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.
Yeah but is TSMC growing diversity in its workforce?
That's it, I'll be dumping the rest of my INTC at the earliest reasonable
timing. Huge highly speculative acquisitions like this are a clear sign
management is floundering about looking for some direction and strategy
from without.
Each process shrink gives you extra room to enhance some combination
of design capability (transistor count), clock frequency, and/or reduce
power consumption. How you choose to use the bonus (more complex
designs with higher IPC, faster clocking designs, or same IPC/freq at
lower cost and power, or some combination thereof) and what aspects of
product operation you dedicate it improving are all choices to be made.
You have a lot of freedom to act but the absolute quantity of process
bonus is limited these days and if you spread it too widely and thinly
then it can effectively disappear into the noise.
For many product generations now Intel has chosen to spend most of the
bonus on the IGPU in its client silicon and the CPU cores got minimal
improvements. This was reasonable at first, AMD was far behind and
Intel wanted to push x86 into mobile computing and also drive towards
eliminating discrete GPUs from traditional PCs for all but serious
gaming desktops and laptops. The rate of IGPU improvement over the
last decade has been remarkable.
I suspect with the resumed challenge from AMD that Intel will ease off
the IGPU lever and spend more effort again on its CPU cores. Keep in
mind that x86 cores have long been extremely well optimized for scalar
integer performance and all the low hanging fruit picked over the last
two decades. Further gains will be extremely costly and limited. That
is why SIMD extensions have been piling on x86/x64 so regularly. You
just can't speed up basic integer/memory/control flow operations of
the core very much more from the microarchitecture or circuit design
side in a world with little desire to raise power budgets.
they will build what they are asked to build.
Exactly.
A lot of people think the tech industry is like Iron Man or something -
lone engineers working on whatever they want and occasionally one
will come up with something revolutionary and the world changes.
MPUs and Semi processes are the combined output of teams of 1000's
of people with many different specialities and responsibilities working
together for years according to some plan of record. What pops out of
the development pipeline today is the result of decisions made and bets
placed five years ago, give or take. Mistakes from the top have a long
half life in the organization and will be paid in full over many years.
chipguy, what is your opinion of Intel's execution in its core markets?
Remarkably indifferent. Probably the result of a lack of serious x86
competition for nearly a decade combined with a CEO who thinks he
is a politician and social activist rather than the man responsible for
keeping a tech leader well ahead of the pack by exploiting all of its
traditional strengths to the max. There is a lack of paranoid survival
instinct and an apparent diffuse sense of responsibility. The MBAs are
running the asylum and they seem to think R&D is a waste of dollars
because everyone else gave up and went home. Everything now looks
like tinkering around the edges in the absence of coherent sustained
strategy and leadership.
I don't have a lot of faith in current management's ability to respond
to true threats like Andy Grove did in the face of RISC's apparent
ascendance in the late 80's/early 90's or the AIM alliance in the mid
90s.
Speaking of the old man, one's mind boggles to think of the strips he
would carefully and painfully strip off the current pathetic CEO if he
came back and saw what BK did to his baby. Shit, I'd pay serious coin
to listen in on *that* reaming.
ROFL, yeah because GPUs are just MPUs done *right* huh?
it won't be the same old swing of the pendulum.
Right, because tough computing problems look like 3D rendering - nearly
infinitely thread divisible, vector style memory access patterns, laughs at
latency, and complete lack of fine grained control and data dependencies.
GPGPU has done really well in the server market
For some definition of really well, sure I guess.
GPGPU* is useful for a minority of HPC work and HPC is a minority
of the server market.
Don't be so quick to dismiss the viability of alternative architectures or even computing styles in a world
I never dismiss the ongoing reality of the same expensive mistakes and
pendulum over-swinging happening over and over again in computing.
Each generation comes up with a "new" old idea under a new name
and they get burnt well short of the hype for the same *old* reasons.
As regular as the sun comes up. I have been watching the computer
industry since the mid 70s. What a show.
