4 over room temp) in 10 seconds, no matter how long the test has been running.

This seems to indicate that the water cooling system is able to move heat way from the IHS very quickly, but the processors simply can't transfer heat to the IHS quickly enough to maintain steady temps at high voltages/speeds. As soon as heat production drops, temperatures stabilize.

On the upside, my chip seems to do pretty well at low voltages. 4400 is 3 hours of IBT Max stable @1.2v (under load) and ~122gflops, with temps staying well within acceptable margins. Under "normal" 100% load (like running Fold @ Home and wow or Prime 95), temperatures stay around 50C. 4200@>1.1v yields temps in the low 40s in Prime 95, but I haven't tested it in IBT yet to see what kind of heat it generates and what kind of numbers it pulls.

I do wish that the TIM intel used was a bit better, but the chip still blows away my old 965BE, and performs very well at moderate speeds/voltages.

 

[Ferzerp]

Which stress level? The 12GB memory one? The rating seems to increase depending upon which you use. 1GB is only like 97-98 GFLOPS at 4.4 for me, and then higher levels rate higher.

Mine does 4.4 with ~1.208V I think it is. Right now I have it stock until some other materials arrive.

 

[GroktheCube]

OK, sorry to sound stupid, but could someone explain exactly what these abbreviations mean and exactly what Intel and the tech site did? I am not at all familiar with the internal parts of a CPU. Dont know what TIM and IHS mean even. Sorry for my lack of knowledge, but I am a bit confused by what is going on here.

TIM= Thermal Interface Material, like what goes between your heat sink and the processor.

IHS = Internal Heat Spreader. This is in charge of getting heat from the CPU cores to your heat sink.

 

[GroktheCube]

Which stress level? The 12GB memory one? The rating seems to increase depending upon which you use. 1GB is only like 97-98 GFLOPS at 4.4 for me, and then higher levels rate higher.

Mine does 4.4 with ~1.208V I think it is. Right now I have it stock until some other materials arrive.

Yup, max is 12GB IIRC (it might actually be 14). I do just about 108gflop at 4.4 on the 1GB setting.

Honestly, these chips are pretty fast at any speed, definitely more than fast enough for anything I'm doing.

 

[rge2]

Give me a break! Everyone used to run with bare dies in the Athlon XP days including myself and we all still run bare dies on our GPUs. It's not that risky. Inadequate mounting pressure is an issue though.

obviously you have never run a bare die cpu like I have, since you are comparing gpu bare die to cpu, they have nothing in common. People that have done it, know exactly what I am talking about by very flimsy bare die that easily warps without IHS, people that havent, wont.

 

[rgallant]

obviously you have never run a bare die cpu like I have, since you are comparing gpu bare die to cpu, they have nothing in common. People that have done it, know exactly what I am talking about by very flimsy bare die that easily warps without IHS, people that havent, wont.

-being 3d you got to wonder if intel had to use tim to act as a cushion to the die.
-more chunky than a 2d film , more filler\top coat even at the nm scale
-solder would not give any pressure relief\flexing of the die.

 

[bononos]

The problem with going block to bare die (without replacing IHS) is 1) the bare die without IHS is very flimsy, I shorted/fried a cpu/board running one in past. 2) You have to remove the socket, and also file/grind down area around the socket, otherwise fail to get adequate pressure. 3) even if you manage to do it, I wouldnt run 24/7, because again the bare die will be so flimsy will tend to bend, cause inadequate contact, short and fry.
........

obviously you have never run a bare die cpu like I have, since you are comparing gpu bare die to cpu, they have nothing in common. People that have done it, know exactly what I am talking about by very flimsy bare die that easily warps without IHS, people that havent, wont.

I agree about your comment about possible problems with the socket blocking the HS from attaching with adequate pressure if the IHS is removed.
But how does the bare die bend/warp so easily, it tends to crack rather than bend. And even if it does bend/warp what component exactly gets shorted? And why does bare die cpu work differently from the gpu when attaching a HS?

 

[Ajay]

Give me a break! Everyone used to run with bare dies in the Athlon XP days including myself and we all still run bare dies on our GPUs. It's not that risky. Inadequate mounting pressure is an issue though.

Well, they weren't *bare*. They had a very hard thin epoxy coating on them. I always used a shim (you could buy them at a variety of specialty online stores) and never cracked a core. Thermalright put out some nice copper shims.