Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Everyone "knows" that at around 3nm the party ends, but they haven't quite "accepted" it yet.
The train is slowing down as it approaches the end of the tracks
but some will be exiting earlier than others due to economics.
I suspect it will be Intel that pushes things to the effective final
limit due its strong in-house material sciences, tight integrated
control over both design methodology and manufacturing, and a
focussed application target driving process development.
If Google bought a huge number of Power systems instead of Xeon it would
hurt Intel.
If Google bought a huge number of Z systems instead of Xeon it would hurt
Intel.
If Google bought a huge number of SPARC systems instead of Xeon it would
hurt Intel.
If Google bought a huge number of AMD based systems instead of Xeon it
would hurt Intel.
If Google figured out a way it could use its existing Xeon systems much
more efficiently and delays all further hardware purchases for two years
it would hurt Intel.
Lot's of hypotheticals could hurt Intel. Lot's of hypotheticals could help
Intel. Let's see what actually happens. Right now the server market is 99%
x86 (of which over 95% is Intel) and 1% everything else. This isn't a random
starting point but the result of the last 20+ years of desperate, vicious, and
highly creative competition among a variety of RISC and CISC platforms.
ARM is just another RISC, late to the party, thrown into the meat grinder
starting at zero.
There's a lot they can do, IMO. Still plenty of things on the iPhone 6s/6s Plus that are far from perfect.
I always got the impression from iPhone owners that each model
was perfect and subsequent models even more perfect than the last.
Let me guess, BK wanted to talk all about diversity achievements, business
achievements not so much.
Of course Google is "interested" in ARM servers. They wouldn't be doing
their corporate due diligence if they weren't leaking such "interest" before
negotiating every new big purchase from Intel.
Let me know when they buy 10k ARM boxes and put them into production
use. Money talks, BS walks.
Qualcomm continues full-speed into Intel's.
Let me know when any ARM incursion into servers is slightly more
effective than the Bay of Pigs invasion.
Is it possible new products have been slowed because the competition does not warrant as fast a pace anymore?
The competition is Intel processors running in customers's current boxes.
If Intel doesn't release compelling new products then upgrade cycles walk
out to the right indefinitely.
In short I want to see products and substantial previews of products on a continuing basis that support a real roadmap for all of Intel's product lines. HR and Bill Holt should represent 10% or less of the column inches if products were properly foremost.
Exactly. I wouldn't be nearly as concerned about Intel's stream of press
releases about diverse hiring if these were far outnumbered by press releases
about great new highly competitive products, processes, and share gains in
new markets.
When all you have to say are platitudes and feel good corporate self-back
patting I get very worried.
Hey Intel, focus on PC - personal computer, not PC - political correctness.
Hey BK, you are a CEO of a huge multinational manufacturing company, not
running for office in Berkeley. Get your priorities right. Outside of the twittering
SJW classes of NA and EU this fluff utterly fails to impress. Get your damn
silicon trains running on time or GTFO.
WWAD - what would Andy do? In today's Intel? Tear a lot of new ones would
be my guess.
Political correctness means excellence is no longer the goal. Mediocrity is.
I don't think mediocrity is the goal, just a discounted side effect of demoting
excellence and achievement to secondary significance.
Isn't that special
http://blogs.wsj.com/digits/2016/02/03/intel-closes-some-diversity-gaps-but-challenges-remain/?mod=yahoo_hs
Is this feel good window dressing going to sell more chips at higher ASP
and GM? Is it going to bring in 10 nm earlier and with fewer problems?
I would like to think that Intel top management spends more of its time
actually taking care of its business than value signalling PR exercises
in political correctness. You know, designing and implementing successful
long term strategies for staying ahead of foundry process technologies,
dealing with the smart phone market and trying to insulate itself and as
much of the PC market as possible from Microsoft's increasingly erratic
and anti-customer tendencies.
