• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Intel Developer Forum 2014

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
"Tablet Innovation: Experiences, Usages, and
Opportunities with Intel’s 2015 Tablet Roadmap" (SF14_TABS002_100f.pdf) touts X86 support for Android SDK apps, NDK apps, and game engines Unity, Marmalde, Epic Unreal and Adobe AIR & Flash. None of this will help apps that are no longer supported or where the code is lost. For everyone else and especially Android gaming applications, does heavy lifting remain to be done? Compiler flags will need to be set to create the correct target. Is code redesign typically required? Will Cherry Trail face a substantively level playing field for Android gaming?
 
TuxDave said:
Wow, this schematic vs layout vs fin count is intense. Fin count is no different from layout count at the end of the day. However the amount that fin/layout count differs from schematic count is higher on a finfet process (assuming iso-performance). Planar transistor can vary its width without counting as a new layout transistor up to a certain point. Finfets, without being able to vary the height of a fin, you can only add more fins.
Click to expand...
So, how many fins does one schematic transistor consist of, on average? I think it's quite high since Mark Bohr said transistors are often made with 2 and 3 or 4 fins and William Holt said it's very uncommon to use 1 or even 2 fins, so wouldn't that mean that if Broadwell on average has 2 fins per schematic transistor, that the apples to Apples comparison would be (in this fictional case) 2.6B transistor for Core M vs 2.0B for A8?

Sanity check: then it would be possible to put 2.82B fins ~ planar transistors in the same area as A8, which is 41% more which means A8 has 30% less transistors, which is close to what Intel claimed (35%), but it's less than 35% because Intel is using a lower density process. Thoughts?

According to Holt, now Intel often uses often 1 or 2 fins. I wonder how the layout transistor count changed from Haswell(-Y) to Broadwell(-Y). Is it half as much? I also wonder what this meant for density. How much does it improve density? Mark Bohr said the 14nm process has a 0.51x neutral density scaling, but the real density scaling is 2.2x (0.45x), so is that difference caused by reduction of fins?
 
witeken said:
So, how many fins does one schematic transistor consist of, on average?
Click to expand...

I would tell you, except I would need to hire you first. 😛

Mark Bohr said the 14nm process has a 0.51x neutral density scaling, but the real density scaling is 2.2x (0.45x), so is that difference caused by reduction of fins?
Click to expand...

I didn't see this quote but it could be several things. Better current/um could improve density as you don't need as large of a transistor to achieve the drive strength you need. Or is he talk about w/ features vs w/o features.
 
TuxDave said:
I would tell you, except I would need to hire you first. 😛
Click to expand...
Okay, any less detailed information you care to share?

I didn't see this quote but it could be several things. Better current/um could improve density as you don't need as large of a transistor to achieve the drive strength you need. Or is he talk about w/ features vs w/o features.
Click to expand...
If you take the transistor estimate (gate*interconnect) and compare it with 22nm, you get a 0.51x scaling, but if you look in the real world, Core M has a 2.2x higher density than Haswell-Y:

14nmFeatureSize.png

BDW-14nm.png


2.2x scaling: http://intelstudios.edgesuite.net/im/2014/pdf/2014_Intel_IM_Holt.pdf page 21

BTW, fast reply. (although I'd rather have a slower but more informative answer)
 
TuxDave said:
The 2.2x area reduction looks like the final result so you can't isolate it to one variable. The improved transistor drive strength helps, but design methodology and design choices can also lead to that 2.2x reduction.
Click to expand...

So the reduction in fins per transistor is negligible, as opposed to what Bohr and Holt have been saying?
 
TuxDave said:
It's definitely not negligible, it's quite significant. I just want to get some credit too. 😛
Click to expand...
Now that I figured you work for Intel, I should try and ask what you are working on? Also, have you ever heard of something called Icelake?
 
witeken said:
Now that I figured you work for Intel, I should try and ask what you are working on? Also, have you ever heard of something called Icelake?
Click to expand...

I can do my best to answer academic questions. So if you want to know what's a schematic, what's layout, what's in a CPU, <insert generic engineering question here>, I'll be glad to help out. But I generally abstain from anything specific to my work. Sorry!
 
TuxDave said:
I can do my best to answer academic questions. So if you want to know what's a schematic, what's layout, what's in a CPU, <insert generic engineering question here>, I'll be glad to help out. But I generally abstain from anything specific to my work. Sorry!
Click to expand...

I already saw that when I just (re)read a whole bunch of your posts. But I tried.

Then what do you work on if we ignore code names?
 
Last edited:
You must log in or register to reply here.

TRENDING THREADS

Back
Top