*BTW, you call it GPGPU. More than 30 years ago the same idea was
called attached array processors. Heck, IBM even had an option (IBM
speak: /VF - vector facility) for their mainframes. Pffft, kids these days.
where mega data center operators are willing to invest in reworking their software to get higher perf/lower TCO.
Seen that sorry story before in the 70s. The mega guys that went that
way got burned big time as general purpose computing marched on
and special purpose/microcoded optimized stuff get EOL'd, LOL.
“As we run out of steam with out-of-order execution people will reexamine dataflow execution and stuff that propels us to the next level,”
Sounds a lot like the thinking of architecture researchers at HP Labs in
the early 90's. They hooked up with Intel to help implement their dreams
of the next big step in computer architecture beyond CISC and RISC.
For those who don't know this story here is the spoiler - OOOE and x86
not only survived but flourished. The highly touted EPIC, not so much.
Microsoft is rich enough to keep pushing for as long as it likes.
That is true.
But Intel is only in trouble when ARM servers have a better story
than x86 for OEMs who want to make money and users who want
to save money. MS setting piles of its own money on fire does
nothing to demonstrate such a reality.
Seen that song and dance a few times before. Cost-em plenty in previous
renditions.
Typical AMD cherry picked comparison to make Ryzen look good.
http://www.theregister.co.uk/2017/03/08/the_italian_job_amd_goes_to_naples/
Compare to a high end SMT (scale-up) Xeon with buffered memory
on the basis of an HPC app? There are much cheaper Xeons which
are optimized (CPU count, frequency, memory bandwidth & latency,
cache etc) for HPC. Funny how AMD missed those.
BTW, I wonder if AMD is going to submit SPEC CPU 2006 scores
for various Ryzen on its own. For many years now the only AMD
processor scores were submitted by Intel.
My peak holding in INTC was May 2009 to Sept 2011 after a
long period of buying moderately on exaggerated dips.
For most of 2013 and 1H14 I was at 74% of peak after a bit
of profit taking.
I dropped to 58% in July 2014, 44% in May 2015, and 22% in
July of last year.
Clearly I am not a fan of the BK regime. I am one idiotic SJW
credential polishing press release away from total divestiture.
Perhaps Intel should focus more on minority charge carriers.
Now, please justify the existence and cost of Intel's $1700 10c/20t workstation chip... WITHOUT using the phrase "Intel Inside".
It is an IQ test for rich tech enthusiasts.
Apparently enough guys (it wouldn't take many) want bragging rights to
justify the incremental marketing and channels costs of re purposing a
workstation/server chip as a low volume enthusiast desktop chip.
I could afford one but it is clear to me that the utility isn't worth the
cost. Same reason I drive a Honda not a Porsche or Mercedes. But
I don't begrudge Porsche and Mercedes customers their right to waste
their money as they see fit to impress the minimum wage teen at the
drive through. Intel has a similar right to take $1700 from such folks
as long as they can.
By the way, Ryzen in my opinion shows just how capable Samsung's 14nm process actually is.
It is really hard to compare process capabilities on the basis of a
couple of point products. These two new AMD chips could have their
frequency and TDP defined in a way that 90% of good die could meet
them or only the very best 10% with the expectation of collecting and
down SKUing the rest with follow on economy offerings. The latter
scenario makes the process's capability look a lot better. It is also a
lot more credible given AMD's ambitious (for them) pricing.
The specs of the high volume, low price Zen parts AMD will eventually
sell will be a much better indicator how good their foundry's process is.
There might be some upside in workstation/server but those markets are smaller and have more inertia to new chips.
Considering that AMD has 0.2% server market share (saw that number
recently in press around Zen release) it has nowhere to go but up.
The first Ryzen parts are pretty impressive, especially getting 8 decent
IPC cores clocked at reasonable speeds at reasonable TDP. Not Intel
killers some here hoped but staggeringly better than anything AMD had
before.
AMD will still lean heavily on price to sell but at least they will be able
to sell a fair amount if they can make and bin them in volume.
The best part is finally we have some fire under Intel's ass to release
more competitive SKUs of existing products and stop pushing out intros
of new products. More better products, less diversity hiring self back
patting press releases - finally get it BK? The fate of my remaining INTC
holdings awaits your answer.
