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.
The Power 8 supposedly has 12 cores per chip* but that version only has 6 and is obviously a bit of a fail per socket chip against Xeon. The 12 core version though should maintain Power's general purpose relevance going forward. Even for the 6 core version where software pricing is done by core e.g. Oracle it will still be attractive.
* http://www.theregister.co.uk/2013/08/27/ibm_power8_server_chip/
As I suspected the Power 8 core is pretty tasty. Intel needs 4 threads per core and probably a better core like Skylake to catch up in per core performance. Haswell with its 4 integer ALUs has already reached the point where 4 threads per core would become a useful performance boost over 2.
Maybe A15 cpu performance and then only a DMIPS number late in 2010 early 2011. I don't think gpu future progress was that clear then.
http://arm.com/about/newsroom/arm-unveils-cortex-a15-mpcore-processor-to-dramatically-accelerate-capabilities.php
http://www.itproportal.com/2011/03/14/exclusive-arm-cortex-a15-40-cent-faster-cortex-a9/
Silvermont is a 2009 design as can be seen from its Westmere era heritage and BYT-T was released in 2013. Goldmont will be the first Atom design that fully takes A15 into account.
But I truly believe that Intel had higher hopes for Bay Trail-T than going up against Rockchip.
I am sure they did ... 4 years ago when the design spec was being finalized . Also BYT-T seems fine both technologically and financially for the Windows market it was intended for 4 years ago. Where was Android 4 years ago, did it even exist ? Was the Snapdragon 800 series specifications known then as a design target to aim for ? You expect too much from educated guesses about future competition 4 years out.
Intel is dumping them into the low end as a "big fish, small pond" proposition to compensate for the fact that its performance is not up to snuff with the high end QCOM parts.
No, the target there is not Qualcomm but Allwinner, Rockchip etc. Intel has bigger ARM fish to fry then just Qualcomm, it's going to disrupt ARM from the bottom questioning over time the whole licensed fabless business model and only the strongest and best licensees will survive the continual relentless pressure.
Bay Trail-T is a failure both technologically
I am pretty sure also Bay Trail is generally faster than Snapdragon as ARM themselves grudgingly acknowledged so I don't see how that is a technical failure. If you are worried about the gpu Moorefield will fix that too as a Silvermont sku. But of course you already know all this ...
Well they could not be that bothered with Windows RT. It would not surprise me if they don't actually wait for the 4 GB memory limit to be breached forcing a 64-bit o/s on them .
A57s will also debut mainly in 32-bit o/s mode making my comment even more relevant.
To remove the effect of the new cryptology instructions to get a sense of the true integer horsepower improvement.
I was comparing A15 to A57 like ARM was in its slide, ARM apples to apples. In 32-bit mode a 2.3 GHz A15 Exynos will outperform a 2 GHz A57 Snapdragon 810. This is evolution not revolution.
Is only 16% the true apples to apples improvement in integer using the same software i.e. same 32-bit gcc ? Considering A57s will struggle to go past 2 GHz that is not that brilliant.
An interesting analysis albeit from an obviously biased source. I know you looked at native game engine ratios before so the question I have is how relevant are the apps and games for which translation is still required ? ISA translation has always been inefficient so no news there. Of course if Intel is first with a 64-bit android platform then going forward this legacy app/game translation factor becomes less and less important.
The slide that interested me was the SoC power one. Firstly is it accurate ? If it is then what is causing it, inefficient gpu and accelerators compared to Snapdragon ? If it's gpu then Merrifield/Moorefield and hopefully Cherry Trail will address that. Like I always thought and said Mobile is going to be a long hard generational slog for Intel like Servers were and these are just the first initial encounters. The important thing is that Intel continues to deliver better performance/price every year and out-price the weakly funded ARM licensees and eventually it will be one of the last left standing.
They paid him exactly what he was asking for e.g. the o/s value of his contract and it sure was one expensive temper tantrum Pickens threw in sacking him
http://www.astrotechcorp.com/news-and-events/news/astrotech-reports-first-quarter-2013-financial-results
http://www.sec.gov/Archives/edgar/data/1001907/000151597112000406/astc10q093012.htm
(14) Commitments and Contingencies
The Company is subject to various lawsuits and other claims in the normal course of business. In addition, from time to time, the Company receives communications from government or regulatory agencies concerning investigations or allegations of
noncompliance with laws or regulations in jurisdictions in which the Company operates.
