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.
Nevertheless, Intel needs to speed up its Core line significantly. They were too concentrated on bringing power down, which was important, but performance needs to increase more than 10% per generation.
It's a question of priorities. Basically from 65nm Core 2 to 14nm Broadwell Intel has increased performance/clock by a factor of 2-3 while reducing power by a factor of 4-5. It would not have been possible to reduce power by a greater factor than performance increased if the performance increase had not been limited by maintaining a constant quad-issue pipeline all through those generations.
This limited the performance gain but will eventually allow the Broadwell-Y to play in the same fanless sockets as traditional ARM chips which is quite an achievement for a traditional high-performance PC processor. The turbo ones will be able to maintain the traditional Intel high performance advantage but in a new fanless environment thereby increasing the TAM for expensive Intel high performance chips. I suspect having reached this important market positioning benchmark Skylake will change the development priority and concentrate more on performance going forward especially as Atom is supporting it well now in mobile performance.
What is funny is the claim JPM make about Apple finding it so easy to increase A8 GB3 performance that much that quickly. They will only be able to do it mainly by clockspeed in single-thread as Apple's ex-Dec guys have already played their Alpha EV6/7 card and I don't see a mobile Cyclone derivative at 2.6 GHz even at 16FF. Cyclone also has trouble repeating its GB3 dominance in common application based benchmarks too questioning again the importance anyone should place on GB3 numbers. The fact that JPM does so much just staggers belief and cries out for the Spec org to release SpecCPU2000 as freeware to end all this bench-marketing nonsense.
How One Tiny Chip Changed The World
http://finance.yahoo.com/news/one-tiny-chip-changed-world-222100079.html
When Robert Noyce and Gordon Moore founded Intel in 1968, they came up with what they called the "Goldilocks strategy.
They wanted to make memory chips, and started with three very different technologies. One was too easy for rivals to copy. Another added too much complexity.
"And the third one, using a different kind of transistor, called a MOS transistor, turned out to be just right," Moore said in an interview years later.
Intel (INTC) would apply the Goldilocks strategy again years later when it made the leap to microprocessors, launching it to global titan status.
Its chips powered the PC revolution, bringing cheap computing to every desk. Its relentless innovation and clever "Intel Inside" branding helped the company reach new heights in the 1990s as the Web made the Internet a global obsession.
Intel shares rose 22,132% from its August 1986 low to its peak in August 2000.
"The company has really pushed leading-edge technology as a corporate philosophy and strategy," ex-CEO Craig Barrett told IBD. "Look at the stuff it's brought the world. Intel has been there every time the rest of the industry says it's time to slow down.
Out Of Memory
But in the 1970s, Intel was making a good living selling memory chips when Japanese companies teamed up to attack and take over the market, according to Barrett, who joined the company as a manager in 1974.
"We were getting our butts kicked in the memory market," in the 1970s and early '80s, Barrett said. "The Japanese were taking over and we were looking for some place to go to excel.
That's when Intel struck gold.
Intel's big break came in 1981 when IBM chose an Intel microprocessor to power the first commercially successful personal computer, the IBM PC.
Intel engineer Ted Hoff had invented the microprocessor in 1971. A microprocessor is a programmable device that fits on a sliver of silicon. Essentially, it's a tiny computer. It stores software and also runs software stored elsewhere on the computer.
By the early 1980s Intel was a leading microprocessor maker, and its chips ran more software than many of its competitors.
High Tech, Low Price
An Intel spokesman recently told IBD that it was a competitive process but for IBM (IBM) "it had a lot to do with ease with which they could deploy their systems.
IBM decided to buy parts on the open market to keep the price down and boost sales. Intel could supply large quantities of fast, cheap chips that could run the software IBM wanted on its PC.
IBM picked Microsoft (MSFT) operating software. The software maker's future co-founders Bill Gates and Paul Allen had built their own personal computer as teenagers using an Intel chip.
