The Intel Atom Thread

[coercitiv]

You forgot that those tablets have 3G or 4G which increases consumption and not to mention higher quality lower latency higher resolution screens which consume considerably more than screen used in that tablet.

If you have put screen on the same level and a 3G or 4G chip then battery life would drop like a rock.

Are you suggesting the modems in the Nexus and iPad are using a significant amount of power even though they're not being used? Turn off the 3G modem on a modern phone, you might be surprised.

I would put the difference in both load & idle times on the OLED screen. At minimum brightness it's probably significantly more efficient over the IPS backlight, while at load we compare 231.1cd/m² for the OLED versus 430 and 357cd/m² for the Nexus and iPad respectively.

However, in the WiFi test all device are set to 150cd/m² and the Dell might have the less efficient screen in this scenario, so it's not like the Intel SoC was given a free pass with this device.

 

[witeken]

A lot of stars are aligning to potentially challenge the top Broadwell Core M SKU. The only saving grace for Intel will probably be Skylake, and even here I'm not as confident as I used to be. I can see why there is a fuss about Apple abandoning Intel for lower spec'd devices, though I don't necessarily agree.

No stars are aligning. Apple has both a process and manufacturing deficiency, and as you say, it will have to compete against Skylake. Also not sure why SKL would be a saving grace. It's of course great if Intel has the fastest SoC on the market, but it just has to be more compelling than other Android SoCs (although Core M serves another class of devices anyway...), unless Apple wants to switch to Intel, which isn't going to happen any time soon.

 

[jdubs03]

No stars are aligning. Apple has both a process and manufacturing deficiency, and as you say, it will have to compete against Skylake. Also not sure why SKL would be a saving grace. It's of course great if Intel has the fastest SoC on the market, but it just has to be more compelling than other Android SoCs (although Core M serves another class of devices anyway...), unless Apple wants to switch to Intel, which isn't going to happen any time soon.

i guess we'll be finding out soon enough. granted intel's direct competition isn't apple while qualcomm, nvidia, and samsung are for SoC's, but there is an indirect effect.

i think you're not giving apple enough credit here. why wouldn't apple be a challenge to intel at the same power envelope? though intel is ahead of samsung and tsmc in process, apple has already proved a potent chip especially compared to atom (nvidia too). there will be a confluence of events that are going to be very beneficial for apple. we may not see the performance uptick like from the a6 to a7, but i surmise it could be close. plus with iOS 9 being more about performance and stability, i see no reason why apple couldn't be a very fast chip.

let's just assume the a72 core is the base providing a 20% increase in IPC, then either a new uarch/enhanced+ cyclone (doubtful, the enhanced+). we could see at least a 30-40% increase on the same node. along with 14nnFF enabling higher frequencies, i see no reason why the a9/x couldn't be >=50% faster per core than the a8/x. using geekbench 3 as a reference; a9 could be ~2400+ single-core|a9x could be ~2700+. not too far off from BDW-5Y71. multi-core a9x scores could be 6500+.

doesn't seem all that radical.

 

[III-V]

i see no reason why the a9/x couldn't be >=50% faster per core than the a8/x.

I don't see how it could. It's already a tremendous core -- it has a larger L1D and L1I cache than Haswell, the same number of ALUs, same size reorder buffer, a larger L2...

Anything Apple could do that would significantly increase its performance per clock would be way too expensive. In fact, their pipeline is a bit on the shorter side by a touch, and if anything, they'd lose IPC if they wanted higher clocks.

Apple skipped the usual iterative stuff and went straight into diminished return land.

There are perhaps a few things they could do, like implement SMT, or add a uop cache (if they haven't already), but they are basically stuck in the same mud Intel is.

 

[jdubs03]

I don't see how it could. It's already a tremendous core -- it has a larger L1D and L1I cache than Haswell, the same number of ALUs, same size reorder buffer, a larger L2...

Anything Apple could do that would significantly increase its performance per clock would be way too expensive. In fact, their pipeline is a bit on the shorter side by a touch, and if anything, they'd lose IPC if they wanted higher clocks.

Apple skipped the usual iterative stuff and went straight into diminished return land.

what's your prediction? you maybe right and it won't be such a big change, maybe a72 is the last in line of uarchs that catch all of the low hanging fruit. but i do think apple will make a statement with the a9.

