2013 core sizes: A7-A15-Jaguar-Atom-Haswell

[witeken]

Intel also said Broxton is coming mid-2015, think logical about it and don't trust in every marketing slide and comment from Intel. Simply forget 58mm², this is nonsense.

Here's my source: http://intelstudios.edgesuite.net/im/2013/archive/ss/archive.html (26:15)

 

[witeken]

Intel 32nm Clovertrail core die size = 5.6mm^2
Intel 22nm(50% reduction) Baytrail core die size = ~2.8mm^2

With 64% reduction from 14nm, Airmont core die size should be close to ~1.01mm^2

How did you come to 0.82mm^2 ???

22nm scaled by a factor of up to 1.7x, not 2x.

 

[witeken]

Going by the same scaling metrics, Sofia 3G would work out at ~25mm^2. (Exactly four of its dies fit into the bay trail one, give or take a bit of rounding)

Given that it is being fabricated on a 28nm process and won't have significant density improvements, is this reasonable?

Yes, and when they port it to 14nm in 2016, the die size will be tiny, with huge power/performance improvements.

 

[Nothingness]

From nowhere, that would imply that a silvermont core is 2.8mm2 while it s 3.5mm2...

According to Hiroshige Goto it is 2.x mm2. And I consider him as a reliable source 🙂

 

[sm625]

Can we get broadwell, arimont, and both Cortex A5x dies appended to the OP's image?

 

[Enigmoid]

According to Hiroshige Goto it is 2.x mm2. And I consider him as a reliable source 🙂

He is not counting the Bus Interface Unit. The other 3.5 mm^2 likely is.

 

[Nothingness]

He is not counting the Bus Interface Unit. The other 3.5 mm^2 likely is.

So an Airmont core alone size should be 1.3mm2 and not the 0.82 mm2 IntelUser2000 quoted. Hence A15 is still smaller :biggrin:

 

[kimmel]

So an Airmont core alone size should be 1.3mm2 and not the 0.82 mm2 IntelUser2000 quoted. Hence A15 is still smaller :biggrin:

My estimated number is smaller than your estimated number. :awe:

 

[IntelUser2000]

Intel 32nm Clovertrail core die size = 5.6mm^2
Intel 22nm(50% reduction) Baytrail core die size = ~2.8mm^2

With 64% reduction from 14nm, Airmont core die size should be close to ~1.01mm^2

Clovertrail is not 5.6mm2. Hans just cropped the entire square section of the chip while bus unit is smaller than what he suggests. Your second picture on the white highlighted section points out the core+bus which turns out to be 5.0mm2, based on the information that CPU+L2 on PCWatch's pic for Silverthorne is at 14mm2. That's not relevant to figuring out Airmont core size though.



Let's see how you calculate the core size

1. According to PCWatch picture, the Bonnel core + L2 = 14mm2
2. Because the image is according to scale, you find out that Silvermont + L2 = 8.5mm2
3. Take this picture: http://www.expreview.com/39814.html
4. Now pixel-count based on that Silvermont core there is 8.5mm2. You end up with...

2.1mm with BIU for 2 cores
1.67mm w/o BIU for 2 cores

That's 0.835mm2 for one core. They probably have more room for optimization in Goldmont. With Silvermont we got practically same core size as Bonnell despite having 50% better IPC.

A15 probably has more transistors.

Irrelevant. A15's IPC is approximately on par with Silvermont. Therefore, its right to compare them. Who cares about architecture and ISA idiosyncrasies? Really, only products matter.

 

[IntelUser2000]

Ashraf measures about 87mm^2 for Cherry Trail (http://www.fool.com/investing/gener...rporation-cherry-trail-die-size-revealed.aspx),

Yea, the 7420 is actually a surprise. So of course everyone fudges their numbers and that includes Intel. It may be that some companies for competitive reasons put conservative claims and go aggressive in actual products.

I don't think the same Ashraf measured it right though. Because using actual die pic you can get it within 2-3%, but you can't with package. If you can take the Airmont core shot and scale it up to Airmont die pic(after you photoshop the pic to be able to compare it) then we should be able to find approximate die size.