Roadmap to 3nm Transistors.

[Katness Everdeen]

Beyond 14nm, as we move to 10 and 7nm, a new fin material will be required — probably silicon-germanium (SiGe), or perhaps just pure germanium. SiGe and Ge have higher electron mobility than Si, allowing for lower voltages, and thus reducing power consumption, tunneling, and leakage. SiGe has been used in commercial CMOS fabrication since the late ’80s, too, so switching from silicon won’t be too painful. (The primary reason that we’ve been using silicon for so long is that the entire industry is based on silicon. The amount of time, money, and R&D that would be required to deploy new machines for handling new materials that we know relatively little about would be astronomical.)

http://www.extremetech.com/computing/162376-7nm-5nm-3nm-the-new-materials-and-transistors-that-will-take-us-to-the-limits-of-moores-law

 

 

[M-ursu]

I can just imagine that heat of 3nm transistors....

[Kuzma]

Quantum computing is imminent

[qwertywarrior]

guys

remember this day and time

 

because when we are old

 

pc enthusiasts OCing etc  wont exist

 

we will be like the last generation

kinda like asking grand dad how things were back in the day

[Beskamir]

guys

remember this day and time

 

because when we are old

 

pc enthusiasts OCing etc  wont exist

 

we will be like the last generation

kinda like asking grand dad how things were back in the day

why wouldn't OCing exist? (sorry if the question sounds stupid but it 4:30am so please don't kill me :P)

[qwertywarrior]

why wouldn't OCing exist? (sorry if the question sounds stupid but it 4:30am so please don't kill me :P)

its something i just believe will happen

[LAwLz]

I can just imagine that heat of 3nm transistors....

The heat would be lower than what he currently got. The whole "ohh Ivy Bridge is hotter than Sandy Bridge because they use smaller transistors" is bullshit Intel pulled to try to cover that they use crappy TIM for Ivy Bridge.

[Kuzma]

The heat would be lower than what he currently got. The whole "ohh Ivy Bridge is hotter than Sandy Bridge because they use smaller transistors" is bullshit Intel pulled to try to cover that they use crappy TIM for Ivy Bridge.

It's not bullshit; the transistors are smaller so they're closer together and have a higher density.

[nicholaswillems]

If they manage to use those transistors very efficiently and at lower power, then the heat won't be too high, I think.

[Frodo]

It's not bullshit; the transistors are smaller so they're closer together and have a higher density.

Higher density yes, but still the heat emitted per transistor is lower which means that at the same number of transistors it has a lower heat output.

[Kuzma]

Higher density yes, but still the heat emitted per transistor is lower which means that at the same number of transistors it has a lower heat output.

Cbb to explain >_> you're obviously set on your idea.

[Chevaishr]

smaller die plus plus slightly lower power consumption would result in higher heat reason being that there is less area to dissipate that heat and with  the use of TIM instead of soldering only makes it worse. 

[LAwLz]

It's not bullshit; the transistors are smaller so they're closer together and have a higher density.

And Intel cheaped out on the TIM so the heat is not transfered to the heatspreader properly. Some people are getting 20+ degrees difference from simply swapping out the horrible TIM Intel uses for some higher quality thermal paste. Again, the whole "och it's hotter because the heat is more concentrated" is just something Intel said to cover up that they have started to cut corners. If they changed back to fluxless solder then the temps on Ivy and Haswell would most likely be the same or lower than Sandy Bridge.

[Kuzma]

And Intel cheaped out on the TIM so the heat is not transfered to the heatspreader properly. Some people are getting 20+ degrees difference from simply swapping out the horrible TIM Intel uses for some higher quality thermal paste. Again, the whole "och it's hotter because the heat is more concentrated" is just something Intel said to cover up that they have started to cut corners. If they changed back to fluxless solder then the temps on Ivy and Haswell would most likely be the same or lower than Sandy Bridge.

xD I never denied the TIM; all I'm saying is if they used fluxless solder then the temps would still be higher.

[YellowDragon]

Cbb to explain >_> you're obviously set on your idea.

Actually he is right and wrong at the same time :P, though you have lower heat output overall, although with the transistors packed so closely together that  you have a higher heat density in a smaller space, coolers have no problems dissipating TDPs of 200w+, however thats if the source of the heat is evenly spread across the cold plate, you concentrate where that heat comes from and it becomes harder and harder to cool as you have less surface area to work with, It dosent help coolers that even the heat spreader has trouble effectively drawing away heat from a processor with a lot of resistors packed close together, even with a good application of TIM. In order for air coolers and water cold plates to function they need surface area to work with as the transfer to one material to the next is not instantaneous and loss less, with tighter packed transistors you are denying a cooler its very reason that it functions properly in the first place.

 

Personally I wonder if it would be possible to interlace coolant channels onto the die itself, actually start with two small copper pipes and build transistors around it and then isolate them electrically from each other that way you avoid the heat spreader altogether and you get a cooling system as physically close as possible to the source of the heat.

