The Intel Atom Thread

[liahos1]

http://gizmodo.com/dell-venue-8-7000-a-terrible-name-for-an-incredible-ta-1680459751

http://www.computerworld.com/articl...a-distinctive-and-premium-android-tablet.html

http://www.wsj.com/articles/dell-venue-8-7000-review-a-thin-tablet-with-real-depth-1420567257

the reviews have been very good. the battery life seems amazing for a thin 8 inch tablet. I wonder if this is all due to using oled vs normal display. Seems like this is a big stepup vs first introduction baytrail tablets last year.

on a side note

http://www.wsj.com/articles/dell-xps-13-review-the-macbook-air-meets-its-match-1421785380

http://www.wsj.com/articles/the-two-day-laptop-battery-is-here-1421779300?mod=e2tw&autologin=y

battery life for the broadwell based laptops also looks very good. 12 hours on the xps 13.

 

[witeken]

Do you mean Intel will use the exact same process and design rules for Cherry as they do for Broadwell? If not, the Broadwell comparison is pointless.

I guarantee you that process does have some effect on frequency

No, it's called common sense. Intel has been designing and optimizing its process nodes for literally more than a decade with high performance and clock speeds in mind, and idem for its architecture. At any given transistor size, Intel has had a vastly, vastly superior transistor than any other company, and without all the architectural optimizations, Core wouldn't clock nearly as high.

Since the architecture (Airmont) is about the same and the process node is only an improvement (0.8x shorter gate, air gaps), it's nothing but logical to assume that there will be an improvement. There are already 2.6GHz Bay Trails, so 2.7 on a vastly superior node is a short stretch.

It seems people think repeating their unsupported ideas many times will somehow make it true or people will believe it's true. But that which can be asserted without evidence, can be dismissed without evidence as well.

 

[Nothingness]

No, it's called common sense. Intel has been designing and optimizing its process nodes for literally more than a decade with high performance and clock speeds in mind, and idem for its architecture. At any given transistor size, Intel has had a vastly, vastly superior transistor than any other company, and without all the architectural optimizations, Core wouldn't clock nearly as high.

Since the architecture (Airmont) is about the same and the process node is only an improvement (0.8x shorter gate, air gaps), it's nothing but logical to assume that there will be an improvement. There are already 2.6GHz Bay Trails, so 2.7 on a vastly superior node is a short stretch.

I have never discussed that. I was just noting you were saying things as if they were known for sure, and that made you sound like you were working for Intel.

It seems people think repeating their unsupported ideas many times will somehow make it true or people will believe it's true.

Where did I do that exactly? I'm asking legitimate questions, as you quite often make claims without proofs except for Intel slides, or their marketing speak.

But that which can be asserted without evidence, can be dismissed without evidence as well.

Definitely, evidence is paramount. Now show us Airmont frequency. If you can't, I'll dismiss your claim, does that sound fair?

 

[witeken]

Sure.

 

[Arachnotronic]

Intel says Cherry Trail devices will roll out in 1H 2015, so it won't be too long before we see devices with production Cherry Trail chips. Should be fun to see how it performs.

 

[Nothingness]

Sure.

This is not what I consider as evidence, this is an Intel slide. I will wait for consumer devices, though I'd be very surprised if Intel couldn't reach what they claim here, after all a 10% max clock increase is not very aggressive (I guess sustained frequency will be better at iso power).

 

[liahos1]

does anyone think intel is building the so-called HPU powering the hololens? What exactly is an HPU?

 

[dahorns]

does anyone think intel is building the so-called HPU powering the hololens? What exactly is an HPU?

I thought I read somewhere that a much smaller specialty company was responsible for the HPU. Doesn't mean Intel isn't providing foundry services, but I don't gather they are responsible for design (also, we'd have heard that from Intel).

 

[III-V]

Intel says Cherry Trail devices will roll out in 1H 2015, so it won't be too long before we see devices with production Cherry Trail chips. Should be fun to see how it performs.

I read somewhere that we'd see them in April.

Oh, it was this thread, lol

does anyone think intel is building the so-called HPU powering the hololens? What exactly is an HPU?

Holographic processing unit.

I thought I read somewhere that a much smaller specialty company was responsible for the HPU. Doesn't mean Intel isn't providing foundry services, but I don't gather they are responsible for design (also, we'd have heard that from Intel).

The device that Microsoft was showing off used Cherry Trail.

 

[dahorns]

I read somewhere that we'd see them in April.

Oh, it was this thread, lol

Holographic processing unit.

The device that Microsoft was showing off used Cherry Trail.

