The Intel Atom Thread

[Nothingness]

Its on Amazon too... http://www.amazon.com/Toshiba-Excite.../dp/B00KF47USG i could pay it with gift cupons.

only if could install windows on it :S

I guess the 8GB flash drive would be the blocking point on top of the firmware.

 

[Shivansps]

nah, the no HDMI thing is the worst.

 

[Sweepr]

Intel and Unity Collaborate to Extend Android Support across Intel-based Devices

SANTA CLARA, Calif. and SEATTLE, Aug. 20, 2014 – Intel Corporation and Unity Technologies today announced a strategic collaboration to advance the development of Android*-based applications on Intel® architecture. The agreement accelerates Intel's mobility push as millions of developers using the Unity development platform can now bring native Android games and other apps to Intel-based mobile devices. Unity adds support for Android across all of Intel's current and future processors including both the Intel® Core™ and Intel® Atom™ processor families.

Unity will ensure Intel product enhancements, including both graphics and CPU performance improvements and features, will be seamlessly integrated into future releases of the Unity 4 and Unity 5 product lines. As Intel architecture continues to gain market segment share on mobile devices, these improvements will help ensure that the Unity developers' games run natively as well as look great and perform beautifully on Intel platforms.

In addition, developers using Unity can now easily add support for Intel architecture in their applications or produce native applications for Intel architecture only with minimal extra effort.

"We've set a goal to ship 40 million Intel-based tablets this year and expect more than 100 Android tablet designs on Intel in the market by the end of this year," said Doug Fisher, Intel corporate vice president and general manager of the Software and Services Group. "Our collaboration with Unity will give its nearly 3 million developers the necessary software tools and support to build amazing Android experiences on Intel architecture."

"Unity is used by half of all mobile game developers, and many of them have been asking for increased support for Intel-based devices running Android," said David Helgason, CEO, Unity Technologies. "We are proud to be working with Intel to ensure that Unity provides the smoothest and highest performing experience possible on Intel platforms.

"As a mobile gaming company, Kabam relies on the Unity game engine and the compelling performance and efficiency it provides us to publish our mobile games for players around the world," said Kent Wakeford, COO of Kabam. "We are very excited to bring Unity-authored content, such as our upcoming title, 'Marvel Contest of Champions,' to the rapidly growing installed base of Intel-powered Android devices."

http://newsroom.intel.com/community...nd-android-support-across-intel-based-devices

Unity powers some very popular Android games like Temple Run, Dead Trigger, Angry Birds Epic and others.

 

[witeken]

It seems the x86 instruction set is becoming less of an issue every month.

How is it with the availability or announcement of Moorefield and Merrifield devices? Any smartphones?

You'd think that Moorefield is a very good competitor for the Snapdragon 800 series, but everyone keeps using Qualcomm. Not a good sign.

The delay of Cherry Trail/14nm was yet another nail in the coffin. Broxton and SoFIA are better going to have some major IPC, BOM, graphics, TTM, price and integration improvements, else we'll have to wait for 10nm until Intel can gain some major market share.

 

[liahos1]

It seems the x86 instruction set is becoming less of an issue every month.

How is it with the availability or announcement of Moorefield and Merrifield devices? Any smartphones?

You'd think that Moorefield is a very good competitor for the Snapdragon 800 series, but everyone keeps using Qualcomm. Not a good sign.

The delay of Cherry Trail/14nm was yet another nail in the coffin. Broxton and SoFIA are better going to have some major IPC, BOM, graphics, TTM, price and integration improvements, else we'll have to wait for 10nm until Intel can gain some major market share.

the most negative ive ever heard you sound on intel. i do agree though you'd think with all the company's resources and mfg advantages they'd hvae started making inroads in the smartphone side this year. i dont understand why they arent building any of their modems on their own fabs yet.

 

[witeken]

the most negative ive ever heard you sound on intel.

Because, like Anand said, this is extremely frustrating. I honestly don't even see a reason why for instance Samsung would pick the Qualcomm S805 with a small frequency bump over Intel's Moorefield in their next Galaxy Note, but that won't happen.

i do agree though you'd think with all the company's resources and mfg advantages they'd hvae started making inroads in the smartphone side this year.

They had to do this a lot earlier. Brian Krzanich fortunately realizes this and now he must do a lot of work to make up for Paul Otellini's wrong decision to ignores Moore's Law.

i dont understand why they arent building any of their modems on their own fabs yet.