IBM gave its new chip partner, wobbly from the Japanese memory market attack, a large capital infusion. Big Blue in December 1982 said it would buy $250 million of Intel stock, giving Intel the resources to fend off Japanese companies that were eyeing the microprocessor and do R&D on its next-generation chip.
Dan Hutcheson, CEO of chip industry market tracking firm VLSI Research, says Intel proved by the early 1980s that it was the undisputed industry leader — partly just by staying alive.
"It's an amazing company, not just for being innovative but in not dying like most companies do. Instead, Intel reinvents itself," on a regular basis, he said.
The IBM PC came to dominate the personal computer market, opening the floodgates for PCs to show up on more than 1 billion home desks and countertops.
IBM decided not to hold Intel to an exclusive contract, so other PC makers such as Dell and Hewlett-Packard (HPQ) began to buy its processors.
Intel quickly became the industry standard, with the exception of Apple (AAPL), which at that time used a Motorola chip. Today, Apple also uses Intel processors.
"Intel basically — I guess you could call it foresight — could see microprocessors would be great and would bring a big computer revolution," Barrett said, adding, "In the early '80s when the PC market took off, it was the same excitement about that market as when the company started.
Barrett credited Noyce, Moore and Andy Grove. Grove was Intel's fourth employee and CEO from 1987 to 1998, during the bulk of the stock's long run.
"It was their vision and foresight that made the company happen," he said.
'Copy Exactly'
Barrett, who succeeded Grove as CEO and ran Intel from 1998 to 2005, in the 1980s developed its vaunted manufacturing network, using what he called the "copy exactly" method. Every Intel factory is, as much as possible, an exact copy of all others. This ensures high, consistent quality no matter where on the globe Intel makes its chips.
Barrett says Intel bet its future on a presentation Moore gave when he was at Fairchild. Moore said that by shrinking the size of chip features, their power could roughly double every two years. That principle came to be known as "Moore's Law.
Moore's Law To Improve
Although it's not a law of physics, the industry has followed it for nearly 50 years.
"Moore had made computing power affordable" for consumers to put on every home desktop, Hutcheson said, adding, "Intel has used Moore's Law as a guide better than any other company in history.
The IBM PC debuted in August 1981, powered by an Intel 8088 chip that ran at 4.77 megahertz. Intel has rolled out ever more powerful chips. Its current flagship Core i7 processor runs at 3.5 gigahertz, about 734 times faster than the 8088.
Intel chips expanded into servers as the Internet boomed, vastly increasing the need for high-powered computing.
Intel remains the PC chip leader, but it's had a tougher time since the dot-com bubble burst in 2000. PC sales are falling fast as consumers shift to mobile devices.
Smartphone Static
Its chips are in HP, Dell, Samsung and Lenovo tablets. But Intel has struggled to crack the smartphone chip market, which is dominated by Qualcomm (QCOM), U.K.-based ARM Holdings (ARMH) and Korea's Samsung, which makes many chips that go into its own smartphones.
Hutcheson says Intel is making headway. "The company has doubled down on mobile phones and accelerated the rate its mobile silicon is coming out," he said.
Barrett said he would like to see Intel stay on the road it's paved for itself.
"I'd like them to do pretty much what the company has done in its history, continue to invest heavily in R&D and move the market forward to bring out new products and not be afraid of competition.
Asked what challenges entrepreneurs wanting to start the next Intel need to be aware of, Barrett said stay focused and ignore negative buzz. "If you've got a great new technology, focus focus focus onto it," he said.
Correct, although they have plenty of promises ...
http://www.qualcomm.com/media/releases/2014/04/07/qualcomm-announces-ultimate-connected-computing-next-generation-snapdragon
These products underscore Qualcomm Technologies focus on 64-bit leadership, accelerating its availability across all product tiers while maintaining a long-term commitment to the continued development of its own next-generation custom 64-bit CPU microarchitecture, with more details expected to be shared later this year.