That is the rumor, yes. But, the device was specifically stated to have (1) a CPU, (2) GPU, and (3) HPU.

(1) and (2) are Cherry Trail (per rumor). I don't think (3) is, nor do I think it is a separate design from Intel.

 

[III-V]

That is the rumor, yes. But, the device was specifically stated to have (1) a CPU, (2) GPU, and (3) HPU.

(1) and (2) are Cherry Trail (per rumor). I don't think (3) is, nor do I think it is a separate design from Intel.

Oh, I must have misunderstood. Didn't look too long at the article.

 

[TechFan1]

I haven't seen anything about the hpu, but a random commenter in that article I linked suggested it was made by Himax. Far from a reliable source.

"The HPU and displays are provided by Himax Technologies. Here's the Himax IR Factsheet. MSFT is listed under ASIC 3D customers. Himax is a leading provider for LCoS displays and is working with LUMUS, ODG, Optinvent, MSFT, and Google Glass. Soon to be released new Google Glass design v2 should be interesting!


http://www.himax.com.tw/en/investor/HIMX_FS.pdf"

 

[ninaholic37]

It seems people think repeating their unsupported ideas many times will somehow make it true or people will believe it's true. But that which can be asserted without evidence, can be dismissed without evidence as well.

Doesn't Cherry Trail-T have to be produced cheaper than Bay Trail-T to avoid contra revenue? Will moving the process to 14nm alone be enough savings to fill the gap?

 

[III-V]

Doesn't Cherry Trail-T have to be produced cheaper than Bay Trail-T to avoid contra revenue? Will moving the process to 14nm alone be enough savings to fill the gap?

Nope. Cherry Trail is very cost focused though (or at least way more than Bay Trail).

 

[Idontcare]

Doesn't Cherry Trail-T have to be produced cheaper than Bay Trail-T to avoid contra revenue? Will moving the process to 14nm alone be enough savings to fill the gap?

Contra revenue is not about Intel's production costs for the chips they are selling (be it Cherry Trail or Bay Trail)...contra revenue is about the BOM cost delta that OEMs experience when building products with the Intel chips (extra stuff the OEM has to buy in order to create the product versus creating a product with an ARM SoC that has higher level of integration).

Cherry Trail may or may not be less costly for Intel to produce, but that is besides the point when it comes to contra-revenue. What Cherry Trail has to do is enable a comparable (or lower) BOM for the OEMs by way of not requiring the OEMs to purchase more expensive components or additional components above and beyond what they would otherwise expend were they to make a comparable product based on an ARM SoC.

Now of course Intel would like to accomplish all of the above, have better margins on Cherry Trail than Bay Trail AND simultaneously integrate enough into the SoC so as to enable an elimination of the BOM delta that the OEMs experience when building products based on Cherry Trail versus an ARM SoC.

 

[Idontcare]

I have never discussed that. I was just noting you were saying things as if they were known for sure, and that made you sound like you were working for Intel.

That is a pretty low bar if that is all it takes for one to assume another individual is an employee of a given company.

I know things about Intel, factually concrete things, that are not public domain. And yet I am most certainly not an employee.

And I am not unique in this regard, there are plenty of knowledgeable forum members here who are privy to information which hasn't officially entered the public domain yet. Why discount the information? You gain nothing by dismissing it, you stand to learn something if you are open-minded to the possibility that the information is legit.

 

[III-V]

Abwx likens contra revenue to Intel wrapping each Bay Trail package with a $20 bill... which isn't all too inaccurate. Intel has stated Bay Trail's BoM delta -- that is, the additional cost involved with implementing Bay Trail instead of an ARM solution -- is ">$15." The SoC itself is cost competitive... it's the additional hardware you need to implement it (or rather, the additional hardware Intel says you need to implement it. Their standards are/were too high). AMD's in the same boat... however people around here make it seem like AMD would have sold millions of SoCs if Intel hadn't implemented their contra revenue program. The truth is, neither of them would have sold anything worth writing home about.

The subsidized part in question, Bay Trail-T, now has a lower BoM. Intel's already managed to reduce the component count from ">850" down to "~600" with discrete VR, and "~550" with an X-powers AXP228 power management IC. That's a 30% and 35% reduction, respectively. There's no doubt most of those components that are no longer required are practically free (cheap surface mount resistors and capacitors), but that's a few less dollars that Intel has to cough up with every Bay Trail-T they sell (for OEMs that implement it).

Contra revenue is not about Intel's production costs for the chips they are selling (be it Cherry Trail or Bay Trail)...contra revenue is about the BOM cost delta that OEMs experience when building products with the Intel chips (extra stuff the OEM has to buy in order to create the product versus creating a product with an ARM SoC that has higher level of integration).

