The Intel Atom Thread

[tempestglen]

Actual Bay Trail-T score @ SunSpider 1.0: 329.6ms

http://www.youtube.com/watch?v=UXCobb134Vo

That makes (up to 2.4GHz) Bay Trail-T >20% faster than 1.3GHz A7 in the 3 Javascript/HTML5 single-thread benchmarks tested by Anand. If they can run Merrifield above 2GHz coupled with a similar PowerVR Rogue GPU (+competitive power levels) then Intel will offer an interesting A7 alternative to Android phones.

Sun spider under IE usually outperforms much better than under android/ios.

So withdraw your 329ms, that's meaningless.

 

[BallaTheFeared]

So withdraw your 329ms, that's meaningless.

And by association the rest of the graph is as well.

 

[SPBHM]

Sun spider under IE usually outperforms much better than under android/ios.

So withdraw your 329ms, that's meaningless.

the A7 is not running Android like the others, and Bay Trail can run Windows, while the other can't, if Windows is better optimized and give better results, that's another argument in favor of bay trail.

edit: just for fun I tried sunspider 1.0 on my ancient Pentium e2140 at 2GHz and got 230ms windows 7/IE1, but Kraken took 6500ms

 

[mavere]

Bay Trail-T score @ WebXPRT: 574 +/- 8

On my 2GHz mobile Penryn Macbook, the latest Chrome 30 beta is 17.5% faster than the latest Safari 7.0 beta in WebXPRT. If one makes a (big) assumption that the relative gap here can be reproduced in hypothetical full-blown Chrome on an A7 chip, the corresponding score in WebXPRT would be 604.

Interestingly, on the 3 common Javascript tests (Sunspider, Kraken, Octane), Safari on this laptop reliably has ~25% better results than Anand's iPhone 5S Safari numbers. That jumps to over 40% on WebXPRT, which makes me wonder exactly why Intel likes that test so much, like Anand claimed.

Not that this matters in terms of Intel's push into the marketplace... The bulk of Apple's customers don't care about specs and would just stick with familiar branding. The A7, then, is just a look into what Qualcomm/Samsung/Nvidia might have 6-12 months from now, which doesn't make for a fair comparison to Bay Trail.

 

[liahos1]

Sun spider under IE usually outperforms much better than under android/ios.

So withdraw your 329ms, that's meaningless.

why is it meaningless. if sunscript is just an exercise in single threaded performance in a most optimized environment then its fair game. you should throw out safari in ios by that logic.

 

[thunng8]

Actual Bay Trail-T score @ SunSpider 1.0: 329.6ms

http://www.youtube.com/watch?v=UXCobb134Vo

That makes (up to 2.4GHz) Bay Trail-T >20% faster than 1.3GHz A7 in the 3 Javascript/HTML5 single-thread benchmarks tested by Anand. If they can run Merrifield above 2GHz coupled with a similar PowerVR Rogue GPU (+competitive power levels) then Intel will offer an interesting A7 alternative to Android phones.

IE10 performs very well in Sunspider, but has has very bad results for Kraken and Google Octane.

http://www.pcmag.com/article2/0,2817,2407701,00.asp

So if Anand chose to use IE10 for the web test benchmarks, then Bay Trail will gain in one benchmark but lose out badly on the other 2.

Anyway - I think all the numbers are for Android phones/tablets (except for the iOS results), so that's why he chose Chrome as ie10 is not available on Android

 

[liahos1]

IE10 performs very well in Sunspider, but has has very bad results for Kraken and Google Octane.

http://www.pcmag.com/article2/0,2817,2407701,00.asp

So if Anand chose to use IE10 for the web test benchmarks, then Bay Trail will gain in one benchmark but lose out badly on the other 2.

Anyway - I think all the numbers are for Android phones/tablets (except for the iOS results), so that's why he chose Chrome as ie10 is not available on Android

i wonder if oems looks at these benches when decided what system to build (taking price out of the equation). Does anyone know how oems actually decide which soc's to work on?

 

[Khato]

Look at the AT battery life tests. There is almost no difference when battery capacities are normalized between the 5 and the 5s.

Unfortunately the battery life tests tell us very little of actual interest except how good the SoC is at idling and how well the rest of the platform was put together... and chances are pretty good that Haswell in the Macbook Air does better in that regard. Really, battery life tests under light workloads are only half of the equation, and no one seems interested in testing the other half when it comes to mobile devices.

As for the GLBenchmark 2.5.1 battery life - it's an interesting idea to provide comparative numbers... but those numbers don't really tell anything. Of course a more powerful GPU running at lower frequency/voltage is going to be more efficient. The best comparison would be to just derive total power draw for each device running a benchmark at its maximum performance and present those results alongside the benchmarks. Similar to how the Server benchmarks had been done previously. Oh well, should get that data from notebookcheck whenever they get their review up.