“The announcement of the Snapdragon 810 and 808 processors underscore Qualcomm Technologies’ continued commitment to technology leadership and a time-to-market advantage for our customers for premium tier 64-bit LTE-enabled smartphones and tablets,” said Murthy Renduchintala, executive vice president, Qualcomm Technologies, Inc., and co-president, QCT. “These product announcements, in combination with the continued development of our next-generation custom 64-bit CPU, will ensure we have a tremendous foundation on which to innovate as we continue to push the boundaries of mobile computing performance in the years to come.”
It only takes 1.5-2 years to put stock already developed ARM cores into SoCs so Qualcomm have not done anything special, in fact they are going to be late to the A57 party compared to all the Asian companies.
The 120 MA provided the critical support at 2.15-2.25 to provide a solid support floor to bounce the stock right back all the way above the 50 MA today. There is still a little moving average resistance left peaking at 2.85 (71-74 MA) but that should not take long to convert back into support and then the stock marches back up into $3+.
http://www.nasdaq.com/symbol/astc/stock-consultant
BULLISH
Confirmation - Strong bullish 3 day chart pattern with Average 3 day accumulation.
Confirmation - CONFIRMED breakout above 2.72, no resistance in area just above.
BREAKOUT
CONFIRMED breakout above 2.72, no resistance in area just above.
Type: Continuation breakout from Single resistance.
Target: 3.08, 12.8% Stop: 2.58, Loss: 5.5%, Profit/Loss ratio: 2.3 : 1 - Good
CURRENT PRICE
2.73, just above resistence, 2.64 ± 0.08, type Single, strength 3
RESISTANCE ABOVE
+15% at 3.14 ± 0.1, type Triple, strength 10
+37% at 3.74 ± 0.12, type Single, strength 3
SUPPORT BELOW
-3.3% at 2.64 ± 0.08, type Single, strength 3
-16.1% at 2.29 ± 0.07, type Single, strength 1
-40.3% at 1.63 ± 0.05, type Single, strength 8
-45.1% at 1.5 ± 0.05, type Double, strength 5
-49.1% at 1.39 ± 0.04, type Double, strength 2
-53.8% at 1.26 ± 0.04, type Single, strength 3
-71.1% at 0.79 ± 0.02, type Double, strength 2
-76.2% at 0.65 ± 0.02, type Single, strength 1
Could Intel be using the "3G part" of the XMM 7260 solution?
Quite possibly or even the 3G part of the XMM 7160 if it is smaller.
I believe SoFIA 3G is actually based on XMM 6360
That was a 40nm chip so more likely the 28nm XMM 7160 then in 3G mode.
Ok but remember the 3G/LTE chip in both variants is the same, the XMM 7260 so the size difference is just the 2 extra cores and maybe a bit more gpu e.g. 1 cluster in the 3G and 2 in the LTE.
I wonder what Intel is sacrificing to get that density.
Probably nothing as it looks like LTE is the 14nm version.
If Bay Trail Entry was so easy and quick to make I really don't get why Intel released Bay Trail first which has the BOM issue. What type of planning was that?
For the Windows market where sales are virtually guaranteed.
How much sense does it make to announce huge contra revenues and then not to release the product and wait instead to make it cheaper?
It makes perfect sense when the investment community reacted so negatively to the 1.5% margin hit that came along with it. As the months rolled by that seemed the most likely reason to me too and said so here many times. They might still need contra-revenue for the ancillary chips like gps but that's more closer to $2 than $20 and the margin hit won't be 1.5% accordingly.
I don't believe the Windows Bay Trails need subsidizing as they have no effective competition in their performance/power class that is fully Windows compliant e.g. Windows RT and AMD chips.
So I believe what will happen now on Android is that the Bay-Trail entry chips will sit in between Clovertrail+ and Merrifield/Moorefield and above those 22nm chips will be Cherry Trail and Broxton when they are released. Intel has done the right thing with Bay Trail on Android as it obviously missed its release window to be an all round top-end chip and will push Moorefield there instead when it is released. Ashraf agrees ...
http://www.fool.com/investing/general/2014/03/30/intel-prepares-more-competitive-tablet-chips.aspx
TA-wise if INTC can stay above 25.85 which is where some sticky 15+ year old moving averages are positioned it should move much more freely in an upward direction if earnings don't disappoint this time.