The Company establishes reserves for the estimated losses on specific contingent liabilities, for regulatory and legal actions where the Company deems a loss to be probable and the amount of the loss can be reasonably estimated. In other instances, the Company is not able to make a reasonable estimate of liability because of the uncertainties related to the outcome or the amount or range of potential loss.
On August 3, 2012, the Company terminated John Porter, the Company’s former Senior Vice President, Chief Financial Officer and Secretary. In a letter to the Company dated October 17, 2012, Mr. Porter demanded substantial monetary compensation totaling approximately $1.0 million from the Company as a result of his termination. The Company has not yet been served with a formal complaint. While the Company intends to vigorously defend any such actions by Mr. Porter, the Company recorded a liability at September 30, 2012 for its estimate of the probable loss in this matter.
They already took the charge in a previous quarter. Pickens knew then he would have to settle as he only fired him because he questioned his behavior which is not proper grounds for dismissal.
Google POWER8 server motherboard
https://plus.google.com/111282580643669107165/posts/Uwh9W3XiZTQ
http://www.wired.com/2014/04/openpower/
Today I'm excited to show off a Google POWER8 server motherboard in the OpenPOWER booth at the Impact 2014 conference in Las Vegas. We're always looking to deliver the highest quality of service for our users, and so we built this server to port our software stack to POWER (which turned out to be easier than expected, thanks in part to the liitle-endian support in P8). A real server platform is also critical for detailed performance measurements and continuous optimizations, and to integrate and test the ongoing advances that become available through OpenPOWER and the extended OpenPOWER community. (Google, IBM and others formed the OpenPOWER Foundation, a non-profit organization dedicated to developing an open ecosystem.
http://www.spacenews.com/article/military-space/40364effort-to-exempt-satellites-from-russia-sanctions-complicated-by-latest
This could force western satellites to use western rockets e.g. Atlas V or Delta IV which could send a whole load of new business ASTC's way.
Intel missed mobile, but won't miss wearables: CEO
http://www.cnbc.com/id/101624638
"We are catching up. I should just start out with admitting that," CEO Brian Krzanich said in an interview with CNBC.
Intel was behind in implementing the latest communication technology in its mobile platforms and has since paid a steep price, he said.
"We were behind. We were late to LTE," Krzanich said, referring to long-term evolution, the most up-to-date standard in wireless communication speed. "But we now believe we have the full suite and we believe we are on the edge or leading edge in almost every category. We think it's critical because I don't think a single product that we make will not have various levels of communication in them."
Merrifield/Moorefield seem to simply drop off the map when Intel talks about 2015.
The way Medfield/Clover Trail+ has dropped off this year ? . The roadmap is for *new* skus but that does not mean older skus will not be sold and Merrifield/Moorefield will be above those 28nm Sofia parts in performance and will therefore outlast them in the marketplace. Don't be surprised to see them around in 2 years time fighting for marketshare.
But the puzzling thing is that at IDF Shenzhen Intel did not talk up Moorefield/Merrifield for tablets even once. All of the presentations centered around Bay Trail-T and Bay Trail-Entry for Android tablets.
That's because those SoCs were specifically designed for the Chinese markets and the info about them was new and interesting. Plenty of time to talk about the premium Android parts when they eventually ship. Merrifield it looks like was not ready to ship at MWC but they committed to a launch then so which turned out to be a paper one but the benches were interesting as a prieview. Moorefield looks like it might be the Nexus 8 64-bit showcase. Yes all these Android parts are late but they will have plenty of time to compete again 28nm ARM parts going forward because that node is going to be in fashion for a very long time.
Intel's process advantage went completely and utterly to waste during the 22nm generation in mobile.
No, Moorefield and Merrifield will go 2nd and 3rd on that list only behind Haswell. The 22nm Atoms will be competing against 28nm ARM parts for a long time to come if the 32nm Atoms are any indication. The Battle against the ARMy has only just been joined.
Could be corner cases where working sets fit into the Power 8's large caches and/or take advantage of its big bandwidth. I expect the Power 8 to spur a re-buying cycle from all previous Power users as it is quite an advance over previous generations, probably the state of the art in modern RISC architecture.