They're still porting it to their own 14nm process, but that also seems to take longer than expected, probably only finished in 2016 or late 2015.

 

[jdubs03]

The thing is its still too early to proclaim Cherry Trail as dead, its supposed to be around EOY, and it will be a significant over Bay Trail, especially in graphics so even though it may be frustrating it is still coming and it'll be around before S810, with likely higher GPU performance.

 

[frozentundra123456]

I see their main problem as lack of an integrated modem, and even then not making it themselves. How hard can it be for a company like Intel to make a modem on 14nm or even an old 22nm process? I think they can compete well in tablets, and of course laptops, etc., but they are way, way behind the curve in phones, and shooting at a moving target.

 

[ShintaiDK]

I see their main problem as lack of an integrated modem, and even then not making it themselves. How hard can it be for a company like Intel to make a modem on 14nm or even an old 22nm process? I think they can compete well in tablets, and of course laptops, etc., but they are way, way behind the curve in phones, and shooting at a moving target.

Its not the node. Its the R&D and/or IP that Qualcomm holds.

 

[witeken]

The thing is its still too early to proclaim Cherry Trail as dead, its supposed to be around EOY, and it will be a significant over Bay Trail, especially in graphics so even though it may be frustrating it is still coming and it'll be around before S810, with likely higher GPU performance.

If you want to gain market share, you can't afford 6 month delays. At the start of the year, things looked good, with CT in Q3 and WT in Q4, with Merrifield and Moorefield capable of some nice phone design wins, and both Broxton and SoFIA in 2015.

 

[Roland00Address]

Guys you do not want a 1024x600 screen in windows. It sucked 5 years ago with netbooks, and it would suck even more with tablets and trying to hit very small lines in desktop windows with your fingers.

You need 720p with 768 or 800p or higher.

Because, like Anand said, this is extremely frustrating. I honestly don't even see a reason why for instance Samsung would pick the Qualcomm S805 with a small frequency bump over Intel's Moorefield in their next Galaxy Note, but that won't happen.

Execution problems where a company has good technology but it just takes too long to get to market are the worse. Effectively Intel in mobile is like AMD today with normal computer chips. Part of that is all the foundries are just so behind intel with the new nodes and nodes do matter, but it also has to do with how much Global Foundry sucks at node execution. The best recent designs from AMD have been TMSC related, and they been doing great with the TMSC video cards.

I see their main problem as lack of an integrated modem, and even then not making it themselves. How hard can it be for a company like Intel to make a modem on 14nm or even an old 22nm process? I think they can compete well in tablets, and of course laptops, etc., but they are way, way behind the curve in phones, and shooting at a moving target.

Its the IP Qualcomm holds, and the fact it is much harder to make "analog" chips on the smaller node processes. The analog nature of the chip means it is harder to miniaturize and you can't use as many design tools to help you with the translation.

To my understanding the LTE modems that intel currently produce are not made at intel fabs but at TMSC 28nm fabs. And the design is going to be (or maybe it already has) been ported to Globalfoundries 28nm fabs. (I am not a 100% sure on this so I will appreciate a confirmation.) To my understanding no intel made modem is on intel's 14nm or 22 nm process yet, but the eventual endpoint is a 28nm cheap SOFIA at TMSC and the nice high end chips at 14nm Intel Fabs.

 

[Arachnotronic]

It seems the x86 instruction set is becoming less of an issue every month.

How is it with the availability or announcement of Moorefield and Merrifield devices? Any smartphones?

You'd think that Moorefield is a very good competitor for the Snapdragon 800 series, but everyone keeps using Qualcomm. Not a good sign.

The delay of Cherry Trail/14nm was yet another nail in the coffin. Broxton and SoFIA are better going to have some major IPC, BOM, graphics, TTM, price and integration improvements, else we'll have to wait for 10nm until Intel can gain some major market share.

Moorefield doesn't have an integrated modem, XMM 7260 was late, it has a fairly weak image signal processor, and its performance isn't anything to write home about relative to Snappy 805 (faster CPU, slower GPU).

Intel needs a clean, clear-cut, no-marketing-nonsense winner that shows up on time before any smartphone OEM is going to risk moving to Intel.

 

[Arachnotronic]

The thing is its still too early to proclaim Cherry Trail as dead, its supposed to be around EOY, and it will be a significant over Bay Trail, especially in graphics so even though it may be frustrating it is still coming and it'll be around before S810, with likely higher GPU performance.

