Intel's 10nm Problems Might Be Worse Than First Thought

[Stefan Payne]

1 hour ago, pipnina said:

And when Apple last changed architecture, didn't they make a program a bit like WINE that could run the PowerPC programs on the Intel macs?

No, they didn't.

They forced the App Developers to include both the PowerPC as well as the x86 Binary with the Programm.

 

I had at the time a PowerMAC G4/933 and that one is still here somewhere, though I don't have a Drive in it, thus not working because no OS. And 10.4 is the last one Supporting PowerPC AFAIR.

[Hunter259]

6 minutes ago, Stefan Payne said:

No, they didn't.

They forced the App Developers to include both the PowerPC as well as the x86 Binary with the Programm.

 

I had at the time a PowerMAC G4/933 and that one is still here somewhere, though I don't have a Drive in it, thus not working because no OS. And 10.4 is the last one Supporting PowerPC AFAIR.

That's not true. Apps could run under rosetta if it was a non-universal binary on a Intel machine. The first intel Macs were about the same as the fastest Power PC macs running Power PC software.

[Stefan Payne]

1 minute ago, Hunter259 said:

That's not true. Apps could run under rosetta if it was a non-universal binary on a Intel machine. The first intel Macs were about the same as the fastest Power PC macs running Power PC software.

Oh, I think you're right and I totally forgot that.

Now that you mention the name, I kinda Remember that there was something.

[rawrdaysgoby]

Intel fan boys have been saying intel will soon come up with 10nm that is denser then tsmc 7nm soon. And then they back track Completely blowing away everything they ever said making regular 10nm that everyone else has been on. 

[Stefan Payne]

...and then their 10nm processors are nowhere to be seen because of Problems, while the other guys all had their Tapeouts...

 

You know, sending out the Tapes with the Prcessor Design....

[GOTSpectrum]

7 hours ago, pipnina said:

I wonder if we will find 10 or 8 nanometers to be the practical limit instead of the theoretical 5. That could be problematic...

Haven't heard anything exciting about graphene chips since 2013, and quantum computers are still very much in the state of normal computers in the 60s (i.e. really bad and they take up a whole room).

 

If they can't get past 10nm, maybe we'll get another 30-50% performance over a decade or two from IPC improvements or larger chips... CPU speed improvements have certainly been slowing over the last two decades.

AFAIK IBM has managed to make a 5nm chip already. The current belief is that 3nm is the theoretical limit now, if you can somehow overcome quantum effects. There are already a number of players started to test 3nm production technologies. 

[mr moose]

48 minutes ago, Ben Quigley said:

AFAIK IBM has managed to make a 5nm chip already. The current belief is that 3nm is the theoretical limit now, if you can somehow overcome quantum effects. There are already a number of players started to test 3nm production technologies. 

yep working with Samsung.

 

https://arstechnica.com/gadgets/2017/06/ibm-5nm-chip/

 

 

 

 

[Taf the Ghost]

1 hour ago, Ben Quigley said:

AFAIK IBM has managed to make a 5nm chip already. The current belief is that 3nm is the theoretical limit now, if you can somehow overcome quantum effects. There are already a number of players started to test 3nm production technologies. 

There's room for an even higher gate height, then there's expected to be several more generations of improved transistors. The current Foundry roadmaps are still full for the next decade, though we can expect Nodes to take longer to move between. The future "Foundry 1nm" node is where things currently seem to be at the speculative research level right now.

[pipnina]

1 hour ago, Taf the Ghost said:

There's room for an even higher gate height, then there's expected to be several more generations of improved transistors. The current Foundry roadmaps are still full for the next decade, though we can expect Nodes to take longer to move between. The future "Foundry 1nm" node is where things currently seem to be at the speculative research level right now.

I find it incredible that since the Pentium 4 15 years ago we have shrunk transistors by almost a factor of 10, and are likely to do so again.

At that low end of the nanometer scale, there's not many things left smaller than the gap. Hell, the 10nm chips still have transistors smaller than the wavelength of light at the border between high-energy ultraviolet and low-power x-ray. If they make it to 1nm it'll be very hard x-ray territory.