In my experience companies that spend much effort telling the world how
virtuous they are often do so to take attention away from how they are
losing their grip and relevance. I am going to watch Intel management
even more carefully from now on with my finger hovering over the dump
button. Smarten the f*** up BK and deliver the goods, not pictures of
rainbows and unicorns.
They may have over-estimated the number of customers sophisticated enough
to use their process technology and wanting its combination of qualities enough
to learn the Intel way of doing things.
The set of such customers may be limited to players who compete directly
with Intel.
That's kind of a huge negative for their foundry efforts.
It does discourage a wide range of customers. They can go to TSMC.
Very few fabless semis need or can readily utilize the particular mix of
pros and cons of Intel's processes. But the few that do generally have a
pretty savvy top notch bunch of engineers who can learn the Intel way
given enough time.
Another negative is the foundry tail will never wag the x86 MPU dog
when it comes to driving features and roadmap of process technologies
at Intel.
People who think a foundry business is simply a matter of accepting
GDSII files when the mood suits you are sorely mistaken. This is taken
one step further when you are talking about leading edge process tech
optimized for narrow range of products expected to be designed using
a specific restrictive style and methodology.
Turning that around, Intel products would be slower, more power hungry,
and more expensive if they were made in TSMC processes and designed
using standard methodologies and uncustomized toolchains that support
a conventional customer/foundry style relationship.
You don't think the fabless chip designers work closely with their foundry partners to do similar things?
LOL, wow, where to start.
The foundry wants its 35 or 40% GM and doesn't give a crap if you can
only get $200 for an iPhone instead of $600. The same for yield. Ever
see "Goodfellas"? They hit PCM spec then they're golden. If you have a
marginal design well f*** you pay up. You got a packaging interaction
issue well f*** you pay up. And so on.
A foundry has thousands of customers and most of them are on the n00b
end of the spectrum. The design rules and methodologies are in bright
primary colours and built with lots of cushioned foam for their protection.
At Intel the designers on the physical side of IC design are expected to
juggle chainsaws and work hand in hand with process developers to learn
how to do it properly. That is part of controlling your destiny.
Yep. It is great, till it isn't!
I remember when Intel's top bin desktop chip was priced around $1k
generation after generation. It wasn't 0.1% of sales to crazy gamers
either, businesses were buying quite a few top end systems - they
were still dirt cheap compared to VAXen and AS/400.
Intel adapted as the market evolved, lowering costs to keep gross margin
just as high if not a tad better than the good 'ol days. You can do that
when you control your destiny - less expensive packaging, design for
yield, repair and salvage, tighter integration of design methodology and
process design etc.
I wonder how Apple would do in a world of $200 iPhones? Can they just
tell Foxconn to start making them for one third the cost?
Apple sells mainly high end fashion accessories with embedded electronics.
Fashions change unpredictably and sometimes rapidly.
So why are you long INTC?
It is a relatively well run company. It controls its own process and
manufacturing destiny. It owns by far the the most lucrative franchise
in the industry with extreme barriers to entry. It pays a good dividend.
I have a long inside view of the industry being part of it and therefore
comfortable being long in a dominant and well understood player. IMO
there's a good chance over long term Intel will be last man standing.
To the victor goes the spoils.
If Apple tries to tighten the price screw even more from there (1000$ for a phone, are you nuts?), I see lots of people switching camps or at least delay their update.
I have been holding off buying a laptop for a while trying to decide whether to
get an older biz system with Win7, a new system with Win10, or a Macbook.
While Apple is treating its Mac users much better than MS is dealing with its
Windows base right now I fear that an Apple under pressure to keep earnings
strong a year down the road might go full Oracle on Mac users (yet again).
I get the distinct vibe that a lot of people on this board want to see Apple fail because
Or, less disparagingly to fellow board regulars, a lot of people here have seen
a long procession of "unstoppable" tech growth monsters plateau and then
fade to various degrees. Back in the 70s and early 80s IBM was seen as even
more dominant and unstoppable than Apple is today. MS was an unstoppable
force in the 90s until it tripped over its own greed and feet.
