Of course, they're claiming existing ones get 4.6 days. Yeah, right. Still, if they can produce one that does 5 days instead of 1, that's a really decent improvement. If they *can* make a decent smartwatch-level CPU with this tech.
POWER (9) is manufactured on a 14HP process, which is exclusively used by IBM. This is a 14nm FinFET SOI process. 22FDX is something completely different and has nothing do to with neither power nor ibm.
Apparently, it is not ideal for large and high performance chips (quoting from there): "When I say that, I should qualify: you are not going to get that from a big chip. If you are making a big chip, you got tons of wire capacitance, so you are not going to beat a FinFET design. FinFET has huge drive current, so if you're dealing with smaller chips, wiring capacitance isn't as a big, but gate capacitance of the transistor is more important. FD-SOI has less gate capacitance than FinFET because you don't have the gate wrapping around the whole device. So it is ideally suited for smaller chips, I say less than 150 square millimeters."
I'd say another drawback is actually availability timeframe: the diagrams there show 22FDX to be "on the same level" as 14nm FinFET (depending on your needs), but from a timing perspective it looks to me like it's rather a "competitor" to 7nm FinFET.
Granted that it will be much better than 40 and 28nm processes, but I didn't get how it has better battery life than FinFET (which I'm assuming to be 16nm)...
FinFET from GloFo is 14nm and 12nm, TSMC has 16nm. But these all mean little on their own. Current FinFET processes are geared for high performance. This is geared for low power consumption at adquate performance. Hence why a low power device functions better on the corresponding low power node. I wouldn't want a GPU or CPU made on this process, I doubt it would clock as well or as efficiently.
For some reason the first use-case that came to mind for me was the Raspberry Pi. I believe its SoC has been stuck at 40nm for a long time now, so I wonder if this would make a nice process for Broadcom if they choose to skip 28nm (assuming 22FDX isn't a bank breaker for this market segment).
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
12 Comments
Back to Article
Rocket321 - Tuesday, July 10, 2018 - link
14 - 20 day battery life for a smart watch? FINALLY it will be possible for companies to make smartwatches worth buying.gfkBill - Tuesday, July 10, 2018 - link
Of course, they're claiming existing ones get 4.6 days. Yeah, right. Still, if they can produce one that does 5 days instead of 1, that's a really decent improvement. If they *can* make a decent smartwatch-level CPU with this tech.close - Sunday, July 15, 2018 - link
So you can expect 2-3 days of battery based on their assessment of "today vs. future". And that's if everything goes well.DefeatedGoat - Tuesday, July 10, 2018 - link
https://i.imgur.com/3QY2gM7.pngMore numbers, you are welcome.
ajp_anton - Wednesday, July 11, 2018 - link
mW/day? There goes its credibility.rocketbuddha - Tuesday, July 10, 2018 - link
Wasn't 22FDX something that came from GF buying the fabs from IBM?Why wasn't POWER mentioned at all?
mczak - Tuesday, July 10, 2018 - link
POWER (9) is manufactured on a 14HP process, which is exclusively used by IBM. This is a 14nm FinFET SOI process.22FDX is something completely different and has nothing do to with neither power nor ibm.
iwod - Tuesday, July 10, 2018 - link
The 22FDX on paper is way too good and blows everything away, unless you absolutely need leading node.Like everything there has to be some trade off, anyone has any pointers?
mczak - Tuesday, July 10, 2018 - link
Anandtech did some interview with GF CTO a while ago:https://www.anandtech.com/show/12438/the-future-of...
Apparently, it is not ideal for large and high performance chips (quoting from there): "When I say that, I should qualify: you are not going to get that from a big chip. If you are making a big chip, you got tons of wire capacitance, so you are not going to beat a FinFET design. FinFET has huge drive current, so if you're dealing with smaller chips, wiring capacitance isn't as a big, but gate capacitance of the transistor is more important. FD-SOI has less gate capacitance than FinFET because you don't have the gate wrapping around the whole device. So it is ideally suited for smaller chips, I say less than 150 square millimeters."
I'd say another drawback is actually availability timeframe: the diagrams there show 22FDX to be "on the same level" as 14nm FinFET (depending on your needs), but from a timing perspective it looks to me like it's rather a "competitor" to 7nm FinFET.
ads295 - Tuesday, July 10, 2018 - link
Granted that it will be much better than 40 and 28nm processes, but I didn't get how it has better battery life than FinFET (which I'm assuming to be 16nm)...Death666Angel - Tuesday, July 10, 2018 - link
FinFET from GloFo is 14nm and 12nm, TSMC has 16nm. But these all mean little on their own. Current FinFET processes are geared for high performance. This is geared for low power consumption at adquate performance. Hence why a low power device functions better on the corresponding low power node. I wouldn't want a GPU or CPU made on this process, I doubt it would clock as well or as efficiently.wizfactor - Tuesday, July 10, 2018 - link
For some reason the first use-case that came to mind for me was the Raspberry Pi. I believe its SoC has been stuck at 40nm for a long time now, so I wonder if this would make a nice process for Broadcom if they choose to skip 28nm (assuming 22FDX isn't a bank breaker for this market segment).