I suspect they anticipate unleashing more clockspeed with future derivatives at smaller lower power processes e.g. 20nm, 16FF etc.
Samsung, Intel, TSMC to account for 52% of semiconductor capex in 2014, says IC Insights
http://www.digitimes.com/news/a20140326PR202.html
Press release, March 26; Alex Wolfgram, DIGITIMES [Wednesday 26 March 2014]
Samsung Electronics and Intel are each forecast to spend at least US$11 billion in semiconductor capital during 2014, and TSMC slightly less than US$10 billion. Collectively the three companies are forecast to account for 51.8% of total semiconductor industry capex during the year, according to IC Insights. This is a decrease from the 55.5% share these three companies held in 2013. Among the top 10, six companies are forecast to spend at least US$3 billion in 2014, and nine suppliers are forecast to spend more than US$1 billion. After keeping their combined spending essentially flat in 2013, the top 10 spenders are forecast to boost capex spending by 10% in 2014.
After chopping its capital spending by 28% in 2012 and 12% in 2013, SanDisk is forecast to show the largest capital spending percentage increase (86%) among the top 10. The company stated that this large increase is needed to expand production of advanced 3D NAND flash memory with its manufacturing partner Toshiba. While SanDisk's capital spending level is expected to be much higher than in 2013, this increased spending is not expected to result in a significant boost to its NAND flash capacity levels. It is worth noting that the combined capex spending increase of Toshiba and SanDisk (on account of their joint venture partnership to build flash memory) is forecast to be US$1.06 billion in 2014, the firm said.
Capital spending budgets are forecast to increase by US$1 billion or more at two companies. Micron is expected to be very aggressive by increasing its spending US$1.12 billion in 2014. Still, this 58% jump in spending is forecast to be less than the 88% jump in sales revenue the company logged in 2013 (after the inclusion of Elpida's sales). Also, pure-play foundry GlobalFoundries is expected to increase its semiconductor capital spending by US$1 billion as well during 2014.
SK Hynix is expected to display an 18% increase in spending in 2014. However, given the company's excellent sales performance in 2013 (43% growth) and its aggressive rebuilding program for its fire-damaged China DRAM fab, IC Insights said it believes there is potential upside to this estimate.
Some of the most eye-catching numbers are the massive amounts of spending expected by Samsung and Intel over the 2012-2014 time period. Over this three-year period, Samsung is forecast to spend US$35.3 billion, with about 60% of this amount targeting memory production. Intel is forecast to be second to Samsung in total outlays over this same time with US$32.6 billion in dedicated to capital expenditures. These huge levels of spending are enough for each company to construct and equip eight or nine US$4 billion leading-edge 300mm wafer fabs, the firm added.
Nine of the top 10 semiconductor industry capital spenders are expected to increase their semiconductor capital expenditures in 2014. In contrast, only four of the top 10 capital spenders increased their spending levels in 2013. Meanwhile, capex spending among "other" suppliers is expected to grow at a much slower 3% rate in 2014, but that is a marked improvement from the 15% decline registered by the "other" segment in 2013. In the long run, IC Insights added it believes that "other" companies will likely increase their spending at a lower rate, or decrease their spending at a higher rate, as compared to the top 10 companies as they implement the fabless or "fab-lite" business models for their IC production.
http://www.fool.com/investing/general/2014/03/11/intels-missing-chip.aspx
A search on LinkedIn revealed that there was not a single reference to a Morganfield platform and, indeed, the only three system-on-chip products associated with smartphones were to be Tangier, the SoC found in Merrifield, Anniedale, the SoC in Moorefield, and Broxton, as shown from this LinkedIn profile:
It is likely that Intel never had any intention of putting the Airmont CPU core in a smartphone-targeted system-on-chip and wanted to skip straight to Goldmont.