Here's something to chew on, though...Intel was quick to talk up Broxton at its investor meeting.

This, to me, signals that Cherry Trail is deficient in some very serious way that Intel wants to move past as quickly as possible. My guess is that the CPU performance was targeted too low (i.e. 5-10% IPC from Silvermont + modest clock bump), no H.265 encode/decode block, weak image signal processor (a common problem for Intel SoCs), and so on.

Braswell will do fine in the traditional PC market, but in the smartphone SoC market, it's a whole different ball-game and Intel needs to figure out how to play it better.

 

[Sweepr]

Moorefield doesn't have an integrated modem, XMM 7260 was late, it has a fairly weak image signal processor, and its performance isn't anything to write home about relative to Snappy 801 (faster CPU, slower GPU).

Sure about this? It's packing a PowerVR G6430 running at up to 533MHz, that should make it slightly faster than S801 or at least equal (given how A7 with its 450MHz PowerVR G6430 holds up against S800/S801 in graphics benchmarks).

http://www.cnx-software.com/wp-content/uploads/2014/06/Intel_MooreField_Snapdragon_801.jpg

 

[Arachnotronic]

Sure about this? It's packing a PowerVR G6430 running at up to 533MHz, that should make it slightly faster than S801 or at least equal (given how A7 with its 450MHz PowerVR G6430 holds up against S800/S801 in graphics benchmarks).

http://www.cnx-software.com/wp-content/uploads/2014/06/Intel_MooreField_Snapdragon_801.jpg

My bad, Sweepr! I meant to say Snapdragon 805. Thank you for correcting me!

 

[ninaholic37]

Guys you do not want a 1024x600 screen in windows. It sucked 5 years ago with netbooks, and it would suck even more with tablets and trying to hit very small lines in desktop windows with your fingers.

You need 720p with 768 or 800p or higher.

Wouldn't a lower resolution make things look bigger? I remember when I switched from a 1024x600 netbook to a 1280x720 one, everything on the 1280x720 one looked tiny! I ended up having to use 125% fonts and 128% zoom in Firefox so it looked as readable as my 1024x600 one, but just got so annoyed with having to unnaturally zoom everything against it's native resolution that I sold it for $70 and kept my 1024x600 one. Modern internet still looks great with 1024 width. :thumbsup:

 

[jdubs03]

Here's something to chew on, though...Intel was quick to talk up Broxton at its investor meeting.

This, to me, signals that Cherry Trail is deficient in some very serious way that Intel wants to move past as quickly as possible. My guess is that the CPU performance was targeted too low (i.e. 5-10% IPC from Silvermont + modest clock bump), no H.265 encode/decode block, weak image signal processor (a common problem for Intel SoCs), and so on.

Braswell will do fine in the traditional PC market, but in the smartphone SoC market, it's a whole different ball-game and Intel needs to figure out how to play it better.

You may have a point, but from based on from what we've seen with leaked CT slides, it is still a considerable improvement over BT. The encoding/decoding block could be right. I mean with Intel their CPU performance must be at least solid or its a disappointment, if there is only a 5-10% IPC increase then yeah that would be a considerable issue with only an extra 300 MHz. That still would be enough to beat the S805, but yeah it would only be in a leadership position for a quarter, but we'll see. What we do know though is that GPU wise Tegra K1 and Intel Gen 8 will most probably outperform Adreno 430 (+30% over S805-Adreno 420). TK1 is around 40% faster than Adreno 420.

IDF can't come soon enough, we'll be in for some nice details; hopefully more about Broxton too.

 

[witeken]

Cherry Trail is a tick, so I don't expect too many CPU improvements, and a 12% clock speed increase isn't going to be enough, although the power consumption will be top-notch. The GPU side should be very solid, but I it will compete against Adreno 430, so the performance delta won't be very big.

 

[jdubs03]

Cherry Trail is a tick, so I don't expect too many CPU improvements, and a 12% clock speed increase isn't going to be enough, although the power consumption will be top-notch. The GPU side should be very solid, but I it will compete against Adreno 430, so the performance delta won't be very big.

If there is an IPC increase of 10%, plus that 12% increase, we still get about +24% in overall performance (I know its over-simplistic, but still illustrates my point). So it won't be a slouch. But it would make sense to be a placeholder until Broxton ups the game; perhaps a 6 month cycle?

