Intel ‘Devil’s Canyon’ chips’ NGPTIM is still not efficient

[Constantine]

I unbderstand where you are coming from but a company such as Intel with what their capable of, I'm sure they would be able to figure it out. And fair enough even if it was the case what you've just mentioned, if thermal paste was the only option then why use that god aweful shit that they used with both Ivy Bridge and Haswell.

 

If it was me and I owned Intel, if it turned out that we couldn't solder anymore then I would have employed a decent thermal paste at least which to me suggests that it is infact money and they won't do it.

 

They're working on improving it, but they're going to take their sweet ass time with it since they already monopolize the CPU market

[FireFox]

They're working on improving it, but they're going to take their sweet ass time with it since they already monopolize the CPU market

 

Yea that sucks, I want to see ARM and NVIDIA come onto the desktop CPU market. I'd definetly buy an NVIDIA CPU

 

Also they are working on improving it, they should contact Noctua or Arctic Cooling, there thermal pastes seem pretty damn good and I've used AC for quite a while.

[Builder]

I unbderstand where you are coming from but a company such as Intel with what their capable of, I'm sure they would be able to figure it out. And fair enough even if it was the case what you've just mentioned, if thermal paste was the only option then why use that god aweful shit that they used with both Ivy Bridge and Haswell.

 

If it was me and I owned Intel, if it turned out that we couldn't solder anymore then I would have employed a decent thermal paste at least which to me suggests that it is infact money and they won't do it.

It's not the paste, there's a gap in between the IHS and die that is no longer completely filled. That's why you see massive improvements no matter if you use crappy aftermarket TIM or Coollaboratory Liquid Ultra. Granted, improvements are higher with CLU, however even with Ceramique there is enormous improvement.

 

Sometimes it's not just a matter of figuring it out either, it's chemistry, there are sometimes things that you just can't do. I think they've decided that we're a small enough portion of the market that it's more worthwhile to push the heavy overclockers and enthusiasts onto LGA2011 and leave the gamers in the 115x sockets instead of spending possibly millions of dollars more to make 115x as overclockable as they used to be. I predict Devil's Canyon will be their last enthusiast-oriented 115x chip.

 

Yea that sucks, I want to see ARM and NVIDIA come onto the desktop CPU market. I'd definetly buy an NVIDIA CPU

 

Also they are working on improving it, they should contact Noctua or Arctic Cooling, there thermal pastes seem pretty damn good and I've used AC for quite a while.

AC on Haswell dies is a bad, bad idea. It has conductive silver in it and Haswell has SMD capacitors right next to the die. Short them out = bye bye CPU. You can still RMA it believe it or not but I can't wait that long. Ain't nobody got time fo that. 

 

ARM doesn't make CPUs, they license standards to Nvidia and smartphone manufacturers. They design chips in accordance with the standards and then sell them.

 

A desktop ARM chip would be pretty much useless, they're designed for low power consumption and plus they use the ARM instruction set, and they're RISC on top of all those other problems. Windows would have to be ported, unless of course you'd like to try Linux or BSD ...

[FireFox]

It's not the paste, there's a gap in between the IHS and die that is no longer completely filled. That's why you see massive improvements no matter if you use crappy aftermarket TIM or Coollaboratory Liquid Ultra. Granted, improvements are higher with CLU, however even with Ceramique there is enormous improvement.

 

Sometimes it's not just a matter of figuring it out either, it's chemistry, there are sometimes things that you just can't do. I think they've decided that we're a small enough portion of the market that it's more worthwhile to push the heavy overclockers and enthusiasts onto LGA2011 and leave the gamers in the 115x sockets instead of spending possibly millions of dollars more to make 115x as overclockable as they used to be. I predict Devil's Canyon will be their last enthusiast-oriented 115x chip.

 

OK then, save millions and not develop a better thermal paste and go back to solder. I still think it is just Intel being f**king wankers.

[Builder]

OK then, save millions and not develop a better thermal paste and go back to solder. I still think it is just Intel being f**king wankers.

Like I've said before, it's chemistry, the new dies are so small that the old solder may be out. I believe they were just using standard solder paste, which is small balls of solder suspended in flux. When you reflow it, the balls melt and the parts are welded together. They may be having troubles with adhesion and on top of that as I said before there are capacitors right next to the die that you can short very easily if not careful. It's possible that those SMD capacitors may also desolder when they reflow the paste, there's a lot of variables you aren't taking into account and with silicon fabs nothing is ever as easy as "just do X."

[FireFox]

Like I've said before, it's chemistry, the new dies are so small that the old solder may be out. I believe they were just using standard solder paste, which is small balls of solder suspended in flux. When you reflow it, the balls melt and the parts are welded together. They may be having troubles with adhesion and on top of that as I said before there are capacitors right next to the die that you can short very easily if not careful. It's possible that those SMD capacitors may also desolder when they reflow the paste, there's a lot of variables you aren't taking into account and with silicon fabs nothing is ever as easy as "just do X."

 

Intel have started full-scale 14nm production of their chips which is an amazing feat and which has been a major bastard aswell and yet they can't figure this out. Intel aren't stupid and it is well within their capability.

 

If I'm not mistaken Ivy Bridge-E has the same thickness IHS and silicon die yet that is soldered and performing perfectly fine, Mmmmm OK.

 

Another thing, why is there a gap between the IHS and the chip because surely Nehalem and Sandy Bridge were the same thickness.

[Builder]

Intel have started full-scale 14nm production of their chips which is an amazing feat and which has been a major bastard aswell and yet they can't figure this out. Intel aren't stupid and it is well within their capability.

 

If I'm not mistaken Ivy Bridge-E has the same thickness IHS and silicon die yet that is soldered and performing perfectly fine, Mmmmm OK.

 

Another thing, why is there a gap between the IHS and the chip because surely Nehalem and Sandy Bridge were the same thickness.

LGA2011 chips are much bigger, surely you know that. The larger surface area makes up. I'm not talking about thickness, thickness barely changes. I'm referring to surface area.

 

Sandy Bridge and Nehalem used solder, they were essentially welded to the IHS. That's why there wasn't a gap, if there was a gap it was filled with molten metal.

 

The gap is caused by the way in which they apply the TIM, I'm not sure quite how they do it. Even then, it's a relatively small gap and thermals are fine if you aren't pushing the limits of your chip. The one thing that's sure is that they don't care too much about enthusiasts using the 115x line chips. They want us to buy LGA2011 stuff.