That also shows the XMM 7160/7260 is not a big chip i.e. not a big incremental cost on a discrete Merrifield.
Then it's looking like a 1 cluster gpu instead of the 4 in Merrifield. They probably should use 2 as the 1 might perform no better and probably worse than CLT+ and how competitive would that be in 2015 ?
Are you sure that's not the 14nm version though ? The cpu blocks are smaller in size to Bay Trail's when in fact they should be bigger.
You have worked this out before so I wonder if you could just remind me what in your considered opinion the annual cpu development costs of Itanium were at its peak and now ?
Follow the proper procedure and ask Admin to review them.
http://investorshub.advfn.com/boards/ReviewRequestQueueBoard.asp?board_id=369
To generate another day's income.
Agreed, very little new but that I suspect that was not the point . I think I am done with all those Y! SA/MF links, not worth the time reading now.
It takes some time to fully assimilate and assess all the information in that post which is a testament to all the hard work you put into it and so thanks for providing it. The number of threads each SPEC/SAP entry has would also have provided useful additional information as a suggestion if you ever update it or do something similar again.
That magic cloud is continuing to drag ASTC up . What does the red part mean in April, resistance ?
I doubt they would put Merrifield into a tablet that is superseding a tablet that had a quad-core in it previously. Also as there have been no Android Bay Trail tablets released yet I doubt there will be any now with Moorefield so close as the latter is more suited to the Android tablet market. Nexus 8 does appear to be the launch vehicle of Moorefield.
This also opens up the question now if there are to be no Android Bay Trail tablets as previously stated then what does this to do to all the contra-revenue that was supposed to be thrown at them ? PO took a lot of criticism for various reasons but at least he was a man of his word but very little of what BK has said so far has come to pass as he stated. Of course it makes much better sense if Moorefield is the 22nm Android Silvermont design rather than the Windows orientated Bay Trail but it would have been nice if shareholders and analysts were informed about this beforehand, who knows they might even have given INTC a share bump as a result .
https://communities.intel.com/servlet/JiveServlet/downloadBody/21856-102-1-25144/Quark_Package_Mech_Drawing.pdf
https://communities.intel.com/servlet/JiveServlet/previewBody/21856-102-1-25144/Quark_Package_Mech_Drawing.pdf
If I am reading this right the (original) 32nm Quark die is 5.324mm * 5.817mm = 31 mm^2.
It's just Moore's law in action. Basically what would have been a top notch ~100W+ server chip 10 years ago is now going into a 5-10W phone/tablet chip e.g. The Alpha EV6/7 -> Apple A7. This is why the Riscs led by ARM are having a mini-resurgence but I suspect lower process sizes will work against them again as x86 has more issue width headroom to play with going forward.
It wouldn't quite fit my theory that Apple is smarter than the others, knowing that more than two cores are a waste in mobile applications. I really hope Apple is not going after this bigger, faster and more bullshit like Samsung.
OTOH Apple like to say they are doubling some metric or another on a pretty slide every year . I would not be surprised if they went quad-core and maybe also upped their L3 to 6-8 MB to increase single-thread performance. I suppose it makes sense to take advantage of the density increase at 20nm and double the cores and then add SMT and increase clockspeed at 16FF when there is only a performance/power increase with negligible density increase which would get swallowed up anyway by adding SMT. Then they could double the SMT threads for each of the next two years (i.e. 2 to 4 and then 4 to 8) while fattening up the core in a quad-core to take advantage of it while also increasing single-thread ipc in the process. That's Apple's mobile chip sorted for the next four years .
A Silvermont cpu pair including its 1 MB L2 cache is ~8 mm^2.
The more performance you have the more sophisticated you can make any design. Whether the slot in between Cortex-A and Cortex-M is worth populating with Quark we are about to find out
Quark is in a different performance class to Cortex-M especially as it's going dual-core at 22nm. The point is will it be sufficiently lower power and cheaper than a small Cortex-A to be relevant, that is the big question mark about it.