 

[teejee]

If there is an IPC increase of 10%, plus that 12% increase, we still get about +24% in overall performance (I know its over-simplistic, but still illustrates my point). So it won't be a slouch. But it would make sense to be a placeholder until Broxton ups the game; perhaps a 6 month cycle?

Intel will have a big product planning issue if cherry trail improves 24%, it will come too close to much higher margin core ULV CPU's. They might do it because of competition though.

 

[witeken]

Intel's 1.86GHz Bay Trail SoC is about on par with Sandragon 801. If CT is 30% faster than BT, that will make up for the 33% increase in performance/watt that 20nm is supposed to bring. If A57 is as "good" (= power hungry) as A15, lower peak performance but much lower power should be a good trade-off.

 

[jdubs03]

Intel will have a big product planning issue if cherry trail improves 24%, it will come too close to much higher margin core ULV CPU's. They might do it because of competition though.

For single-thread, CT would be significantly less than Broadwell-Y at 30%. And if we get 30% higher performance I'd be satisfied. It wouldn't even be half of Broadwell-U. Now multi-thread it may be close to Broadwell-Y, but still it would probably benchmark better and feel faster.

With Atom, they can't afford to be out designed with a +1.5 node lead, especially considering CPUs are supposed to be their expertise (even though I think Nvidia has a nice lead with TK1 (quad ARMv7 and dual Denver ARMv8), Qualcomm not as much)...so that will pressure them to make some decent gains in preparation for a big Broxton jump. They're going to have to keep up with the performance improvements irregardless of Y-series, as they use the Atom uArch in Xeon Phi (significantly enhanced) and through dense-servers. But now that Broadwell-Y is down to 4.5W, -> Skylake-Y @ 4W, they can focus on IPC there as well. The motivation to continue lowering power consumption should lessen.

 

[dealcorn]

Here's something to chew on, though...Intel was quick to talk up Broxton at its investor meeting.

This, to me, signals that Cherry Trail is deficient in some very serious way that Intel wants to move past as quickly as possible.

That interpretation is plausible, but there are alternatives. Brian may appreciate that Intel has fallen into the "once burned twice shy" trap. By repeatedly setting expectations too high and then not delivering mobile market share, Intel lost credibility. Brian has not yet set a high bar for CT, but share accretion looks certain. CT should sport class leading graphics efficiency and graphics performance adequate to satisfy the bottom 90% of the target market. Other than perhaps DCC, most or all the Broadwell M graphics good stuff gets mirrored in CT. Again looking at the bottom 90% of the market, CT's CPU will sport class leading efficiency and with a 5% IPC gain and something on the frequency side, CT should be OK for the bottom 80% of the target market. Obviously, the words "target market" provide wiggle room where you can exclude almost any potential competitor based on SDP. However, in tablet world, SDP is king.

Bay Trail-T was a marketing disaster. I believe BT-T met it's design specifications but there was no demand for the specified design at the original price. It makes sense to me that Brian fired much or most of the Marketing Department: the BT-T failure was a marketing boo boo rather than a technology failure. Yes, BT-T's graphics were not adequate for the top of the market so it was wrong to target BT-T for the upper end. However, despite the fiasco, BT-T delivered every important strategic objective. Unity will support X86 because with contra revenue, BT-T delivered share and credibility. BT-T developed a supply chain familiar with X86 and more than willing to ramp fast if demand is adequate. The nonsense about X86 not being efficient has been effectively rebutted. As a fiasco, BT-T was overpriced, but still not too shabby. For the middle market, CT appears primed to gain significant market share. The most glaring weakness is the lack of a credible n-1 CPU to target the bottom of the tablet market. CT's target range is too large because of the BT-T boo boo. We already know Apple is off the table for CT. Do not get distracted by the upper end niche CT will not address. If CT captures the bottom half, it gives Intel breathing room to sort out the other half over time.

 

[Roland00Address]

For single-thread, CT would be significantly less than Broadwell-Y at 30%. And if we get 30% higher performance I'd be satisfied. It wouldn't even be half of Broadwell-U. Now multi-thread it may be close to Broadwell-Y, but still it would probably benchmark better and feel faster.

Exactly, there will still be a placed for i5s or i7s. Less so for the pentium broadwells, but I don't think Intel will be wanting to push those skus (they have not been announced yet to my knowledge) and preferring to push atoms and the core Ms.

For comparisons of single threaded performance this info is from anandtech bench and anandtech reviews. Cinebench 11.5 single thread.

