TSMC's struggle in 20nm may have benefited INTC. QCOM and APPL go to Samsung and GF for their next chip at 14nm. Which is good for NV who could have better attention at TSMC's 16nm. Will Intel's 14nm too late for mobile is very depend on Samsung & TSMC screw up or not on their FinFET at 16nm/14nm.
Keep in mind that there were multiple rumors, back in April, that David Einhorn was building a position in NVDA. When the funds released their holdings, in mid-May, it didn't show up in his report.
Now, we will be having the release from the 2nd quarter, sometime Thursday (45 days after the end of the quarter), probably after the close, and IF David Einhorn has indeed built a stake, that will be when we find out. IF he has, the stock will be 19 1/2-20 within minutes.
Just an fyi.
As always, we'll know in the fullness of time.
Sentiment: Strong Buy
Much about Core M is not known as test platforms where tech bloggers can play with and benchmark may be months away. This tells us that Intel is cutting it close in terms of timing for the holidays. We also do not know about clock speeds, which would give away a better feel of trade-offs between performance (valued in traditional PC/servers) and efficiency for Broadwell generally. Apples-to-apples mainstream Core Broadwell will hit the market sometime in 1H15, which we view as a year late relative to last year’s Haswell (22nm) launch. What this may do to Intel’s tick-tock product cadence with the upcoming Skylake processor in 2015 will be interesting to watch, but we would say that product positioning and/or transitions are in flux.
PS: Didn't realize Y! cut off part 3. Here it is again.
Nvidia was always cautious when talking about upcoming Maxwell parts, the first of which was launched back in March and based on the first-generation Maxwell architecture.
This left quite a bit of room for speculation about the second generation chip which was the new architecture shrunk to 20nm. We learned that Geforce GT 750 TI and 750 cards were closer to the Kepler architecture than originally thought, but they were also much more power efficient than Kepler chips, so Nvidia decided to call this transitional architecture Maxwell.
Now the company is ready for the Geforce 800 series and it looks like this chip will be power efficient and powerful, too. For Nvidia it was all about getting the 20nm manufacturing process to the point where it makes financial sense, while gaining a performance uplift substantial to excite customers and get them to buy new cards.
Nvidia will invite a few chaps to its headquarters in the first half of September, but this doesn’t mean that the actual event talks place at the same time. We expect Nvidia to announce the second generation Maxwell before its Q3 2015 financial conference call, simply as it will make it look good. With that in mind Nvidia will look better in the eyes of investors and the potential Denver Tegra K1 64 Nexus design win should help, too.
Mobile Maxwell in 20nm based on the upcoming Geforce 800 series will also come before the holiday shopping season. As far as we know 20nm GPUs from AMD won’t launch this year, but we could be wrong. It is too early to say for sure.
Sentiment: Strong Buy
Intel’s 14nm sees better- than-normal scaling, achieving 2x improvement in perf per watt, compared to the historical 1.6x per gen. In addition to being 1yr ahead in FinFET, Intel expects to have greater transistor density at 14nm (volume now) compared to competitors’ 14/16nmFF (volume in 2015?). Intel was clear in its expectation that greater density translates to a lower cost per transistor relative to competition. Intel’s 2nd-gen tri-gate uses taller and thinner fins, improving drive current (performance), translating to fewer fins per circuit, resulting in better scaling (lower cost) and capacitance reduction (lower power).
Lipacis sees opportunities for Intel to further distance itself using its manufacturing edge:
Having overcome the yield challenges of 14nm 2nd-gen tri-gate using advanced double-patterning techniques, we would not be surprised to hear Intel announce an advancement on III-V materials that further extends its manufacturing lead on future process technologies.
Not everyone is satisfied.