It's hard to imagine developing such a wide core as A7 and not multi-threading it eventually if the architect technical competence is available.
Most if not all commercial server software has probably been running longer on Power than on x86 although of course bespoke software that sites write themselves is I expect heavily in x86's favor now.
With up to 8 threads per core it may be closer than you think.
http://www.fool.com/investing/general/2014/04/22/why-intel-shouldnt-give-up-on-cellular.aspx
Just one thing Ashraf, this point you made 'most of Intel's business today in this division is modems, not apps processors' I would say is generally true but probably not last quarter where I suspect the Atom revenue outnumbered the 3G revenue. However as the XMM7260 modem is in a future Galaxy S5 premium variant the modem revenue is about to stage a recovery.
http://www.sammobile.com/2014/04/22/exynos-chip-with-intel-lte-2k-display-support-confirmed-could-replace-snapdragon-cpus-on-samsung-devices/
OpenPower is a much bigger threat to DCG than any ARM server effort especially if Google is interested it in volume. Power 8 will lose very few benchmarks if any.
Whole article was a load of meandering waffle which bears little relation to reality.
Manufacturing on the same process node also levels out the competitive field for the various CPU architectures
Not really. Design and ISA still dominate.
At the 14nm process node Intel’s lead is being reduced to just a quarter or two.
LOL, in the foundries dreams. Has anybody even seen a foundry 20nm part yet in the wild ? 14FF will be at least a year behind that and it will be less denser than Intel's.
I meant the initial Windows version of it i.e. Cherry Trail.
Braswell might also affect competitors android sales too especially for those customers where gpu performance is important. Basically it's the first opportunity for Intel's 14nm transistors to shine in mobile along with fanless Broadwells. The important point about that this year though is what will the volume availability for Q4 holiday season be as Intel has been missing volume of its new premier products in Q4s lately.
or just plain continued difficulties with drivers/system software to work with Intel's platforms?
Probably this although Merrifield was late too and Moorefield will not be long after. Maybe the phone OEMs are waiting for the 7260 too to pair it with for maximum impact and longevity. Obviously Intel's android mobile efforts are still a WIP so the next big driver on the mobile side this year is going to be Braswell(Cherry Trail) which will neutralize the gpu reason for buying an AMD low power chip leaving just price.
we still have no announcements of Merrifield based designs. What do you think is going on here?
All of the Silvermont Android designs are late which would seem to point to a software bottleneck.
If any Android OEM was seriously bothered about 64-bit Intel could have released a 64-bit version of Clovertrail+ for them as Atom has been 64-bit from birth with the 32-bit versions only fused off versions. So yet more alarmist nonsense from you. If any OEM is seriously bothered about 64-bit phones which I doubt Merrifield and Moorefield will be available in 2014 for them well before your A57 Snapdragon.
What do you mean by "endurance"?
How long a mobile device lasts in x hours/days in typical usage profiles
and the fact that Intel is putting 16 EUs into CHT should be a big clue that Intel's customers are demanding gaming performance as well.
That was not put in to fight ARM tablet chips but AMD low power laptop chips which have been traditionally strong on gpu against Atom. Gaming is more important for a traditional mobile PC user than a typical media consumption tablet user.
If gpu was so important to tablets Clover Trail would not have single-handedly stopped Windows RT dead in its tracks with its feeble one. You are getting confused as to what the real customers want with the benchmarketing games mobile OEMs indulge in the absence of real differentiation.
An A57 SoC will be more like a hot power hungry FX than Moorefield.
The only reason I care about this stuff is because the OEMs probably do. Notice how phone vendors will cheat on benchmarks just to "win"? That's how important differentiating yourself on performance in these devices is.
It's just games they play in the absence of real differentiation. Phones don't sell on performance they sell on design, features and endurance. Who seriously cares about the absolute performance of a phone chip apart from you and some small group of geek buyers ?
You tried to tell me earlier that gaming was all so important to tablet buyers and the best you could do for evidence was find a survey which ranked it third after internet and e-mail lol. I am sure hard-core PC gamers are concerned about internet and e-mail usage too ... not .
Intel is crazy if it thinks it's going to successfully fight a Snapdragon 808/810 on 20nm with Moorefield.
In bench-marketing and all the other features that seem to excite you perhaps not but in real world usage it will continue the fine reputation for performance allied to endurance x86 phone chips have quickly established.