Well, it would certainly help if Bay Trail were free. If there were the case, contra revenue wouldn't be needed, since the total cost to implement would be much lower... you'd just have to pay for the power supply circuitry and whatnot.

 

[Shivansps]

The Power Management IC is even needed? the EC of my BT tablet is not working since day 1 because of a failed firmware update and other than no battery indicator there are 0 issues.

 

[III-V]

The Power Management IC is even needed? the EC of my BT tablet is not working since day 1 because of a failed firmware update and other than no battery indicator there are 0 issues.

It integrates a number of other power delivery circuits.

 

[Nothingness]

That is a pretty low bar if that is all it takes for one to assume another individual is an employee of a given company.

I know things about Intel, factually concrete things, that are not public domain. And yet I am most certainly not an employee.

And I am not unique in this regard, there are plenty of knowledgeable forum members here who are privy to information which hasn't officially entered the public domain yet. Why discount the information? You gain nothing by dismissing it, you stand to learn something if you are open-minded to the possibility that the information is legit.

My original post had a smiley

FWIW I also have and share various insider information (including Intel and TI for which I worked for a few years in the DSP division), and I typically assume that it's not because someone works for a given company that he's a fanboy of that company. OTOH just using Intel slides to prove something is a bit too much, you have to be cautious when using that information and IMHO witeken is sometimes blindly trusting it.

 

[coercitiv]

Abwx likens contra revenue to Intel wrapping each Bay Trail package with a $20 bill... which isn't all too inaccurate. Intel has stated Bay Trail's BoM delta -- that is, the additional cost involved with implementing Bay Trail instead of an ARM solution -- is ">$15." The SoC itself is cost competitive... it's the additional hardware you need to implement it (or rather, the additional hardware Intel says you need to implement it. Their standards are/were too high).

There's a swarm of very low cost tablets with Intel Inside hitting the stores, it's a race to the bottom with prices of under 100$ a piece (Europe taxes included). I wasn't joking when I said in another thread that I bought a BT tablet just to use it as a photo album: it's that cheap, and to be honest it also works great.

As an outside observer I really can't tell if they're still in the "establish foothold in new market" phase or they have enough momentum from CR to go for the jugular.

Meanwhile, I really like the new Dell Venue 8 tablet, it may be my next.

 

[Sweepr]

Lenovo Tab S8 Tablet Review (NotebookCheck)

Light as a feather. Lenovo's Tab S8 does not even weigh 300 grams but is quite a bargain. It offers a Full HD screen, an 8-megapixel camera, and front-facing speakers for 199 Euros (~$225). Discover whether Lenovo's tight calculation is also worthwhile for the customer in our review.

...Lenovo's Tab S8 sports an IPS screen with a resolution of 1920x1200 pixels, which is slightly higher than Full HD. A higher resolution is only found in the considerably more expensive comparison devices. The equally priced LG G Pad 8.0 can only offer 1280x800 pixels.

...Intel's Atom Z3745 is a favorably used SoC with four cores and a clock rate of 1.33 - 1.86 GHz. The user is thus well-equipped for the majority of apps in the Android field. Although the processor slightly lags behind the considerably more expensive devices in Geekbench 3 (multi-core), it should be tolerable considering the price difference of up to 100%. The performance of the equally expensive LG G Pad is clearly surpassed.

The Intel HD Graphics (Bay Trail) can even beat the graphics unit in Samsung's Galaxy Tab S 8.4 - an ARM Mali-T628 MP6, in many benchmarks. The graphics power should thus at least come close to that of the pricier comparison devices. Lenovo's Tab S8 is sometimes even faster than its rivals when browsing via Chrome.

...The graphics unit easily copes with graphically sophisticated games like Asphalt 8 even in high settings and without losses in frame rate. Controlling via position sensor and touchscreen are also beyond reproach. The part about temperatures is quickly closed in Lenovo's Tab S8. A temperature increase is actually not at all noticed in idle mode, and it remains very low with a maximum of 33.8 °C on the back even during load. Although some waste heat is felt, it would not really be enough to warm the hands with it.

The minimum idle consumption of Lenovo's Tab S8 is a relatively high 1.9 watts. Sony's Xperia Z3 Tablet Compact is happy with just 0.6, for example. The other idle rates do not provide any top values with 3.3 and 3.7 watts. Our review sample is also clearly undercut from the higher-performance Sony Xperia Z3 Tablet Compact during load. Otherwise, it is more frugal than its other rivals. 7.2 watts is the absolute maximum consumption. Lenovo's Tab S8 is definitely fit for use with over 10 hours of Wi-Fi browsing time.

