I'm wondering how much of their roadmap is affected with all those Spectre revelations. There must be a massive re-engineering effort going on ? Speculation mechanisms were mostly implemented in silicon, right ?
> Thanks for your email. We are aware of issues with our HTTPS and currently trying to upgrade our systems and the security certificates. Unfortunately, it will take us a while until all the testing is done, but hope to migrate to new environment by the end of this year.
Wasn't it just a year ago when people were making big claims about Intel using its "manufacturing edge" to compete against ARM and whatnot, and about them making other companies' chips and stealing customers from TSMC and the others?
Outsource Chipset production to TSMC. Not Chip / CPU, I would be surprised if they do. I wonder if this is planned or not, because TSMC just doesn't have spare 12nm capacity sitting there, at least not in the quantity Intel requires. They tried moving those chipset back to their own 22nm and it is obviously not working out for them well.
If not, it is another sign of Intel management failure.
TSMC doesn't have a 14nm process, it's "16nm" process is 20nm, and considerably inferior to intel's 14nm same goes for the 12nm which is a BEOL improvement of 16nm with no density improvements.
The only process that Intel might use if they outsource to TSMC is going to be the high performance 7nm node when it's out in full swing which is insanely expensive and not available just yet for mass production.
For non CPUs it doesn't really matter but I doubt PCH, network chipset and modems are even built on 14nm but there is no process that Intel can use to tap out CPUs on especially it's high core count ones while keeping the same frequencies it's used to even with 7nm.
TBH this sounds more like some BS than a valid rumor.
I’m a little surprised that they’re continuing to push shrinking transistors so hard. While it’s impressive, it introduces massive new problems to solve across the entire design and manufacturing process, every time. There are other ways to achieve their goals of reduced area, etc.
One example is to save area by making the chip do less, and an outstanding place to start would have been to banish all of this DRM crap that no customer has ever asked for. It costs complexity, no doubt adding so much crap to the chip that they need to save area by shrinking the whole chip on an accelerated schedule.
If there are other ways, you should patent it because it'll be worth trillions.
They're not shrinking it just for the fun of it, they're progressing the technologies to accommodate the explosive growth in computing every year.
The major problem with the growth is the energy consumption, we can't sustain it without reducing it.
Shrinking the process node allows them to double the transistor capacity without doubling the power requirements and they were on the path of doing it every few years, alas also the Moore's law.
Also, the mobile devices are outpacing the desktops, which means the market is demanding lighter, more power-efficient devices.
It takes several years and tens of billions of dollars to develop a new fab, so Intel has to predict and work on future technologies continuously the best ways they can; they can't risk losing their businesses to other fabs that is also doing the same thing.
DRM is the least of the problem, the biggest part of the Intel CPUs are the integrated graphics now.
I did give an example of another way: reducing the features of the chip. DRM is just one case.
I am aware of the reasons for shrinking. I’m not saying they should never pursue a new process but they’re attempting to achieve it on a timeline that isn’t realistic.
Also, power-inefficient design is something that is hard for Intel to change because too much depends on their architecture being the way it is.
So between not wanting to get rid of stupid features and not being able to sell a backward-incompatible chip with power-efficient architecture changes, they’re left trying to shrink before everybody else.
12 comments
[ 3.1 ms ] story [ 31.8 ms ] threadalternative link https://outline.com/RstMtX
https://www.ssllabs.com/ssltest/analyze.html?d=www.digitimes...
> Thanks for your email. We are aware of issues with our HTTPS and currently trying to upgrade our systems and the security certificates. Unfortunately, it will take us a while until all the testing is done, but hope to migrate to new environment by the end of this year.
If not, it is another sign of Intel management failure.
The only process that Intel might use if they outsource to TSMC is going to be the high performance 7nm node when it's out in full swing which is insanely expensive and not available just yet for mass production.
For non CPUs it doesn't really matter but I doubt PCH, network chipset and modems are even built on 14nm but there is no process that Intel can use to tap out CPUs on especially it's high core count ones while keeping the same frequencies it's used to even with 7nm.
TBH this sounds more like some BS than a valid rumor.
One example is to save area by making the chip do less, and an outstanding place to start would have been to banish all of this DRM crap that no customer has ever asked for. It costs complexity, no doubt adding so much crap to the chip that they need to save area by shrinking the whole chip on an accelerated schedule.
They're not shrinking it just for the fun of it, they're progressing the technologies to accommodate the explosive growth in computing every year.
The major problem with the growth is the energy consumption, we can't sustain it without reducing it.
Shrinking the process node allows them to double the transistor capacity without doubling the power requirements and they were on the path of doing it every few years, alas also the Moore's law.
Also, the mobile devices are outpacing the desktops, which means the market is demanding lighter, more power-efficient devices.
It takes several years and tens of billions of dollars to develop a new fab, so Intel has to predict and work on future technologies continuously the best ways they can; they can't risk losing their businesses to other fabs that is also doing the same thing.
DRM is the least of the problem, the biggest part of the Intel CPUs are the integrated graphics now.
I am aware of the reasons for shrinking. I’m not saying they should never pursue a new process but they’re attempting to achieve it on a timeline that isn’t realistic.
Also, power-inefficient design is something that is hard for Intel to change because too much depends on their architecture being the way it is.
So between not wanting to get rid of stupid features and not being able to sell a backward-incompatible chip with power-efficient architecture changes, they’re left trying to shrink before everybody else.