The Kinetis KL03 will be sampling to Freescale partners in March for testing with the FRDM-KL03Z Freescale Freedom development board with assorted developer tools. The MCU will be made available at volume for $0.75 per unit in 100,000-unit quantities beginning in June, the company said.
The KL03's 48MHz ARM Cortex-MO+ has a bit manipulation engine for "faster, more code-efficient handling of peripheral registers," according to Freescale. The tiny MCU has 32KB of flash memory, 2KB of RAM, and 8K of ROM with an on-chip boot loader, with high-speed 12-bit ADC and a high-speed analog comparator. It operates on a 1.71-3.6-volt charge and utilizes low-power wake up and other low-power modes for different tasks.
This stock will only behave with abandon when it clears $26 as it briefly did in January. Under that price moving average resistance is too high because it is of a very long term nature and thus the toughest to break down.
That does not back up your original claim that ARM processors were crippled on GB2 just ones that run on iOS. Silvermont does relatively better in GB2 (integer) against Android ARM chips than in GB3 even though it has an AES advantage in GB3. Something else for you to ponder before deciding GB3 is the first, last and only word in low-level cpu benchmarks.
Good for him to have taken the initiative to TRY to come up with a decent low level benchmark in a barren mobile benchmarking landscape.
So just because he's come up with a free benchmark we should just accept it as the gold standard by which you boldly claimed A7 was twice as fast as Merrifield no ifs buts or maybe ? The comparison you used was flawed. Flawed in that Merrifield's score was 32-bit so missed out on a ~10% 64-bit uplift. Flawed in that Silvermont does not have SHA accelerator instructions so A7 has another ~10% apples to pears advantage. Flawed in that WTH are cryptology tests doing in a general purpose integer benchmark in the first place.
Everybody has to deal with imperfect benchmarks, the real analytical deal is recognizing those imperfections and parsing what is relevant out of those benchmarks to the conclusive statement you are trying to make. You did not do that, you just saw a headline figure, took it as gospel and then went and wrote a sensationalist and what you thought was news-breaking article out of it. This is how tabloid journalists work, shock and then print the real facts later if they don't tally with the original piece.
You also still have not proven how Geekbench is more relevant to a mobile user than WebXPRT in revealing to him/her which processor is faster for their general needs or that WebXPRT is a useless tainted benchmark which you implied.
Earlier versions are crippled on ARMH with sub-optimal compiler flags.
Prove it and I mean both showing the flags and the proof they hindered performance.
he only low-level, cross-platform CPU benchmark we have
This is not true, there's the earlier versions and more importantly there's SpecCPU2000/6 so stop spreading mistruths.
So, I guess to you FPU performance is completely irrelevant on a phone/tablet.
Not completely but basically yes and I have been consistent in that over years and am in good company (e.g. L. Torvalds). There is no SpecCpu2006 differential between two processors that is about 100% different than the SpecCpu2000 differential which shows GB3 and GB2 are not fundamentally measuring the same thing. For you only to use GB3 with no provisos about general applicability to other tests measuring compute ability shows an inadequate analysis and a distortion of the whole true picture. It's like saying AMD make better general purpose gpus than Nvidia because they are better at mining litecoins because that is more what GB3 is testing for now.
As soon as Intel apologizes for: ...
Your true motive reveals itself.
The author acknowledged that part of it is bugged for x86 and fixed it for GB3. The deal is you have taken an unrepresentative specialized (because of the crypto) benchmark (GB3) and boldly claimed that A7 is twice as fast as Merrifield when the previous generation of that benchmark actually showed it is generally slower in the known bug-free sub-tests. You owe your readers and Intel a big apology for such an idiotic ill-researched badly analyzed article but as it was so bad on so many levels I think it is irredeemable and I suggest changing the authorname to Kofi Bofah and hope no-one notices.