 

[III-V]

what's your prediction? you maybe right and it won't be such a big change, maybe a72 is the last in line of uarchs that catch all of the low hanging fruit. but i do think apple will make a statement with the a9.

Well Apple is much further ahead than ARM is, in terms of having a "fat" core. ARM's still only a 3-wide design, while Apple's Cyclone is 6-wide. They had very different design goals in mind -- the A57 is supposedly small enough to be able to have four fit inside a single enhanced Cyclone core.

I expect ARM to eventually let go of its extreme pursuit of keeping costs down with their "big" cores, and let the A17 and its eventual successors take over with balancing performance and cost.

The A72 will likely be pretty similar to the A8, perhaps a bit narrower of a design.

 

[lopri]

With regard to the Dell tablet - Please test it out in person instead of believing benchmarks posted around the web. After playing with it for half an hour, I am pretty sure there is DFVS (or is it DVFS? Forgot the order) manipulation for select benchmarks to "ensure" the SOC is running full steam during the benchmarks. The tablet just does not perform as well as Galaxys and iPads.

I thought it'd be better. It looked like the most compelling offering from Intel to date, so while I had enough dose of skepticism before seeing it person I wanted to be surprised. Quality tablets are becoming hard to find due to race to the bottom and I want more competition in mid-to-high end tablets, so if Intel wants to provide quality tablets at cost, I am all for it. I do not care about corporations losing money. (although I wonder what the legality of the aforementioned practice is)

Alas, the tablet did not perform like a high-end tablet as its price tag might want you to believe. Stutters and lags galore, and I experienced crashes twice in that 30 minutes. I suspect a very aggressive throttling is at work to keep power/temp in check in the tablet's 6 mm thick body. I have seen Atoms perform better than that.

2560x1600 ambition obviously does not help in that regard. And Dell somehow managed to make an AMOLED screen look like a TN panel. I did not know AMOLED screen is capable of looking like that.

Frequency-controlled benchmarks will give not just a misleading but an opposite performance impression under such circumstance. (that notebookcheck.com review is a joke, btw)

There are a lot other shortcomings I noticed, not directly related to the SOC. Such as ergonomics, speaker/mic locations, camera locations (does anyone like herself/himself to be pictured from chin up?), unbelievably slow GPS,.. You get the idea. Even if the tablet was riding on a Cyclone or a Krait, I would call the whole package a novel experiment at best. As it is, the tablet is a dud.

 

[Shivansps]

With regard to the Dell tablet - Please test it out in person instead of believing benchmarks posted around the web. After playing with it for half an hour, I am pretty sure there is DFVS (or is it DVFS? Forgot the order) manipulation for select benchmarks to "ensure" the SOC is running full steam during the benchmarks. The tablet just does not perform as well as Galaxys and iPads.

I thought it'd be better. It looked like the most compelling offering from Intel to date, so while I had enough dose of skepticism before seeing it person I wanted to be surprised. Quality tablets are becoming hard to find due to race to the bottom and I want more competition in mid-to-high end tablets, so if Intel wants to provide quality tablets at cost, I am all for it. I do not care about corporations losing money. (although I wonder what the legality of the aforementioned practice is)

Alas, the tablet did not perform like a high-end tablet as its price tag might want you to believe. Stutters and lags galore, and I experienced crashes twice in that 30 minutes. I suspect a very aggressive throttling is at work to keep power/temp in check in the tablet's 6 mm thick body. I have seen Atoms perform better than that.

2560x1600 ambition obviously does not help in that regard. And Dell somehow managed to make an AMOLED screen look like a TN panel. I did not know AMOLED screen is capable of looking like that.

Frequency-controlled benchmarks will give not just a misleading but an opposite performance impression under such circumstance. (that notebookcheck.com review is a joke, btw)

There are a lot other shortcomings I noticed, not directly related to the SOC. Such as ergonomics, speaker/mic locations, camera locations (does anyone like herself/himself to be pictured from chin up?), unbelievably slow GPS,.. You get the idea. Even if the tablet was riding on a Cyclone or a Krait, I would call the whole package a novel experiment at best. As it is, the tablet is a dud.

There is problably something going on with the modems, powervr and souch. BT-T on tablets never goes below the base freq unless its at idle, its the cpu turbo that dont last very long if all cores are loaded and/or the igp is in use.