IIRC these chips are made by bombarding the silicon wafers with light... In order to get 1nm precision I dread to think of how high energy the photons would need to be.

[asus killer]

I'm not happy with this. We can easily go from a place where there was only one relevant CPU maker to a place where... nothing changes really, only the name of the company in control.

AMD will be no better than Intel if they corner the market.

[pipnina]

11 minutes ago, asus killer said:

I'm not happy with this. We can easily go from a place where there was only one relevant CPU maker to a place where... nothing changes really, only the name of the company in control.

AMD will be no better than Intel if they corner the market.

This is very true, but how on earth is anyone going to be able to compete with two companies that have spent billions a year on R&D for 40 years to get to the top of the CPU ladder, as well as also owning the sole rights to the instruction set that makes up 94% of the consumer market?

 

ARM manufacturers are far worse than AMD and Intel as well, AMD and Intel offer free software drivers that can be redistributed and modified, while Qualcomm gives "2 years of support" for their chip's drivers and the only drivers they offer or allow to be run on those chips are proprietary. They are the reason why a lot of mobile phones can't have kernels made in the last 4 years. Phones upgrading to Oreo don't even need to have a kernel newer than 3.18 (Which came out in 2014) and android only requires that new Oreo phones use a minimum of kernel 4.4. At least 4.4 is still supported, though it did come out in January 2016 so it is already almost half way to EOL.

 

Computing is a shady business.

[schwellmo92]

10 hours ago, Stefan Payne said:

Seen that thing? Look at the Writings on the Die, ignore everything else.

https://www.techpowerup.com/gpudb/2977/xbox-one-x-gpu

https://www.techpowerup.com/gpudb/2866/xbox-one-s-gpu

https://www.techpowerup.com/gpudb/2086/xbox-one-gpu

 

 

OR that?

https://www.techpowerup.com/gpudb/2876/playstation-4-pro-gpu

https://www.techpowerup.com/gpudb/2085/playstation-4-gpu

 

So why shouldn't there be a contract where AMD designes the GPU, holds the licenses and everything else is up to Apple??

Also the fact that AMD don’t do the fabrication..

[Sakkura]

1 hour ago, pipnina said:

I find it incredible that since the Pentium 4 15 years ago we have shrunk transistors by almost a factor of 10, and are likely to do so again.

At that low end of the nanometer scale, there's not many things left smaller than the gap. Hell, the 10nm chips still have transistors smaller than the wavelength of light at the border between high-energy ultraviolet and low-power x-ray. If they make it to 1nm it'll be very hard x-ray territory.

IIRC these chips are made by bombarding the silicon wafers with light... In order to get 1nm precision I dread to think of how high energy the photons would need to be.

The Van der Waals radius of a silicon atom is 0.2nm (so the diameter is 0.4nm), so we're getting into the range of a couple dozen atoms. Pretty wild that we are mass manufacturing that kind of stuff - it's one thing to do it in a lab, but in a big manufacturing plant making millions of chips...

 

We have techniques to etch features that are significantly smaller than the wavelength of the light being used, so we don't have to get quite as drastic about the wavelengths... but it's still getting into the very high-energy UV radiation.

[schwellmo92]

12 hours ago, abazigal said:

I am talking about custom ARM processors replacing Intel chips altogether. 

 

There might be some switching pains, but if anyone knows how to manage such a transition, it's Apple. I would view it as short term pain for long term gain. With Intel increasingly falling behind when it comes to updating their processors, it doesn't make sense for Apple to continue to rely on a company who has demonstrated that they simply can't deliver. 

 

And it's not as though their Mac line really has a lot to lose at this point. 

ARM would not be good for high power computing due to the inherit pitfalls of RISC (Reduced Instruction Ser Computing).

[Stefan Payne]

1 hour ago, pipnina said:

I find it incredible that since the Pentium 4 15 years ago we have shrunk transistors by almost a factor of 10, and are likely to do so again.

Yes, but the clockspeed did not increase that much

 

The 15years before that, from 1988 The Clockspeed increased almost a hundred(twenty(eight)) fold. From just 25MHz to 3GHz.