Trying to put the ATI humpty dumpty back together again?
If Zen CPU costs 300 dollars and performs like a 450 intel equivalent - why pay more?
That has been AMD's proposition since the 286 days.
Obviously it is more limited and simplistic in practice than you think
a pre-emptive move against ZEN?.......seems plausable
So you think dual core i3s are the natural competitor to $500 octa core Zens?
Gartner saw slower server sales in Q3
http://www.gartner.com/newsroom/id/3530117
In the third quarter of 2016, worldwide server revenue declined 5.8 percent year over year, and shipments declined 2.6 percent from the third quarter of 2015, according to Gartner, Inc.
...
x86 servers declined 2.3 percent in shipments and 1.6 percent in revenue in the third quarter of 2016.
Considering that recently x86 has been about 80% of total server sales
by revenue and 99% by units these Gartner numbers suggest that non-x86
server sales dropped by about 23% in revenue and 30% in units.
IDC numbers will provide a second view of Q3 in a day or two.
Underestimating the competition is a really dumb thing to do. Only the paranoid survive. You don't sound paranoid.
I have been following the computer and microprocessor industries since
the 1970's. In the 60's and 70's IBM utterly dominated computing. Now
they are a bit player. The same fate may await Intel some day.
Now listen very carefully biosbob. Here is the clue that you obviously
so sorely lack. IBM was not done in by Amdahl making cheap software
compatible knock-offs of their mainframes. The competition came from
entirely different architectures, platforms, and business models.
Intel should be paranoid. But AMD is the very least of its worries. If
and when Intel is reduced to a bit player in computing it sure the f***
won't be AMD doing it with cheap softwate compatible knock-offs of its
x86 processors. Now run along kid, go chase some Zen moonbeams
or something.
A troll wants a reaction.
On the AMD board they will just say "amen brother!" and "straight to the moon!".
Where is the fun in that?
You going to cash out on AMD soon or fly the pig all the way back down
into the weeds? Don't expect any sympathy after all your pom-pom shaking
and rude noise making on this board.
Intel will have to respond on Zen somehow.
Why? It is not like Zen will match let alone greatly exceed even
current Intel products across all attributes.
Most likely outcome is AMD will go from being unprecedentedly
behind Intel to a more traditional somewhat behind but competing
mainly on price and their limited enthusiast fan base.
The biggest competitor to an Intel MPU sale is the Intel MPU a
customer is already using. Intel has to keep making better MPUs
to generate replacement/upgrade sales. Zen is a second order
effect at best, third order noise most likely.
Zen's gonna cost $500. That's a lot.
It isn't for a server part.
It may be for an enthusiast part.
It definitely is for a mainstream part.
With out further elaboration a single figure ($500) is meaningless.
AMD's conundrum is pricing the *various* varieties and speed grades of
Zen to under price comparable Intel models deeply enough to generate
sales and still have high enough margins to pay back years of development
costs and shore up corporate finances.
Historically speaking, AMD has rarely accomplished this feat.
Over 80 million shares traded with over 3 hours to go. That is thinly traded?
It is in the context of countless billions in excess capital sloshing
around the global markets looking for a new bet. The "masters of
the universe" class aren't at personal risk for poor bets so why
bother worrying about drilling down into business details, just
follow the animal spirits of the moment.
Funny how you clam up on days it is down 10%.
Volatility is a sign of speculation, thin trading, and/or possible
manipulation. I.e. risk. You cash out yet or is greed still winning
over fear?
Otis who?
AMD's "One-Time" Charges Are Starting to Add Up
http://www.fool.com/investing/2016/11/10/amds-one-time-charges-are-starting-to-add-up.aspx?source=yahoo-2-news&utm_campaign=article&utm_medium=feed&utm_source=yahoo-2-news
Advanced Micro Devices (NASDAQ:AMD) reported a non-GAAP profit during its latest quarter, a vast improvement over recent quarters. But on a GAAP basis, the company posted a massive $406 million net loss. The discrepancy was mostly due to a $340 million charge that AMD took related to its wafer supply agreement with GlobalFoundries. The GAAP numbers include this charge; the non-GAAP numbers don't.