And i dont think how thick the body is its gona matter, these atoms heatsink is a aluminium sheet that is about 0.5mm.

To me its just a crappy android build, specially if you said it crashes.

 

[Nothingness]

There is problably something going on with the modems, powervr and souch. BT-T on tablets never goes below the base freq unless its at idle, its the cpu turbo that dont last very long if all cores are loaded and/or the igp is in use.

As far as I know Power VR works very well in all SoC and I'm confident that it's more power efficient than the Intel GPU (or else why would Intel have chosen it for its phone SoC?), so I'm not sure one could put the blame on it.

To me its just a crappy android build, specially if you said it crashes.

That'd be surprising as the basis most likely comes from Intel Android team which has done a lot of work on the x86 porting. Or perhaps Dell added a bad layer on it...

 

[witeken]

As far as I know Power VR works very well in all SoC and I'm confident that it's more power efficient than the Intel GPU (or else why would Intel have chosen it for its phone SoC?), so I'm not sure one could put the blame on it.

Because Merrifield and Moorefield are evolutions of the Medfield platform.

 

[Nothingness]

Because Merrifield and Moorefield are evolutions of the Medfield platform.

It'd have been less expensive and far easier for Intel to use the same IP for all of its Silvermont-based SoC, no matter what they did in the past (the SoC are very different from previous generations and even the PowerVR chips is completely different). So the choice of IMG tech was made on purpose, and power efficiency is most likely one of the reasons.

 

[dahorns]

With regard to the Dell tablet - Please test it out in person instead of believing benchmarks posted around the web. After playing with it for half an hour, I am pretty sure there is DFVS (or is it DVFS? Forgot the order) manipulation for select benchmarks to "ensure" the SOC is running full steam during the benchmarks. The tablet just does not perform as well as Galaxys and iPads.

I thought it'd be better. It looked like the most compelling offering from Intel to date, so while I had enough dose of skepticism before seeing it person I wanted to be surprised. Quality tablets are becoming hard to find due to race to the bottom and I want more competition in mid-to-high end tablets, so if Intel wants to provide quality tablets at cost, I am all for it. I do not care about corporations losing money. (although I wonder what the legality of the aforementioned practice is)

Alas, the tablet did not perform like a high-end tablet as its price tag might want you to believe. Stutters and lags galore, and I experienced crashes twice in that 30 minutes. I suspect a very aggressive throttling is at work to keep power/temp in check in the tablet's 6 mm thick body. I have seen Atoms perform better than that.

2560x1600 ambition obviously does not help in that regard. And Dell somehow managed to make an AMOLED screen look like a TN panel. I did not know AMOLED screen is capable of looking like that.

Frequency-controlled benchmarks will give not just a misleading but an opposite performance impression under such circumstance. (that notebookcheck.com review is a joke, btw)

There are a lot other shortcomings I noticed, not directly related to the SOC. Such as ergonomics, speaker/mic locations, camera locations (does anyone like herself/himself to be pictured from chin up?), unbelievably slow GPS,.. You get the idea. Even if the tablet was riding on a Cyclone or a Krait, I would call the whole package a novel experiment at best. As it is, the tablet is a dud.

I agree that on the ergonomic side there are problems. I completely disagree on the performance side. I didn't see any problem running apps, games, or browsing. I agree (as most reviews have said) there is some stutter in the android menu system. Given how everything else runs, I'm assuming that is a driver/build issue. It doesn't seem to affect performance anywhere else.

And, you act like the benchmarks are out of the norm for the newer atoms. Most the benchmarks line up as a nice iterative increase on Bay-Trail. It isn't like this thing is benching at A8 levels, so I'm not sure how you are concluding that benchmark cheating is occurring.

Also, if you were playing with a store display model, you probably should have reset the tablet.

Edit: Finally, are you really complaining about the display? If there is one thing all the reviews have agreed on is that the screen is really good.

 

[Arachnotronic]

With regard to the Dell tablet - Please test it out in person instead of believing benchmarks posted around the web. After playing with it for half an hour, I am pretty sure there is DFVS (or is it DVFS? Forgot the order) manipulation for select benchmarks to "ensure" the SOC is running full steam during the benchmarks. The tablet just does not perform as well as Galaxys and iPads.

