By: Purana Archer (ancientarcher.delete@this.gmail.com), August 18, 2014 2:32 am
Room: Moderated Discussions
David Kanter (dkanter.delete@this.realworldtech.com) on August 17, 2014 6:28 pm wrote:
> > > > History is not an strong argument. Despite flaweless execution of last years., Intel is now having delay
> > > > problems with Broadwell and problems with TSX.
> > >
> > > History is an incredibly strong guide to the future. I've been
> > > hearing "We will catch up with Intel on manufacturing"
> > > for 15 years from various competitors. It was bullshit then,
> > > and I have yet to hear why it isn't bullshit now.
> > > Why is TSMC any more competent today than they were 5 years
> > > ago? Do you think Intel is less competent now than
> > > they were before? Or do you think that their latest problems will impact the whole industry?
> > >
> > > My guess is that Intel ran into some nasty issues, and TSMC will hit them sooner or later.
> >
> > I already said that history is not a strong argument. History can not be used to
> > make infallible predictions, otherwise historians would be all rich people.
>
> Why don't you explain why you think this time is different instead? What
> reasons do you have to believe that TSMC is better than before?
Why is TSMC better than before? In one word - money. TSMC just has vastly more money to spend that before. Can money buy expertise? you bet!
5 years ago (2008)
Intel's capex = $5.2bn, Intel R&D = $5.7bn
TSMC's capex = $1.9bn, TSMC R&D = $0.7bn
2013
Intel's capex = $10.7bn, Intel R&D = $10.6bn
TSMC's capex = $9.6bn, TSMC R&D = $1.6bn
Has the gap shrunk??
The WHY is easier. It's because Intel didn't or couldn't choke off the flow of money into its competitors the way it did with AMD and the RISC processor companies before that. And why was that - one word - smartphones! Intel missed the wave and that not only hamstrung it but also enabled its competitors.
Before you start crowing about how Intel spends so much more in R&D, just a couple of things:
1) Intel's R&D is on the entire stack of design and manufacture. TSMC's R&D is just on manufacturing. The other parts of process are taken up by other parties, viz ARM, Qcom/Nvidia/Apple, Cadence/Synopsis, ...
2) The biggest cost of R&D is cost of manpower. TSMC's engineers in Taiwan are far cheaper than those employed by Intel in the land of the free and home of the brave (or is the other way around, I forget)
Nonetheless, in the last 5 years, you have seen
1) TSMC's capex has gone from 1/3 of Intel to nearly equal
2) TSMC's R&D has increased 2.5x while Intel's has gone up only 1.8x
So, compared to where TSMC was 5 years ago vs Intel, does it have a few billion reasons be be better. Hell yeah!!
>
> > > Also, TSX is irrelevant to manufacturing, stop making deceptive arguments.
> >
> > Don't misinterpret me. My mention of TSX was to bring one example of how recent history >couldn't be used
> > to predict the problems that Intel has now with TSX. I never said that was relevant to manufacturing.
>
> OK, so history is imperfect - I agree. I still don't hear you explaining why you think that TSMC has
> somehow improved, or Intel has gotten worse. In the absence of a compelling argument you simply aren't
> going to convince me (and I doubt you'll convince anyone else) that the process technology gap between
> Intel and the rest of the world has narrowed. If anything it has widened over time...
>
> David
> > > > History is not an strong argument. Despite flaweless execution of last years., Intel is now having delay
> > > > problems with Broadwell and problems with TSX.
> > >
> > > History is an incredibly strong guide to the future. I've been
> > > hearing "We will catch up with Intel on manufacturing"
> > > for 15 years from various competitors. It was bullshit then,
> > > and I have yet to hear why it isn't bullshit now.
> > > Why is TSMC any more competent today than they were 5 years
> > > ago? Do you think Intel is less competent now than
> > > they were before? Or do you think that their latest problems will impact the whole industry?
> > >
> > > My guess is that Intel ran into some nasty issues, and TSMC will hit them sooner or later.
> >
> > I already said that history is not a strong argument. History can not be used to
> > make infallible predictions, otherwise historians would be all rich people.
>
> Why don't you explain why you think this time is different instead? What
> reasons do you have to believe that TSMC is better than before?
Why is TSMC better than before? In one word - money. TSMC just has vastly more money to spend that before. Can money buy expertise? you bet!
5 years ago (2008)
Intel's capex = $5.2bn, Intel R&D = $5.7bn
TSMC's capex = $1.9bn, TSMC R&D = $0.7bn
2013
Intel's capex = $10.7bn, Intel R&D = $10.6bn
TSMC's capex = $9.6bn, TSMC R&D = $1.6bn
Has the gap shrunk??
The WHY is easier. It's because Intel didn't or couldn't choke off the flow of money into its competitors the way it did with AMD and the RISC processor companies before that. And why was that - one word - smartphones! Intel missed the wave and that not only hamstrung it but also enabled its competitors.
Before you start crowing about how Intel spends so much more in R&D, just a couple of things:
1) Intel's R&D is on the entire stack of design and manufacture. TSMC's R&D is just on manufacturing. The other parts of process are taken up by other parties, viz ARM, Qcom/Nvidia/Apple, Cadence/Synopsis, ...
2) The biggest cost of R&D is cost of manpower. TSMC's engineers in Taiwan are far cheaper than those employed by Intel in the land of the free and home of the brave (or is the other way around, I forget)
Nonetheless, in the last 5 years, you have seen
1) TSMC's capex has gone from 1/3 of Intel to nearly equal
2) TSMC's R&D has increased 2.5x while Intel's has gone up only 1.8x
So, compared to where TSMC was 5 years ago vs Intel, does it have a few billion reasons be be better. Hell yeah!!
>
> > > Also, TSX is irrelevant to manufacturing, stop making deceptive arguments.
> >
> > Don't misinterpret me. My mention of TSX was to bring one example of how recent history >couldn't be used
> > to predict the problems that Intel has now with TSX. I never said that was relevant to manufacturing.
>
> OK, so history is imperfect - I agree. I still don't hear you explaining why you think that TSMC has
> somehow improved, or Intel has gotten worse. In the absence of a compelling argument you simply aren't
> going to convince me (and I doubt you'll convince anyone else) that the process technology gap between
> Intel and the rest of the world has narrowed. If anything it has widened over time...
>
> David