 

As for manufactoring I think the step was a bit bigger as well...

Its from 1,5µ to 130nm in that timeframe. (0,13µ AFAIR)

Right now we have 130nm to 12/14nm...

 

Sources:

http://processortimeline.info/proc1980.htm

http://processortimeline.info/proc2003.htm

 

In 1988, the 25MHz 80386 CPU was just introduced....

 

 

The bigger Problem is the breaking Voltage of Silicon because that is around 0,7V.

That means that under that voltage Silcion does not conduct.

And that was one of the Reason for the development of Germanium semiconductors as the breaking voltage is less than half of that.

 

That mean that you can run your semiconductor at around ,5V or so...

 

 

Notice that the VCore of the CPU didn't rally change in years??

You can say that its almost 10 years and we only lowered the Powre Consumption by like ,1 to ,2V.

[Hellion]

How about intel concern themselves with fixing the countless variants of security flaws in their products first?

 

It makes no difference to shrink the node size when the product itself essentially takes a 15%+ performance hit after bios and OS patches are implemented to fix their engineering neglegence because they are too busy playing the fat cat role pushing marginal 5%~ performance increases each generation.

 

What a joke intel has become. 

 

 

For the record I have 3 machines with Intel CPU's,  so in before the "fanboy" accusations start to fly.

[Clanscorpia]

20 hours ago, Deus Voltage said:

Is it possible to do both at the same time? Manufacturing chips at your own fabs and using "outsider" fabs as a temporary thing? 

No, because the process each uses to make chips is entirely different. Theyd basically be 2 completely different CPUs

[Zodiark1593]

30 minutes ago, Hellion said:

How about intel concern themselves with fixing the countless variants of security flaws in their products first?

 

It makes no difference to shink the node size when the product itself essentially takes a 15%+ performance hit after bios and OS patches are implemented to fix their engineering neglegence because they are too busy playing the fat cat role pushing marginal 5%~ performance increases each generation.

 

What a joke intel has become. 

 

 

For the record I have 3 machines with Intel CPU's,  so in before the "fanboy" accusations start to fly.

Intel aside, Arm's Cortex A75 and (heavily implied by Apple themselves) Apple's own CPU cores both exhibit the "Meltdown" security flaw. However, Cortex A75 had not yet been released at the time the flaws were made public, and Apple can easily patch their iOS devices with a minimum of attention.

[Okjoek]

Intel had the last half a decade or more to figure this out.

[cj09beira]

1 hour ago, Hellion said:

How about intel concern themselves with fixing the countless variants of security flaws in their products first?

 

It makes no difference to shrink the node size when the product itself essentially takes a 15%+ performance hit after bios and OS patches are implemented to fix their engineering neglegence because they are too busy playing the fat cat role pushing marginal 5%~ performance increases each generation.

 

What a joke intel has become. 

 

 

For the record I have 3 machines with Intel CPU's,  so in before the "fanboy" accusations start to fly.

the people doing those things aren't the same ones so don't worry they are working on it, its hard to fix these kinds of problems 

[Taf the Ghost]

10 hours ago, pipnina said:

I find it incredible that since the Pentium 4 15 years ago we have shrunk transistors by almost a factor of 10, and are likely to do so again.

At that low end of the nanometer scale, there's not many things left smaller than the gap. Hell, the 10nm chips still have transistors smaller than the wavelength of light at the border between high-energy ultraviolet and low-power x-ray. If they make it to 1nm it'll be very hard x-ray territory.

IIRC these chips are made by bombarding the silicon wafers with light... In order to get 1nm precision I dread to think of how high energy the photons would need to be.

Well, the move from Planar (design-wise 2D) to Fin-based transistors actually improved the density a lot more than people realize, because it moved things into a vertical space. As a result, nothing in the 14nm or 10nm generations are actually at that spacing. Thus, there is room to shrink within a 3D space on a 2D design approach. We haven't even gotten to interesting approaches to 3D Stacking.