The wonder should be why should someone with a brain expects exponential
growth without end.
When the party does come to an end there is always a group of cheerleaders
who do a 180 and retroactively claim to have seen all the signs of deep trouble
years before. The only thing that doesn't ever change is human nature.
sadly I am also hearing that 10nm is just a shrink of 14nm :(
At these feature sizes nothing is "just a shrink".
When a critical layer starts 100 atoms thick a 30% shrink means one thing.
When a critical layer starts 5 atoms thick a 30% shrink means something
else altogether.
So you think that the original poster was being absolutely literal when making his point?
Did you even read the original poster's comment?
"Apple had intended to go with Samsung entirely for the A9 but because Samsung's yields weren't where they needed to be, they did a quick port of the A9 design to TSMC 16FF+."
Such that you feel the need to point out that taking a design from one process to another takes a non-trivial amount of time and effort?
Clearly the original poster has absolutely no idea what is really involved in
even something as "simple" as porting an existing design to a new process.
By the time Apple is producing enough production quality Samsung A9s to
gauge yields the decision to make a TSMC version it is already several years
behind.
To anyone who has the first clue about the IC industry knows Apple would
have had to have two implementation efforts operating in parallel for a long
time to have both Samsung and TSMC based versions of the A9 exist in
the same product cycle at the same time.
If you produce a Samsung A9 and decide Samsung yields suck then you are
basically f***ed for this product cycle. You don't start a TSMC A9 then, you
decide to make A11 or A12 at TSMC. You either fix the yield issues with
the Samsung A9, down spec the part, or you skip A9 altogether and hope for
better luck with A10.
Me, I interpreted it simply as taking an existing design and moving it to a different process, which is easier than designing a processor from scratch.
Clearly you have very little knowledge about IC design methodologies and
the industry in general given your fit about Elmer's reply to the original
post.
I don't talk about my own work publicly but here's a little relevant story
I got first hand from an Alpha designer at DEC back in the day. When EV6
was ported from 0.35 um to 0.25 um (i.e. the EV67) there was a serious
transistor sizing error in a critical path in the FPU that knocked down
top bin clock rates by about 20%. It was an "easy" fix (once they found
the speed path and tracked it down) but it would have required a respin
and restarting the verification process at the silicon and system levels.
The EV6 ports to 0.18 um (one to a Al process, one to a Cu process) called
EV68 were already underway and well along. DEC didn't respin the EV67 to
fix the bug. The fixed EV67 would have ended up reaching market after the
first EV68s. As a result DEC Alpha systems didn't reach the top frequency
target originally envisioned during that year long EV67 market window.
That is the reality of complex ICs like MPUs and SoCs. There is no such
a thing as a simple fix or a simple port.
Not being an expert is not a bad thing. Being an ignorant tw*t actively
being condescending to people who do this for a living is. Even worse is
someone who does understand this and is intellectual dishonest for the
purpose of trolling.
TSMC forecasts 11% YoY decline in Q1 revenues
http://finance.yahoo.com/news/fewer-orders-apple-suppliers-could-014332497.html
But TSMC, which makes some of the chips that go into iPhones, forecast this month that first-quarter revenues would likely fall by up to 11 percent year-on-year, adding that demand for high-end smartphones would also be weak.
An 11 percent quarterly decline would be the steepest revenue drop for TSMC in almost 7 years, Thomson Reuters data shows.
I guess this means the smart phone is dead right?
Donno bout that, but i searched the page for "watt" and found nothing ?
Good point I forgot to mention.
Despite being designed to be able to clock nearly twice as fast as A9X
Intel big x86 cores are comparable in power consumption to A9X at similar
frequencies.
Think about how power efficient Intel big x86 cores would be if they were
designed to be able to only clock as fast as A9X.
