Intel Tock-Ticks Back to 22nm

[PaulAlcorn]

Intel's H310C chipset is fabbed on the 22nm process, meaning the chip-making giant has taken a step back to an older process as it struggles with a 14nm shortage.

Intel Tock-Ticks Back to 22nm : Read more

 

[takeshi7]

Is this the first ever instance of redesigning a chip to work on a larger process node?! It's the only example I'm aware of.

 

[slybootz]

Take that, Moore's Law

 

[InvalidError]

Is this the first ever instance of redesigning a chip to work on a larger process node?! It's the only example I'm aware of.

I bet this is far more common than you might think, we just don't hear about it as much because most of it is happening on chips nobody other than affected product engineers are tracking. For example, few to no people know or care about what chips are being used in their SmartTVs, SmartFridge, wireless speakers, etc.

 

[slybootz]

I just now re-read the title of the article. Good one

 

[wifiburger]

And.... nobody cares .

 

[pct8750]

Z370, H310C?200 Series Chipset.
Please compare H310C and Z370 pics with 200 Series Chipset pics.
Those all look the same to me!

USB 3.0 xHCI Host Controller on 200 Series Chipset, Z370, and probably H310C
DeviceID:A2AF
There are USB drivers for Windows 7.

USB 3.0 xHCI Host Controller on H370, B360, H310
DeviceID:A36D
There are no USB drivers for Windows 7.

Intel has not released Windows 7 USB 3.0 drivers for DeviceID:A36D.

 

[Paskek]

Well the 22nm works fine it just takes about 30% more power draw compare to the 14nm chipsets.
Since Intel currently lives and die by their 14nm++ it's logical to sacrifice a little on the efficiency for Intel's customers that got the older gen process boards in order for Intel to pump out more of their higher margin products like processors so Intel could make more money.

 

[tuetano]

H310C, which measures 10 x 7mm. Ok, here you are: 10 & 7 nm. Success.

 

[DavidC1]

Intel also needs to ship 200 million 14nm LTE modems to Apple. That is likely a big reason for supply constraints.

 

[stdragon]

Well the 22nm works fine it just takes about 30% more power draw compare to the 14nm chipsets.
Since Intel currently lives and die by their 14nm++ it's logical to sacrifice a little on the efficiency for Intel's customers that got the older gen process boards in order for Intel to pump out more of their higher margin products like processors so Intel could make more money.

So long as it's used for desktop and not laptops, I guess it will have to suffice for now.

 

[King_V]

Take that, Moore's Law

ROFL!! :lol:

 

[brooks2]

Intel had expected to have their 10nm lines producing at high volume by Q4 2017. This would have shifted processor production away from 14nm. Now the 14nm lines are still producing Intel's high end processors. Pushing the 14nm lines to produce faster processors means using more multi-patterning steps (SAMP). This increases process time and increases the utilization of 14nm lithography and etch tools. This reduces the availability of the 14nm lines for other processes.

This does not only effect the 14nm lines. By shifting 14nm designs to production on their 22nm lines for H130 chipsets, Intel will be need to push the 22nm lines by introducing more multi-patterning steps. This reduces the production capacity of their 22nm lines for other products.

 

[takeshi7]

Is this the first ever instance of redesigning a chip to work on a larger process node?! It's the only example I'm aware of.

I bet this is far more common than you might think, we just don't hear about it as much because most of it is happening on chips nobody other than affected product engineers are tracking. For example, few to no people know or care about what chips are being used in their SmartTVs, SmartFridge, wireless speakers, etc.

I highly doubt that. Even in SmartTVs and SmartFridges, they don't take an existing chip and redesign it to be built on a larger piece of silicon because that costs more money.

 

[smartcom5]

Thing is, since they have to, it's that simple. They were forced to do so, by law. That's the result of the California Energy Commission's 2019 regulations. So they literally have to move those to 14nm – since their chipsets ain't energy efficient enough when being fabbed on 22nm.


Speaking about happily shooting your on foot;
To my knowledge, that very law which demands computer peripheral components being energy-efficiency, Intel even happily was lobbying for and was actively involved in shaping such regulations. That why they proudly announced it joyfully in the first place. So their own actions are now coming back to bite their ass.


Did I mention that I love boomerangs? Their ways are so friggin‘ predictable! xD


We're writing the Year two thousand and eightteen as we witness when Intel shots their own face …

 

[ryan.gebauer]

I'm going to wait for Intel to release 32nm before I finish my rig

 

[stdragon]

Thing is, since they have to, it's that simple. They were forced to do so, by law. That's the result of the California Energy Commission's 2019 regulations.

Unless there's existing legislation that supersedes state via the U.S. Department of Energy (DOE), this can always go back to congress to craft new law that makes Cali requirements moot.

 

[Ali_200]

There is one major aspect for transitioning from 22nm to 14nm which is overlooked here. These are CHIPSETs, so they must interface with several fixed standards for I/O. The I/O specs must drive huge voltages (1.8V - 3.3V), while 14nm transistors are much weaker and can't drive the same voltages/currents reliably. Hence ALL I/O circuits must be re-designed, more cascoded transistors to meet high voltages, and more transistors in parallel (or longer gates) must be used to maintain standard I/O drives, slew rates, controlled/tunalbe impedances, etc. So if you shift PCH into 14 nm, ALL I/O interfaces (HDMI, eDP, SATA, PCIe, USB 3, DDR4, LPC, I2C, SMBus, etc. etc.) must be re-qualified and re-tested to meet external standards. This is a no easy feat task. Now since Intel had its engineering workforce "diversified with inclusivity" and all seasoned engineers were restructured with H2 juniors or juniors in far-far away places, with mostly C++ background, the resut should be expected.

 

[King_V]

I'm going to wait for Intel to release 32nm before I finish my rig

Pfft, whatever, dude - I'm holding out for 1µm . . . go big or go home! :lol:

 

[Tim Gueguen]

I'm going to wait for Intel to release 32nm before I finish my rig

Pfft, whatever, dude - I'm holding out for 1µm . . . go big or go home! :lol:

The truly hip users will be using germanium transistors combined with vacuum tubes before long.

 

[InvalidError]

I highly doubt that. Even in SmartTVs and SmartFridges, they don't take an existing chip and redesign it to be built on a larger piece of silicon because that costs more money.

Most chips used in SmartThings aren't designed by the company making the appliances and aren't designed for any specific appliance, they are made by chip companies marketing their products to appliance manufacturers. If your chip is much more expensive to make on 14nm or cannot be made on 14nm in sufficient volume from the fab contracts you are able to secure in a timely manner due to fabs being over-booked, then you are better off re-spinning your non-critical chips on 22nm to free up 14nm wafer starts for chips that must be on 14nm than losing millions of unit sales altogether when your clients are forced to source competitors' chips and re-design their products accordingly.

Also, when you go from one fab to another, you usually have to re-spin the whole chip due to each fab having its own primitive libraries and hard-macros with different electrical characteristics and footprints. So, even if you stay at 14nm, you still have to bear the brunt of the re-layout, re-qualification, re-tooling, etc. costs when outsourcing chip fab to multiple companies.

 

[ryan.gebauer]

I'm going to wait for Intel to release 32nm before I finish my rig

Pfft, whatever, dude - I'm holding out for 1µm . . . go big or go home! :lol:

The truly hip users will be using germanium transistors combined with vacuum tubes before long.

:lol:

 

[jasonelmore]

@Brooks They don't need more patterns to scale up. They simply need more silicone. The will get less yields because the die has been expanded to fit all of the features.