• 0.15 Atom N550 (pineview Atom the best 45nm atom, clovertrail was just a 32nm die shrink with a small increase in clock speed.)
• 0.31 Atom Z3740
• 0.32 AMD E-350 (1.6 ghz no turbo bobcat)
• 0.39 AMD A4-5000 (1.5 ghz no turbo jaguar, mullins does not increase much on ipc it just has better power consumption and thus better dynamic range.)
• 0.40 Atom Z3770
• 0.68 i3-4020Y (1.5 ghz, note this size chip in theory can be fanless), MS Surface Pro 3 i3 version
• 0.96 i5-4200u (1.6 to 2.6 ghz, Surface Pro 2)
• 1.17 i5-4300u (1.9 to 2.9 ghz, Surface Pro 3, 11% higher max ghz but the score went up 22% compared to the surface pro 2, so it looks like it is a better bin chip and thus can stay at the highest turbo longer)

So a speculated 30% increase of 0.31 (middle atom) and 0.40 (high end atom) would be 0.40 (middle atom) and 0.52 (high end atom) respectively. But do not forget that the i3s will be going a little faster and the i5s will be able to hit those speeds without using a fan.

Now most of the i5 haswell Y series are not using the cTDP down so they can be fanless. One sku you do see in the market is the i5 4202y which is 1.6 ghz to 2.0 ghz. There are faster Ys out there going all the way up to 2.3 ghz turbo but I have not seen any fanless devices with the 2.3 ghz model.

0.88 for the fanless i5 4202y according to notebookcheck 1.85 for the multithreaded score due to 2 cores and hyperthreading for the fanless tablet. Multithreading is not as impressive for comparison the atom z3770 gets 1.48, but remember that we are talking a chip that already gets the desired tdp before the die shrink. Both the atom and the core i5 will get the predicted 30% increase in energy efficiency before the die shrink.

 

[Sweepr]

Intel’s next-gen NUC desktop PCs coming in 2015 (Broadwell & Braswell) - Leaked Roadmaps

Q2-2015 launch for Braswell NUC, ready for BTS season. ''Light mainstream gaming'' and ''Solid CPU/GPU performance'' suggests Intel is confident about Braswell's gaming potential (16 EUs iGPU - Gen 8).

www.fanlesstech.com/2014/08/exclusive-intel-readying-nuc-20.html

Acer Chromebook 11 First to Use Bay Trail Processor

Chromebook sales started out slow, but are starting to grow at a fast rate – they are light, uncomplicated, generally cheap and improvements and more exposure are making them more accepted. Several manufacturers have jumped on board – Toshiba and LG have a couple models, HP makes a couple, Samsung is a major contributor as well, but nobody makes as many as Acer.

Acer is adding to its Chromebook line with its newest model, the Acer Chromebook 11 – it showed up in a brochure in July at a German event and now they have finally posted a promotional video (see below) which promises to be a great value. I know what you are thinking…another 11.6-inch Chromebook, big deal – but what makes this new addition such a big deal is that it is their first one with an Intel Celeron Bay Trail processor. It joins Acer’s existing 720 Chromebook series, which feature a Celeron 2955U or Core i3 Haswell chips and a recently launched Chromebook 13 which uses a NVIDIA Tegra K1 processor. [...]

www.androidheadlines.com/2014/08/acer-chromebook-11-first-use-bay-trail-processor.html

Asus Transformer Pad TF303 tablets with FHD screen hits Europe

Most of the new Android tablets Asus unveiled this summer are entry-level models with Intel Atom Bay Trail processors and 1280 x 800 pixel displays. But if you prefer tablets with full HD screens, Asus has you covered.

The company’s new Transformer Pad TF303 line of Android tablets feature 10.1 inch, 1920 x 1080 pixel displays. They’re not yet available in the US, but Mobile Geeks reports that at least two versions of the Transformer Pad TF303 should be available in Europe this September for 299 Euros and up.

The Asus Transformer Pad tablets are defined by the optional keyboard docks that let you use the devices like notebooks. I recently reviewed the Transformer Pad TF103 and found the tablet to be nice, but I didn’t find the keyboard all that useful.

As for the new models, the Asus Transformer Pad TF303CL features an Intel Atom Z3745 processor, 2GB of RAM, 16GB to 32GB of storage, and a 1920 x 1200 pixel display. It’s basically what you’d get if you took the TF103 and added more memory and a higher-resolution display. [...]

http://liliputing.com/2014/08/asus-transformer-pad-tf303-tablets-with-fhd-screen-hits-europe.html