Oh, but you were trying to go for the snide cheap shot. Sorry...
Anand runs SPECint2006 (10 out 12 actually) on A9X
http://www.anandtech.com/show/9766/the-apple-ipad-pro-review/4
The A9X significantly reduces the gap with Intel big core x86 at similar
frequency compared to A8X. The trick though is the big Intel cores can
clock nearly twice as fast.
Still, the A9X is a pretty good processor for a n00b.
LOL, it is funny to see low single digit volt capable FETs called "high voltage".
At least if you ever worked with silicon where HV meant high three figures,
Idsats were measured in two figure Amps, and a timing or logic bug could
announce itself with a sound like a rifle shot rather than a blue screen.
AMD: SYSmark, stop looking so closely at our flabby little CPUs
http://www.theregister.co.uk/2016/01/19/amd_accuses_intel_of_vwlike_results_fudging/
In SYSmark, Salinas claims “there is an excessive[ly] high amount of CPU tasking being done – that is, that the benchmark is only evaluating the CPU side of the system.
IOW, our CPUs are so weak it is unfair not to dilute compute performance
with other factors like graphics and memory bandwidth.
If AMD is calling out their tired old wambulance yet again at this time it is
hardly inspiring confidence about upcoming Zen core's oomph.
BTW, AMD releases financial results after close today. Look for another big
decline in compute revenues caught in double whammy between dropping
PC sales and probable share gain by Intel.
IBM has clearly shown that OpenPOWER is starting to flourish and is beginning to pose a pretty serious alternative threat to Intel’s high-end HPC offerings, including their Xeon and Xeon Phi family of processors.
Market data from IDC and Gartner show x86's momentum and RISC's shrinking
market share continue unabated. Historical precedents suggests IBM's dumping
its in-house semi operation will impede its ability to keep POWER competitive
(such as it is).
The markets have been a flaming disaster so far in 2016 and currency
swings over the past 6 months have been insane.
Is it time yet to stick a toe in and start buying or is there still a lot of
downward momentum left to play out in this train wreck? Thoughts?
geekbench is to SPEC CPU what bubba the redneck launching propane cylinders
skyward from his junkyard is to NASA conducting an interplanetary mission.
One is a half educated amateur messing around for lolz and the other is a
large accomplished organization including many technical specialists from a
variety of fields building on the successes and failures of the past using hard
won institutional knowledge.
http://files.shareholder.com/downloads/INTC/1345169316x0x870281/53447830-5FC8-4FB7-87D2-E35A49795B44/Earnings_Release_Final_cs.pdf
Q4 Business Unit Trends
• Client Computing Group revenue of $8.8 billion, up 3 percent sequentially and down 1 percent
year-over-year.
Pretty good considering IDC and Gartner's recent bad news about PC sales.
I can't believe AMD has much hide left for Intel to take share from so
maybe those buying PCs are buying higher end processors on average.
AMD finally launches its ARM server chip.
http://finance.yahoo.com/news/amd-key-industry-partners-welcome-140000015.html
They call it "Opteron A1100 SoC".
Curious it they call this chip Opteron since it is not x86 like every other
Opteron AMD has ever released. OTOH it will likely be as uncompetitive
to its ARM competitors (such as they are) as its x86 brethren are to their
Xeon competitors so maybe the name is appropriate even if misleading.
Gartner says 2015 semi sales fell 1.9%
http://www.theregister.co.uk/2016/01/07/2015_chip_market_figures/
That same weakness in PC demand hit Intel, which saw overall revenues slip 1.2 per cent to $51.7bn. Nevertheless, it maintained its top spot in the industry, and actually increased its market share by half a percentage point to 15.5 per cent.
Samsung saw revenues grow 11.8 per cent to $38.9bn, while SK Hynix saw 3.1 per cent growth to $16.5bn. Qualcomm saw revenues slump 17.5 per cent to $15.9bn, while Micron’s revenue slid 11.2 per cent to $14.4bn.