[Kuzma]

Actually he is right and wrong at the same time :P, though you have lower heat output overall, although with the transistors packed so closely together that  you have a higher heat density in a smaller space, coolers have no problems dissipating TDPs of 200w+, however thats if the source of the heat is evenly spread across the cold plate, you concentrate where that heat comes from and it becomes harder and harder to cool as you have less surface area to work with, It dosent help coolers that even the heat spreader has trouble effectively drawing away heat from a processor with a lot of resistors packed close together, even with a good application of TIM. In order for air coolers and water cold plates to function they need surface area to work with as the transfer to one material to the next is not instantaneous and loss less, with tighter packed transistors you are denying a cooler its very reason that it functions properly in the first place.

 

Personally I wonder if it would be possible to interlace coolant channels onto the die itself, actually start with two small copper pipes and build transistors around it and then isolate them electrically from each other that way you avoid the heat spreader altogether and you get a cooling system as physically close as possible to the source of the heat.

Sounds interesting and could work :p

[LAwLz]

Actually he is right and wrong at the same time :P, though you have lower heat output overall, although with the transistors packed so closely together that  you have a higher heat density in a smaller space, coolers have no problems dissipating TDPs of 200w+, however thats if the source of the heat is evenly spread across the cold plate, you concentrate where that heat comes from and it becomes harder and harder to cool as you have less surface area to work with, It dosent help coolers that even the heat spreader has trouble effectively drawing away heat from a processor with a lot of resistors packed close together, even with a good application of TIM. In order for air coolers and water cold plates to function they need surface area to work with as the transfer to one material to the next is not instantaneous and loss less, with tighter packed transistors you are denying a cooler its very reason that it functions properly in the first place.

 

Personally I wonder if it would be possible to interlace coolant channels onto the die itself, actually start with two small copper pipes and build transistors around it and then isolate them electrically from each other that way you avoid the heat spreader altogether and you get a cooling system as physically close as possible to the source of the heat.

It's the TIM's job to transfer the heat from the CPU die to the IHS, and Intel cheaped out on that part starting with Ivy Bridge. You get a ~20 degrees difference if you change the crappy stuff Intel uses to for example Freeze Extreme (on stock clocks, ~30 degrees difference at 4.6GHz according to some tests) or Indigo Xtreme.

Again, it's not the whole "more stuff in a smaller area" to blame, it's simply Intel trying to cut corners. They might even do it on purpose to hinder people from buying cheaper chips and just overclocking them.

Smaller manufacturing process != Higher temps.

Shitty job with the thermal design and trying to save a few cents for each processor == higher temps.

 

As for the "put heatpipes on the die" idea, it doesn't sound like a good idea. I mean, where would you put them? You can't just put them between cores or something like that

 

Interesting tests:

TIM intel i7 3770K Review | AS5 | MX4 | PK1 | LQP | IX [noticias3d.com]

TIM is Behind Ivy Bridge Temperatures After All [TechPowerUp.com]

Removed the IHS from an Ivy Bridge i5 3570K [Hardforum.com] (reports on lower temps by simply swapping the TIM is on page ~7 or so, and then forward from there).

 

I am not trying to bash Intel. Hell, I get called an Intel fanboy/shill several times a week, but they did make their processors worse by not using fluxless solder between the die and the IHS.

[Glenwing]

Ivy Bridge is certainly not the first time we've done transistor process shrinks.  Suddenly it creates more heat, coincidentally at the same time they start using crappy thermal compound?  I don't think so.  It wasn't a problem when we moved from 40nm to 28nm on GPUs.  Quite the opposite in fact.  This is HardwareCanuck's comments on Westmere (the die shrink on Nehalem, like Ivy was to Sandy):

 

"We used Real Temp 3.50 RC6 Beta to collect these temperature figures, and if the results are correct the i5-661's cool running nature is simply betrayed by the mediocre stock cooler that Intel bundles with their new mainstream chips. Having said that, with a high-end CPU cooler like the TRUE the temperatures were extremely low, a testament to the new 32nm process. We believe that passive cooling might definitely be a possibility with Clarkdale, at least in systems with decent case airflow."

 

Same story 2 generations before that with Penryn, when we moved to 45nm...

 

"The new processors, which are now produced on a 45 nm fabrication process, not only consume phenomenally little energy, they also offer outstanding overclocking potential. Even with the processor overclocked to its limit, its thermal dissipation and power consumption are almost on par with those of today's Core 2 CPUs - at their default settings."

 

It makes sense too; smaller transistors require less voltage to switch, and voltage is what creates heat.  Furthermore Intel also switched to a triple gate transistor design, which reduces electron leakage and lowers power consumption even more on top of the process shrink.  Being packed closer together won't matter much; it is just about the amount of heat generated overall, and the surface area of the die, which hardly changes, while the power consumption drops a lot.

 

Higher temperatures have never been a result of a process shrink, not once, not in CPUs and not in GPUs.  From where I stand, Intel's excuse for Ivy Bridge's temperatures is a flat-out lie.

[martinrox1568]

It's not bullshit; the transistors are smaller so they're closer together and have a higher density.

They also use less voltage because each transistor is closer together

[dyent10]

They also use less voltage because each transistor is closer together

the proximity of transistors have nothing to do with voltage levels. unless your using very high voltages that it creates a strong magnetic field around the transistors. in which case is irrelevant to your post.

 

yes there are voltage drops from one end of the bus to the other, but its only relevant in longer distance.