I thought it'd be better. It looked like the most compelling offering from Intel to date, so while I had enough dose of skepticism before seeing it person I wanted to be surprised. Quality tablets are becoming hard to find due to race to the bottom and I want more competition in mid-to-high end tablets, so if Intel wants to provide quality tablets at cost, I am all for it. I do not care about corporations losing money. (although I wonder what the legality of the aforementioned practice is)

Alas, the tablet did not perform like a high-end tablet as its price tag might want you to believe. Stutters and lags galore, and I experienced crashes twice in that 30 minutes. I suspect a very aggressive throttling is at work to keep power/temp in check in the tablet's 6 mm thick body. I have seen Atoms perform better than that.

2560x1600 ambition obviously does not help in that regard. And Dell somehow managed to make an AMOLED screen look like a TN panel. I did not know AMOLED screen is capable of looking like that.

Frequency-controlled benchmarks will give not just a misleading but an opposite performance impression under such circumstance. (that notebookcheck.com review is a joke, btw)

There are a lot other shortcomings I noticed, not directly related to the SOC. Such as ergonomics, speaker/mic locations, camera locations (does anyone like herself/himself to be pictured from chin up?), unbelievably slow GPS,.. You get the idea. Even if the tablet was riding on a Cyclone or a Krait, I would call the whole package a novel experiment at best. As it is, the tablet is a dud.

As somebody who purchased a Venue 8 7000, I completely agree. The tablet "looks" nice and checks a lot of interesting boxes, but it's really not pleasant to use. It also is pretty laggy.

A "gimmicky rip-off" is probably the best way to describe it. I'll be eBaying mine soon...

 

[Sweepr]

Another Moorefield tablet review, this time based on the 1.83GHz Atom Z3560. Looks like a spiritual succesor to the ASUS Nexus 7 (2013) tablet. Once again excelent longevity in the battery tests. Compared to the Snapdragon S4 Pro-based Nexus 7 2013 which packs the same battery (3950mAh) you get 3 extra hours of web browsing and video playback - better CPU performance and more than twice the GPU performance is another bonus. 2015 is already shaping up to be a better year for Intel's mobile offerings.

Asus Memo Pad ME572C review: A pad to remember

Key test findings:

Build quality is solid
The display is superb across the board
The hardware delivers excellent performance
Battery life is great
Audio quality is good
Camera is a major letdown when it comes to image quality
ZenUI adds features to stock Android without being too heavy
Built-in video player comes with very limited codec support

The battery in the Memo Pad 7 ME572C is the same 3950mAh unit as in the ME176C model, which has less than half the pixels of the currently reviewed unit. That said, the ME176C posted great scores in our dedicated battery test.

The Memo Pad 7 ME572C showed excellent longevity in the web browsing routine with just shy of 12 hours, actually beating the ME176C by close to an hour. Video playback turned out more taxing on the battery, but 10 whole hours is a very respectable score nonetheless.

The Asus Memo Pad 7 ME572C is pretty heavily armed for a brand name midrange tablet, costing only €200. The company has been employing Intel silicon for some time and the powerful Atom Z3560 was chosen to power the ME572C. The chipset is manufactured using a 22nm process for improved performance and power efficiency over the majority of Qualcomm and Mediatek offerings. Its four cores are clocked at 1.83MHz, but Intel's architecture allows for better per clock performance, which should be on par with the Snapdragon 800. The GPU is an import, namely the PowerVR G6430 found in Apple's iPhone 5S, where it performed admirably.

http://www.gsmarena.com/asus_memo_pad_7_me572c-review-1205.php

Moorefield would be a good fit for a $199 7-8'' Nexus 7 successor. It fully supports 64-bit Lollipop and smokes Snadragon 615 performance-wise in most benchmarks.

Dell launches Bay Trail Chromebook 11, Latitude 11 for school use

Dell has also redesigned the case of the Chromebook 11 and introduced a Windows version of the laptop. That model is called the Latitude 11 Education Series. Both laptops are designed for students and educators, but they’ll be available for anyone to buy. The laptops feature Intel Celeron N2840 dual-core processors, 2GB to 4GB of RAM, 16GB of storage, and 11.6 inch displays with optional touchscreens.

Since these laptops are designed for use int he classroom, they have rugged features including Gorilla Glass screens, a MIL-STD 810G tested case for pressure, temperature, shock, and vibration resistance, and a sealed keyboard and touchpad for spill protection. [...]

