It is happening: 10 nanometer Chip technology is here


FBnil

They'll own everything and be miserable.
Joined
Dec 14, 2012
Messages
5,114
Location
Yurp
disclaimer: The power consumption comes mostly from the screen background leds.

So, in 2014 ARM released the A15 (28mn) technology. Which was available one year later to non-insiders. Look at the power reduction...
powerconsumption.png

And now, in 2017 Artemis arrives on 10nm, and look at the sustainable burst:
8-1.png

Smaller chip means less power and means less heat.

However: not all is gold there. They changed the 3D from Mali to Mimir (which probably means incompatibility once the daughterboard is replaced... until the software catches up.)
Now ARM makes the SoC, but this isn’t an SoC that ARM will ever bring to market. It licenses out their design, and somebody (big chip companies) would have to make them (and pay for the license) before consumers can buy them

So this happened last month:
https://www.extremetech.com/computi...rm-chips-for-lg-on-upcoming-10nm-foundry-node
And SnapDragon will follow in Q1 of 2018:
https://www.tweaktown.com/news/57648/qualcomm-snapdragon-845-10nm-lpddr4x-due-q1-2018/index.html

So hopefully by the end of 2018 the prices would have dropped enough due to competition (although bigger companies might ask for such large volumes that it could stay unavailable for a year or two)

FYI: Iphone 8's SoC is also 10mn tech.

But 10nm is old news... 7nm is where it is at. Though the prices are too high and nobody wants them at that pricepoint, says Intel.
 
Yes, new size reduced parts are always more expensive to make, because yields tend to be low until they can figure out how to do the lithography better. Presumably we're well into the scope for quantum tunneling and things to mess everything up, but I haven't heard about that for a few generations so I don't know what they've had to do to get around it, although that would be initial R&D cost and be peanuts compared to that lost through duff printings. In the long run they're cheaper because they can print more parts on the same wafer, although I guess that's more a benefit for big chips like x86 chips.

I'll also mention that 750mW/core stat they're using for the graphs in the second diagram looks kind of high to me. A phone/tablet/pyra type thing won't be able to sustain that across quad or even dual cores for all that long I'd have thought.
 
The A15 was in SOCs used by products like the Shield Portable (Tegra 4) etc in mid 2013. Even the Omap5 is a mid 2013 release. So it goes back a year earlier than that poster states
 
The A15 was in SOCs used by products like the Shield Portable (Tegra 4) etc in mid 2013. Even the Omap5 is a mid 2013 release. So it goes back a year earlier than that poster states

There's even the Samsung Chromebook featuring dual-A15 Exynos 5250, which I picked up when it came out in October 2012.
 
There's even the Samsung Chromebook featuring dual-A15 Exynos 5250, which I picked up when it came out in October 2012.

Yeah I was just reading the ARM wikipedia page, it said the following:

"Produced In production late 2011, to market late 2012"

It's pretty old when you think about it.
 
Last edited:
Is this the moment when someone has to point out that the utility and value-for-money of a device is (very nearly) independent of competing devices and their prices?

Edit: This is my 6,666th post. Hooray!
 
Last edited:
Back
Top