Intel's Core i7-4960X Exposed: CPU Die Soldered to the IHS

[N.Broekhuijsen]

A tweaker from the Coolaler forums has de-lidded an i7-4960X and noted that the CPU die is soldered to the IHS.

Intel's Core i7-4960X Exposed: CPU Die Soldered to the IHS : Read more

 

[ingtar33]

Well, the theory for why Intel used thermal grease and not solder on Haswell and Ivy Bridge, the one that said they couldn't be soldered because of the shrunken die size is now proven untrue.

No other explanation then Intel either was saving money or artificially gimping Haswell/IB or both.

 

[ojas]

But than wont the IVB-E die be bigger than standard?

 

[basketcase87]

With the LGA 1155 Ivy-Bridge chips a handful of users would opt to remove the lid from their chips in order to get around Intel's decision to solder, though obviously at the cost of warranty.

Should be Intel's decision NOT to solder (or Intel's decision to use crappy TIM).

 

[RedJaron]

Well, the theory for why Intel used thermal grease and not solder on Haswell and Ivy Bridge, the one that said they couldn't be soldered because of the shrunken die size is now proven untrue.

No other explanation then Intel either was saving money or artificially gimping Haswell/IB or both.

Possibly. It could also be that Intel needed extra time to devise a solder method for the smaller die. Not saying you're wrong, but I'm willing to give Intel the benefit of the doubt until proven wrong.

This does mean that Intel has no excuse to not solder normal Ivy Bridges now, though.

 

[dragonsqrrl]

This shouldn't come as much of a surprise. I think it's pretty clear now what Intel is doing, if it wasn't abundantly clear already. The whole point of the thermal compound on Ivy and Haswell isn't to pinch pennies or out of necessity as some have suggested. It's a way to further segment their desktop line-up by artificially limiting the overclocking potential of their LGA 1155 processors. This gives both Sandy and Ivy Bridge-E a sort of 'manufactured' advantage over Ivy and Haswell by offsetting any IPC disadvantage with additional overclocking headroom. It's a clever move on their end, it just sucks for the average enthusiast. Of course I got thumbed down to oblivion when Ivy Bridge first released for even suggesting that this might be the case. The prevailing theory at the time seemed to be that the engineers and scientists at Intel simply didn't know what the hell they were doing... lol.

 

[SteelCity1981]

it's an engineering sample the 4770k ES's didn't use thermal paste either, but the OEM versions do. So take this article with a grain of salt.

 

[nokiddingboss]

and i thought amd was cheap... intel is cheaper but in a bad way. naughty naughty naughty. so i guess they can use their unused tp's to grease up their own butt's now that they are back to soldering.

 

[balister]

With the LGA 1155 Ivy-Bridge chips a handful of users would opt to remove the lid from their chips in order to get around Intel's decision to solder, though obviously at the cost of warranty.

Should be Intel's decision NOT to solder (or Intel's decision to use crappy TIM).

Actually, the problem isn't that they used the TIM (although soldering does improve the heat transfer slightly), the real problem is that there is a very tiny gap between the TIM and the IHS causing the heat transfer from TIM to the IHS to end up being radiative instead of conductive (radiative is the worst form of heat transfer). This has been proven a number of times from people that have de-lidded both IB and Haswell CPUs.

 

[wh3resmycar]

lol. leaking something and you can only leak "one" photo? fishy fishy.

 

[physical]

Being an engineering sample this doesn't really mean very much. IIRC the IB engineering samples were also soldered on.

De-lid a production chip and get back to me.

 

[eklipz330]

a lot of people complained about them using TIM in Ivy_bridge, but only a few people that delidded actually saw temperature improvements from the forums that i scoured when i was interested.

not worth it unless your temps are seriously out of line. and if you're watercooling.

 

[zero2dash]

http://www.coolaler.com/showthread.php/305405
You forgot to say that he ripped the core in half and killed the CPU in the process.

 

[Chaoss]

So basically 100mhz faster than the current top of the line CPU. Not all that impressive a difference

 

[alextheblue]

I agree with Physical, we need production chips to tell for certain. Anyway if this is only on their hexacore chips I'm not all that interested.

 

[JPNpower]

Remind me how you can safely rip off a soldered on heatsink from a flimsy piece of silicon.

Also, I read that using TIM is in order to reduce costs, not because the TIM process is much cheaper, but because the soldering process is easily botched and leads to more failures. The TIM process gives good enough performance at a better success rate. Also, I heard that the small gap between the heatspreader and die is good for big heavy heatsinks, as they tend to crush the die otherwise.

All of this I got from a Toms commenter who probably posted above me already

 

[vaughn2k]

Just a question though... unless otherwise this is Ag sintered or the die has a sintered TiNiAu (where both process are also very expensive), how does Intel manage the adhesion of solder to silicon? I think this is a hoax...

 

[cats_Paw]

I dont really see much reason to trust this guy...

 

[sarinaide]

I love these threads, just for the tin foil hat brigade.

 

[InvalidError]

how does Intel manage the adhesion of solder to silicon? I think this is a hoax...

They can do copper deposition using the same process as normal metal layers, use a chemical plating step to apply tin to that copper and solder the IHS to that. This is basically the same thing that happens on the side where uBGA balls are attached.

 

[Duckhunt]

Well as Intel starts to release that AMD is less and less relevant. It will try to wring out every dime it can. Just wait. The bean counters are sitting around trying to work out how to make themselves richer. The question is if the shareholders will also get richer.

 

[shadowfamicom]

Hmm... still no word on new 2011 socket CPUs?

 

[InvalidError]

Hmm... still no word on new 2011 socket CPUs?

Last time I checked, i7-4960X which this article was about is exactly that and expected later this year.

 

[sarinaide]

Hmm... still no word on new 2011 socket CPUs?

The i7 4960X is a Intel Extreme Socket 2011 chip

 

[shadowfamicom]

Op my mistake must have read it wrong!