 

It's pretty impressive, but it's also a century of research and trillions in R&D.

[Taf the Ghost]

2 hours ago, Okjoek said:

Intel had the last half a decade or more to figure this out.

Intel should have made the decision to step back from the issues with 10nm in 2015. That's why they fired their previous CEO.

 

Krzanich hired in 2013.

 

$300 Million for Diversity efforts, announced in January 2015.

 

Which is about the exact moment that decisions around 10nm were botched. While these types of mistakes are not exclusive to Diversity pushes, when a CEO is making big investments/statements that do nothing to help the company function better you always end up with these results. If Intel wasn't so big & profitable, this would be a far, far worse problem. (If 14nm wasn't such an amazing process & AMD wasn't in the midst of the Dozer Disaster, Intel would have big, big problems right now.)

[cj09beira]

2 minutes ago, Taf the Ghost said:

Intel should have made the decision to step back from the issues with 10nm in 2015. That's why they fired their previous CEO.

 

Krzanich hired in 2013.

 

$300 Million for Diversity efforts, announced in January 2015.

 

Which is about the exact moment that decisions around 10nm were botched. While these types of mistakes are not exclusive to Diversity pushes, when a CEO is making big investments/statements that do nothing to help the company function better you always end up with these results. If Intel wasn't so big & profitable, this would be a far, far worse problem. (If 14nm wasn't such an amazing process & AMD wasn't in the midst of the Dozer Disaster, Intel would have big, big problems right now.)

should i mark this as informative or funny

hm.... decisions decisions 

[Stefan Payne]

The Diversity Thing might have been the Problem because without it you can have a sheet of paper without a name, sex/gender (the Biological Thing), without a photo and so on.

So you don't know how this person looks, if its a Man or a Woman, Long hair or short, Blond or Grey. Nothing about the Apereance.

 

All you know of this Person is how hard he/she works and what he/she accomplished, nothing more.

 

The Problem is that this isn't good for diversity for whatever reason.

 

So if you implement the Diversity Thing, you prefer a person who is not the best choice for the Job because of to what Parents this person is born (ie Race), if its a male or female...

 

That also isn't positive for the Company climate because nobody really respects these people because nobody knows if he/she got the position because he/she earned it or because of other factors. And if you think that this sounds pretty racist and/or sexist, you are pretty darn right! 

THAT is Racism and Sexism at work, basically...

 

 

And also big changes in Company personel means that it throws back the Company!

Especially in higher managment positions or in development. If you do it right, it throws the Company back years!

Because the new Guy has to learn what whoever came before had done, what is actually going on. So you can expect 3-6 Months without good leadership for that reson...

 

Well, anyway in the End these Decisions might be good for the consumer, as Intel looses Market Share and we can actually hope for a somewhat equal marketshare and the choice between two good products for real good prices...

The Optimum would be if the x86 Patents would be sold to ARM (or another Licensing company  that makes money with the Licenses), so that there is a possibility that for example Apple and/or Samsung can join the x86 Market so that we have 3 or 4 competitiors like we did in the early to mid ninetys...

I kinda miss Cyrix. They always had some interesting, innovative Products...

[Taf the Ghost]

20 minutes ago, cj09beira said:

should i mark this as informative or funny

hm.... decisions decisions 

Informative. The classic move is to look to Short any stock when a CEO either builds a big house or they announce something for Optics reasons. For a bit over a decade, it's been the Diversity Initiatives. In decades past, it was other things, normally a big, new HQ. Especially if the CEO spends 100s of 1000s on his office. The 90s had a bout when a company went big on Gay Rights, but so much was just straight up from 1995 to 2000 that it didn't matter much.

 

It's quite straight forward why it happens. The company is sitting on a lot of money and the CEO is looking to impress his connections. In this case, it's fluffing his own ego at the cost of the company. Big expenditures of money that do nothing for the core functions of a company always do this. There's never "free" money in a company, regardless of how much money they are making. Fortune 500 Executive level is something of a really tight club, so a lot of decisions get made to show off to those also in the club. It's also the reason for the Golden Parachutes and why incompetents keep getting rehired.