...Buyers who do not need the prestige of an iPad Mini 3 or the coolness of a Nexus 9 will find Lenovo's offer attractive. The Tab S8 sports a sharp Full HD screen, decent camera equipment, relatively rich-sounding speakers, high performance, and feasible battery runtimes. It is, however, surprising that Lenovo's Tab S8 clearly outperforms the considerably more expensive models from Apple and Google in some categories. Lenovo's tablet has most definitely earned a performance-for-money title and a clear purchase recommendation.

+10 hours Wi-Fi with a meager 16.3 watt-hour battery (5.3W idle/load delta for the whole device), which means it lasts longer than iPad mini 3 (23.8 watt-hour) and Nexus 9 (25 watt-hour) with a smaller battery. Bay Trail is about to be replaced but it still packs a punch in performance/$ and power efficiency, Lenovo did a good job here.

Ps: The high idle power consumption number is probably wrong, otherwise it wouldn't last 24 hours @ idle like it did on the battery test.

http://www.notebookcheck.net/Lenovo-Tab-S8-Tablet-Review.135139.0.html

Windows PC stick + touchscreen monitor = All-in-one desktop PC

The Hannspree Micro PC is a pocket-sized Windows computer that you can plug into the HDMI port on a display to run Windows apps on a TV or monitor. The recently-launched PC-on-a-stick is available from a handful of stores for around $200 and up. But UK-based retailer The Stick PC Store is offering an interesting bundle: For a little over $500 you can get the Hannspree Micro PC plus a touchscreen monitor and a wireless keyboard and mouse.

...The Hannspree Micro PC features an Intel Atom Z3735F Bay Trail processor, 2GB of RAM, and 32GB of storage, plus a microSD card slot, 802.11b/g/n WiFi, Bluetooth 4.0, and USB and micro USB ports. The included display is a 23 inch, 1920 x 1080 pixel HannsG touchscreen monitor with 10-point multitouch input and stereo 1.5W speakers. Also included are the Cherry DW300 wireless mouse and keyboard set which comes with a USB receiver that you can connect to the Micro PC.

http://liliputing.com/2015/01/windows-pc-stick-touchscreen-monitor-one-desktop-pc.html

Cheap and easy to carry PCs-on-a-stick have a lot of potential. Bay Trail enabled some interesting form factors for x86 Windows PCs.

http://gizmodo.com/dell-venue-8-7000-a-terrible-name-for-an-incredible-ta-1680459751

http://www.computerworld.com/articl...a-distinctive-and-premium-android-tablet.html

http://www.wsj.com/articles/dell-venue-8-7000-review-a-thin-tablet-with-real-depth-1420567257

the reviews have been very good. the battery life seems amazing for a thin 8 inch tablet. I wonder if this is all due to using oled vs normal display. Seems like this is a big stepup vs first introduction baytrail tablets last year.

Lenovo Tab S8 is another example, quite a bit better than the first Bay Trail tablets. Time to bring some some Cherry Trail tablets with great build quality and 1080p IPS/OLED screens for a reasonable price now.

 

[Abwx]

Lenovo Tab S8 Tablet Review (NotebookCheck)

The minimum idle consumption of Lenovo's Tab S8 is a relatively high 1.9 watts.

+10 hours Wi-Fi with a meager 16.3 watt-hour battery

Yes, 10 hours when surfing, with a 16.3Wh battery and 1.9W comsumption at idle, i guess that once you browse comsumption get down from 1.9W to 1.6W, first time i see a negative power delta from idle to light loading, must be the famous perf/watt methodology i talked about in another thread, or more simply Notebookcheck forgot to unplug the main...

 

[dahorns]

Yes, 10 hours when surfing, with a 16.3Wh battery and 1.9W comsumption at idle, i guess that once you browse comsumption get down from 1.9W to 1.6W, first time i see a negative power delta from idle to light loading, must be the famous perf/watt methodology i talked about in another thread, or more simply Notebookcheck forgot to unplug the main...

Something went wrong with their testing of this device. There are plenty of consumer complaints about 3-4 hours of battery life while surfing, which seems reasonable given the small battery. Maybe the other consumers were are all running at 100% brightness, I don't know.

 

[mikk]

http://browser.primatelabs.com/geekbench3/1815592

Here is a first Cherry Trail-T Geekbench submission, and for the first time with SKU name. It's called Z8700.

http://browser.primatelabs.com/geekbench3/1764699

Here another one from Z8700.