Raymond James’s Hans Mosesmann reiterates an Underperform rating on Intel today, writing that Intel is trying to create a firewall with Core-M against the low-cost parts of competitors using technology from ARM Holdings (ARMH): “Core M, in our view, is Intel’s attempt to create an impregnable barrier for ARM-based players to penetrate platforms above the tablet space. Alternatively, low cost versions of mainstream Core and/or Atom variants (e.g., Bay Trail today) may not be doing the job as hoped or needed.”
But the timing of this is questionable, and Intel is late to market relative to where it should be, thinks Mosesmann: (more...)
from Barrons: part 1
By Tiernan Ray
After Intel (INTC) yesterday discussed some key forthcoming chips, “Broadwell” for PCs, and a lower-power part for laptop designs that can be thin and light and avoid having a fan to cool them, called “Core-M,” the Street is debating what it means for Intel’s positioning.
The parts are distinguished by using Intel’s chip-making technology for transistors measuring 14 nanometers.
Core-M is expected to arrive in devices by this holiday season, while the higher-power Broadwell parts will come sometime in the first half of next year. VentureBeat’s Dean Takahashi was on the scene covering the press event, and AnandTech’s Jarred Walton offers the site’s usual excellent deep dive into the details.
Drexel Hamilton’s Rick Whittington reiterates a Buy rating on Intel, and a $50 price target, writing that the company is ”Brandishing The 14nm Mailed Fist.”
Intel is “the only logic company still on Moore’s Law improving performance, diminishing cost curve, with others falling further behind,” he writes, noting that “14nm half the die size in XY axis has longer, thinner Z fins, Intel is leading with inherently higher yielding small, low power mobile parts.”
Jefferies & Co.’s Mark Lipacis reiterates a Buy rating and a $45 price target, writing that the unveiling is a “watershed” for the company, “as we believe it is the first process node where Intel will have both a transistor density and cost advantage.”
“We ultimately believe this will lead to share gains in tablets and handsets.”
Aside from just transistor size, what’s important to Intel, notes Lipacis, is the fact that Intel has already been ahead of competitors such as Taiwan Semiconductor Manufacturing (TSM) in making chips that have 3-D transistors, which it calls “TriGate” and which the industry broadly refers to as “FinFet”:
"they are comparing their 2.5 GHz Denver only against a 1.6 GHz Bay Trail Celeron and a 1.4 GHz Haswell Celeron"
A legitimate comparison for device intended for the same target devices.
"should be a nice profitable price range under Core into which it could fit safe from any price bombing Intel does on Atom"
correct, this is where the ARM ecosystem (or at least this iteration) exposes intel despite all the noise about process
OUTHOUSE! we missed you.
lobstirpee 1 post | Last Activity: 15 minutes ago Member since: Aug 12, 2014
Keep talking it down, just as you have been doing since the low teens.
Nvidia's valuation is still dirt cheap, especially after the strong Q2 results and guidance. Hopefully we'll bust into the $20s soon after a brief rest. I noticed that the short interest climbed that last 2 weeks of July, though many of the shorts must have covered last Friday. Still can't believe so many folks are betting against this company. Nuts...
I got the impression from my non computer science/electrical engineer background the cores were instruction set agnostic. If that is the case it may open doors for licensing the design to IBM for their own use. In the last conference call JSH mentioned they are open to licensing designs not just patents.
i have no idea, i think it is more to do with loving intel. it seems they cant love both.
remember also Nvidia is starting from a low volume base so it will not take much interest to start making a material difference. This chip is like Nvidia's version of Itanium but higher clocked with lower performance and power.
There is more potential for volume when it's made into a quad at 20nm because at the moment it's not going to be much faster than a A57 at the single thread core level and Samsung/Qualcomm will be churning out loads of those 20nm A57 quads next year. Still it does not look a dud assuming there are no glass jaws with the particular in-order concept they have chosen and even the 28nm dual-core version should sell some if the A15 K1 is managing to sell.
How portable do you think this translation layer is? It would be really impressive if it can be easily tweaked to run MIPS, Power, etc.