Dell’s new Chromebook 11 launches today for $250 and up, while the Latitude 11 Education Series with Windows 8.1 will launch on March 3rd if you want a model without a touchscreen.

Starting with the Chromebook 11, the device features a Celeron N2840 processor, anti-glare 1366x768 LCD with optional touch, and 2GB RAM/16GB eMMC storage (and there’s an upgraded model with 4GB RAM). It has a sealed keyboard and touchpad, and as noted above a 180-degree durable hinge so it can lay flat. Hinges have been a major point of failure for Chromebooks in education, and Dell says they specifically went back to redesign and strengthen this aspect so that if, for example, a student picks up the laptop by the screen, it won’t break.

...The Latitude 11 Education (3160) at first looks a lot like the Chromebook 11 but with Windows instead of Chrome OS, but there are actually a few other points of differentiation. Yes, it does run Windows 8.1 or Windows 7 (or an education SKU) and it’s launching on Bay Trail N2840. However, Dell will be moving to Braswell later in June/July, and the Latitude 3160 includes a Gigabit Ethernet port, true SSD (starting at 64GB), and it also has upgradeable RAM.

http://anandtech.com/show/8974/dell...k-11-latitude-11-and-venue-1010-pro-offerings
http://liliputing.com/2015/02/dell-launches-bay-trail-chromebook-11-latitude-11-school-use.html

AnandTech's Toshiba Chromebook 2 Review: A Feast for the Eyes

Besides the inclusion of a great display – more on that in a moment – the remaining specs are pretty typical of Chromebooks. The processor of choice this time is Intel’s Celeron N2840, a dual-core Bay Trail chip running at 2.16-2.58GHz. The GPU portion of the chip is based on Intel’s HD Graphics architecture, the same architecture in Ivy Bridge and Haswell processors, but with only 4 EUs active.

...The CPU of the N2840 is able to surpass most ARM processors (at least, those that we’ve tested) in terms of performance. The problem is that the GPU is quite a bit slower than the competition. Take NVIDIA’s Tegra K1 SoC, which pairs one of the fastest SoC GPUs with a respectable ARM-based CPU; by contrast, the N2840’s CPU is generally faster than the Tegra K1’s CPU, but the GPU ends up being substantially slower – less than half the performance by our numbers. For a browser-centric OS like Chrome OS, I’m not sure it matters all that much, but as we discussed earlier there are cases where graphics performance still matters – like YouTube 1080p60 content.

...In light of the lack of a clear winner in all areas, what we’re left with is a decision of where to compromise. Among the 13” Chromebooks the Toshiba Chromebook 2 (specifically, the CB35-B3340 model) is currently my favorite option, and it’s really all about the screen. If you appreciate a good display it’s practically the only game in town for Chromebooks right now, with the 15” 1080p Acer model due in the near future.

http://anandtech.com/show/8961/toshiba-chromebook-2-a-feast-for-the-eyes

 

[dahorns]

Another Cherry Trail Geekbench score:

http://browser.primatelabs.com/geekbench3/1898242

990 Single, 3451 Multi - running GB 3.1.5 pro, Windows 64-bit.

 

[Sweepr]

The best Bay Trail-T multi-core scores are around 3100 so that's actually pretty good. I'm fine with 10-20% better CPU performance and ~2x GPU performance, especially if they can further reduce power consumption.

Seeing 64-bit is refreshing too, we need more 4GB Windows tablets.

 

[Arachnotronic]

The best Bay Trail-T multi-core scores are around 3100 so that's actually pretty good. I'm fine with 10-20% better CPU performance and ~2x GPU performance, especially if they can further reduce power consumption.

If GB3 is representative of real CPU performance, then this is disappointing.

Look at the Exynos 5433's performance...

 

[dahorns]

If GB3 is representative of real CPU performance, then this is disappointing.

Look at the Exynos 5433's performance...

Yeah, if this is a final product it is very disappointing. There doesn't appear to be any real gain from Bay Trail.

http://browser.primatelabs.com/geekbench3/432517

 

[Nothingness]

The best Bay Trail score for Windows 64-bit seems to be ST 1028 / MT 3322

http://browser.primatelabs.com/geekbench3/1203078

So 990/3451 would be very disappointing.