Companies report non-GAAP figures in an effort to give investors a number that better reflects their true performance. Unfortunately, many companies back out recurring costs like stock-based compensation, a practice which I strongly disagree with. Backing out one-time costs, like the wafer supply agreement charge, is often a reasonable thing to do. But when a pattern of "one-time" costs begins to emerge, investors should think twice about accepting those non-GAAP numbers.
A history of "one-time" costs
That $340 million wafer supply agreement charge was the biggest one-time cost that AMD has reported in the past few years, but it certainly wasn't the only one. Between the beginning of 2014 and the third quarter of this year, AMD has taken 15 different one-time charges, with a whopping total of nearly $1.1 billion.
the AMD needle?.......a good chance
LOL. The pity f*** sales OEMs throw to AMD will move from X million
dozer based parts per quarter to X million Zen based parts. Maybe a
somewhat higher ASP will slow AMD's need to borrow. The idea of
some big sea change or reversal of fortune is child like in its naivety
and utterly blind to the lessons of actually observing the industry for
a few decades.
If Zen is even halfway decent I am sure major OEMs will throw AMD
a bone with a minor socket win here and a minor socket win there
to point to when they negotiate their next billion dollar order pricing
schedule with Intel.
But ultimately the customer calls the shot. Tech enthusiast forums
are not representative so don't let a few wild AMD fans fool you
into some idea the wider market is waiting with bated breath. The
vast majority of big buyers don't give a damn about some f***-up
prone, barely alive, alternative supplier of non-socket compatible
x86 processors. They simply want their existing codes to run problem
free and be able to seamlessly upgrade/supplement their thousands
of existing Intel based systems.
AMD going from slinging unbelievably noncompetitive products to
somewhat inferior me-too products really doesn't change anything
in the greater scheme of things.
I don't think the faithful realise just how hard it is to match Intel's x86
products across the spectrum. Besides raw silicon performance there
are the issues of software stacks, developer tools, chipset support,
motherboard support, reference designs, application support and so
on.
I think Zen has the potential to be a much better product than anything
that AMD is selling now. But it is Intel that dictates the goalposts, hell
basically the entire playing field. Even if Zen is an out of the park home
run that basically performs the same as Skylake at similar TDP (very
unlikely IMO) the potential to move the needle is still rather limited.
Don't forget to sell while everyone is talking about the Zen sizzle, you
know, before 3rd party reviewers get to chew on the fat and gristle.
To the moon baby!
Talk is cheap. Back up the truck.
Careful with that Zen pumping. Barcelona and Bulldozer pumpers had do a lot
of backpedalling when the smoke cleared. Yeah, yeah its different this time, LOL.
Weak. Very weak. You having your trolls ghost written by unpaid interns now?
Microsoft ends OEM sales of Windows 7 Pro and Windows 8.1
http://www.theregister.co.uk/2016/11/02/microsoft_ends_oem_sales_of_windows_7_pro_and_windows_81/
Glad I got my new Skylake system in under the wire.
Even if Intel and AMD sold identical chips Intel would still command a
significant price premium because of branding, diversity and redundancy
of global manufacturing/assembly/test operations, and financial stability/
long term prospects.
Zen will return AMD to its traditional level of product inferiority vis
a vis Intel rather than the extraordinary depths of ineptitude that its
MPUs currently reside.
The problem is that traditional trailing edge was never enough to live
comfortably, grow the business, and pay down debt even when the PC
market was going gangbusters and AMD had its manufacturing in-house.
Best case scenario success for Zen simply slows AMD's descent into
further irrelevance and eventual insolvency.
Win-win.
Funny how the articles about Intel's Q3 revolves around the server outlook
(which reflects server market recent history as reported by IDC so absolutely
no surprise to anyone whose head wasn't up their fundament) when client
computing group revenue is up 21% sequentially and 5% YoY.
I guess "PCs no longer dying" isn't a calamitous message to excite page
views, not like say... "OMG servers about to die".