In most cases the kids on Andy's lawn he shouts at are visible only to him.
That article repeatedly mentions the "ailing PC business". It is fine to
bring in expertise in other markets but the PC business is still absolutely
huge and very profitable. Let's hope the new suits at Intel don't forget to
keep feeding the cash cow even if it is giving a bit less milk these days.
If I see the beginning of a duplication of AMD's headless chicken act I'll
dump all my INTC lickety split.
How does the GPU in the two devices compare?
Each process shrinks buys a bit more transistors and a bit faster logic
at the same voltage. Intel has using the former for more muscular GPU
and the latter to drop voltage and keep *device* power the same or a
bit lower at a similar frequency.
These choices are a logical response to the totality of market challenges
and opportunities. You don't like it buy your x86 processor from someone
else.
Interesting that Broadwell-C/H never made it to particularly high frequencies either; lower than their 22-nanometer Haswell counterparts actually (Intel ships 4GHz turbo HSW chips for laptops and 4.4ghz turbo for the 4790k; BDW-H tops out at 3.7GHz-ish IIRC).
I definitely think it's a process-related problem.
Why?
Intel has been dialling down TDP while increasing the proportion of
area and power budget dedicated to integrated graphics. Desktop no
longer drives core design, mobile does. Generally speaking increasing
power efficiency at 2-3 GHz reduces headroom for operating above 4
GHz. That is a conscious business decision by Intel. With no effective
competition from AMD all overclockers can do is cry in their beer.
Yes the Skylake 6700k inventory is low and so is the Core i5 unlocked part.
I think Intel was surprised by strength of desktop sales and the level of
persistent demand for the 6700k in particular.
But if supplies of 6700k remain tight after another quarter then that would
suggest a frequency bin yield problem for this SKU.
Apple buys a small fab. What are they going to do with it?
Probably to experiment with exotic Si and/or non Si process technologies.
It makes no sense to have a small fab for plain Si process.
AMD doesn't own any fabs any more. Both AMD and GF are on life support.
Germany has basically no say over the future of either company.
Governments have terrible records picking winners and losers in general,
even worse in tech. Dresden fab is a concern for German tax payers, not
Intel and its shareholders.
Gartner Says Worldwide Server Revenue Grew 7.5 Percent in the Third Quarter of 2015, While Shipments Increased 9.2 Percent
http://www.gartner.com/newsroom/id/3174018
Mr. Hewitt said, "x86 servers managed to produce an increase with growth of 9.2 percent in shipments and 9.7 percent in revenue year over year. RISC/Itanium Unix server revenue declined 11.5 percent in the third quarter, while shipments grew 1.1 percent. The 'other' CPU category, which is primarily mainframes, showed an increase of 3.5 percent in terms of revenue."
Meanwhile, IDC says:
http://www.idc.com/getdoc.jsp?containerId=prUS40674215
December 2 , 2015 – According to the International Data Corporation (IDC ) Worldwide Quarterly Server Tracker , vendor revenue in the worldwide server market increased 5.1% year over year to $13.4 billion in the third quarter of 2015 (3Q15), the sixth consecutive quarter of year-over-year revenue growth. During this quarter, revenue grew in rack optimized, blade, and density-optimized servers, while towers declined. Worldwide server shipments totaled 2.49 million units in 3Q15, an increase of 4.5% when compared with the third quarter of 2014.
...
Demand for x86 servers improved again in 3Q15 with revenues increasing 7.1% year over year to $11.5 billion worldwide as unit shipments increased 4.5% to 2.47 million servers.
No surprise here, x86's immense stranglehold* over servers continues to
strengthen. Come on ARM chippers, toss yet more hard cash down the
toilet trying to break into this market.
*85.8% revenue share, 99.2% unit share by IDC figures.
Sorry to see you go but I fully understand your sentiments. Best wishes.