 

[Sweepr]

The best Bay Trail score for Windows 64-bit seems to be ST 1028 / MT 3322

http://browser.primatelabs.com/geekbench3/1203078

So 990/3451 would be very disappointing.

To be fair that Lenovo submission is far ahead of the rest of the top 25. From 3rd to 25th place the multi-core scores vary from 3140-3195. I'm sure that's not the best Cherry Trail-T submission we will see, there's not a single retail device out there yet and at the same stage BT-T scores were (probably) quite a bit lower.

I'm not convinced GB is the best way to compare x86 and ARM CPUs either but that's another topic.

 

[Nothingness]

To be fair that Lenovo submission is far ahead of the rest of the top 25. From 3rd to 25th place the multi-core scores vary from 3140-3195.

You were talking about the highest score, I picked the highest I found :biggrin:

I'm sure that's not the best Cherry Trail-T submission we will see, there's not a single retail device out there yet and at the same stage BT-T scores were (probably) quite a bit lower.

Agreed.

I'm not convinced GB is the best way to compare x86 and ARM CPUs either but that's another topic.

And I'm convinced GB is definitely interesting to compare x86 and ARM, but it certainly shouldn't be the only benchmark to look at

 

[kpkp]

Another Moorefield tablet review, this time based on the 1.83GHz Atom Z3560. Looks like a spiritual succesor to the ASUS Nexus 7 (2013) tablet. Once again excelent longevity in the battery tests. Compared to the Snapdragon 600-based Nexus 7 2013 which packs the same battery (3950mAh) you get 3 extra hours of web browsing and video playback - better CPU performance and more than twice the GPU performance is another bonus. 2015 is already shaping up to be a better year for Intel's mobile offerings.

I am not sure if they still use the 50% brightness method in the battery tests as they used to, in case they do, the battery gains have more to do with the display then the SoC, as it's evident from the review itself:

Brightness on white at 50%:
Display test | Black cd/m2 | White cd/m2
Asus Memo Pad 7 ME572C | 0.15 | 155
Asus Google Nexus 7 | 0.25 | 244

 

[mikk]

The Pentium N3700 is 14nm quad core clocked at 1.6GHz, with the ability to jump all the way to 2.4GHz. The Bay Trail-M based Pentium N3540 is clocked at 2.16GHz base clock and 2.66GHz Turbo.
As you can see, Braswell 14nm has a significantly lower clock and Turbo at a 6W TDP (Thermal Design Power) and 4W Scenario Design Power (SDP).

http://www.fudzilla.com/news/notebooks/37040-mobile-braswell-pentium-coming-in-q3-15


That doesn't bode well if this is the fastest Cherry View-M SKU. And also delayed into Q3.

 

[III-V]

http://www.fudzilla.com/news/notebooks/37040-mobile-braswell-pentium-coming-in-q3-15


That doesn't bode well if this is the fastest Cherry View-M SKU. And also delayed into Q3.

There's no way that's the fastest model... even smartphones hit higher clocks.

 

[IntelUser2000]

http://www.fudzilla.com/news/notebooks/37040-mobile-braswell-pentium-coming-in-q3-15


That doesn't bode well if this is the fastest Cherry View-M SKU. And also delayed into Q3.

Basically this is what's happening. When Intel went from Nehalem-based Lynnfield to a much more power efficient Sandy Bridge, they increased base clocks quite a bit. Low base clocks mean that its inherent power profiles are probably not very good.

7.5W on the N3540 to 6W on the N3700 is only a 25% difference, so it shouldn't result in dropping base clocks that much. Process alone should have done more than enough to go from 7.5W to 6W without dropping ANY clocks.

Now early Cherry Trail submissions show that performance isn't that good. It's faster in MT, but almost identical in ST. It probably has 5-10% better IPC but they had to clock that down, so max clock scenarios like ST is similar. It's perhaps possible MT does slightly better because it can sustain frequency a bit better.

Extremely disappointing performance on Intel's 14nm process.

2015 is already shaping up to be a better year for Intel's mobile offerings.

Highly doubt it. Last time Intel did really good was in 2011, and it was down every year until 2014, which they did good. Now they had good basis for doing so. While Haswell wasn't as good as expected, it was a good improvement for Laptops. And even though financially and realistically mobile is a disaster, most do not think so, based on that hopefully they'll turn the losses around. What does 2015 bring? Disappointing Core M and Atom CT chips, and delays and delays.