CPU production

[ItsMinJunLol]

Hi! I don't have much knowledge on CPUs and processors but why can't manufacturers just make the chip big? That way they can add more transistors and cores, etc. Like, they should make a socket like LGA 1151 XL that has this massive cpu with 40 cores and 6 ghz. I'm not sure but it's just been in my head for a while. Can someone tell me why companies like Intel doesn't do this? thanks!

[patrickjp93]

6 minutes ago, ItsMinJunLol said:

Hi! I don't have much knowledge on CPUs and processors but why can't manufacturers just make the chip big? That way they can add more transistors and cores, etc. Like, they should make a socket like LGA 1151 XL that has this massive cpu with 40 cores and 6 ghz. I'm not sure but it's just been in my head for a while. Can someone tell me why companies like Intel doesn't do this? thanks!

Silicon wafers have a given average defect density. The bigger the die, the higher the risk a chip comes out non-functional because of a defect in the silicon. Then there are additional risks with ever laser pass and chemical dispersion. The bigger the die, the higher every risk becomes that the die will come out dead. You also reduce the viable die count per wafer and increase the waste area.

 

Intel's biggest dies were 720mm sq. on the 90nm platform iirc, and that seems like a hard limit most foundries don't want to exceed.

[SLAYR]

They do have bigger dies then lga 1151 ,those are dwarfed by the server socket sizes of amd g34,  or intel 1366, 2011, 2011-3, or the monster lga 3647 which has up up to 72 cores. 

 

http://ark.intel.com/m/products/95831/Intel-Xeon-Phi-Processor-7290F-16GB-1_50-GHz-72-core#@product/specifications

 

[EthanCGamer]

I'm by no means an expert here, but I'll give it a shot. (I could be wrong on a few of these)

 

I can think of a few reasons:

1. The current ATX standards and size of motherboards and components on them limit how big a CPU can be without taking other component's space.

2. While a bigger die can contain more cores, this also means that the complexity and cost to design and manufacture increases. At a certain point, it becomes too complex and costly to put more cores on, which most users won't even use.

3. Power limits: If a hypothetical CPU was made that had 40 cores each running at 6 ghz, the power consumption would be massive. High end Xeons run at about 130 Watts, where the 6700K maxes out at 91 Watts. If we take the Xeon E7-8890 v4 that runs at 165 watts with 24 cores at a max of 3.4 Ghz (not all at once) and make some extremely rough estimates, the processor you propose would run at about 500 Watts or so, depending on the load. Of course, no one at home will hit that, unless you're doing some extreme protein folding.

4. You can't have a high core count AND a high clock speed, not yet anyway. Xeons have lots of cores (up to 72!), but run at about 2 Ghz on average, while i7's have 4 or so cores that run at ~4 Ghz on average. 

 

If you look at the ARK data for the CPU @SLAYR posted, the cores don't even get above 2 Ghz. It's simply not possible to do that with current manufacturing processes we have. Also, I can't recall any CPU that is factory clocked at anything above 4.5 Ghz. The record clock speed for any CPU was about 8.8 GHz, and that was with some insane cooling. Maybe in the next 10 years the enterprise grade gear could get to that level, but for now we just have to wait.