 

[Enigmoid]

Unfortunately the battery life tests tell us very little of actual interest except how good the SoC is at idling and how well the rest of the platform was put together... and chances are pretty good that Haswell in the Macbook Air does better in that regard. Really, battery life tests under light workloads are only half of the equation, and no one seems interested in testing the other half when it comes to mobile devices.

As for the GLBenchmark 2.5.1 battery life - it's an interesting idea to provide comparative numbers... but those numbers don't really tell anything. Of course a more powerful GPU running at lower frequency/voltage is going to be more efficient. The best comparison would be to just derive total power draw for each device running a benchmark at its maximum performance and present those results alongside the benchmarks. Similar to how the Server benchmarks had been done previously. Oh well, should get that data from notebookcheck whenever they get their review up.

Light workloads are something like 99.9% of what the device is used for. I'm just saying that even with A7 we did not see or saw a slight regression in idle/basic use.

I don't particularity care about power under intense loads because on a phone that will rarely come up. But power under common conditions is very important.

 

[Eug]

Yeah, I generally don't transcode full-length movies or batch CFD calculations on my cell phone.

I guess I do play the odd game though.

For my laptop my choice for SSD was based on idle power. I got a high idle power SSD and it actually made a small but noticeable impact on battery life. It was one that had a higher rated power at idle than most 2.5" platter drives actually, and it was meant more for desktops. I returned it and got a super low power sipping SSD, and my battery life increased.

 

[Nothingness]

IE10 performs very well in Sunspider, but has has very bad results for Kraken and Google Octane.

http://www.pcmag.com/article2/0,2817,2407701,00.asp

So if Anand chose to use IE10 for the web test benchmarks, then Bay Trail will gain in one benchmark but lose out badly on the other 2.

Anyway - I think all the numbers are for Android phones/tablets (except for the iOS results), so that's why he chose Chrome as ie10 is not available on Android

Nah he should have picked IE for Bay Trail too so that WebXPRT would have cratered to 370: http://www.hardwarezone.com.sg/feat...-benchmarks-intels-latest-bay-trail-processor

 

[Khato]

Light workloads are something like 99.9% of what the device is used for. I'm just saying that even with A7 we did not see or saw a slight regression in idle/basic use.

I don't particularity care about power under intense loads because on a phone that will rarely come up. But power under common conditions is very important.

Agree completely with the above. The battery life benchmarks that are performed by the majority of reviews are definitely what actually matter from the practical perspective... but they still only give half the picture from the technical perspective Does it matter that the A7's performance might come at the cost of decreased battery life if you actually use it? Not really, because that's a very atypical use case. But for discussing the merits/performance per watt of the architecture it does.

Sadly all we can do is speculate as to how Silvermont based products will perform in light workload battery life... but isn't it pretty safe to guess that it's going to be at least as good as Haswell ULT in that regard? And we already know how Haswell ULT does in comparison to the A6X equipped iPad 4... So as long as Baytrail and Merrifield don't completely drop the ball on light workload battery life, there's a pretty decent chance that they'll offer both better performance and battery life than what Apple has to offer.

 

[Nothingness]

Agree completely with the above. The battery life benchmarks that are performed by the majority of reviews are definitely what actually matter from the practical perspective... but they still only give half the picture from the technical perspective Does it matter that the A7's performance might come at the cost of decreased battery life if you actually use it? Not really, because that's a very atypical use case. But for discussing the merits/performance per watt of the architecture it does.

Sadly all we can do is speculate as to how Silvermont based products will perform in light workload battery life... but isn't it pretty safe to guess that it's going to be at least as good as Haswell ULT in that regard? And we already know how Haswell ULT does in comparison to the A6X equipped iPad 4... So as long as Baytrail and Merrifield don't completely drop the ball on light workload battery life, there's a pretty decent chance that they'll offer both better performance and battery life than what Apple has to offer.

I might have missed it, but I don't think we saw any battery life test for Bay Trail, be it light or heavy. We just got instantaneous power consumption figures.

 

[Sweepr]

HP announces Bay Trail powered Pavilion 11 X2 2-in-1 and Omni 10 tablet

HP Omni10

First, let’s talk tablet. If you’re looking for a tablet-first device that will handle your everyday tasks (and play games) with ease, you need to check this 10-inch tab out. Since it’s all about the screen, let’s start there: A 10-inch screen that’ll give you 1920 by 1200 pixel resolution. With Windows 8.1 and the Intel Atom Z3000-series CPU and 2GB of RAM under the hood, I hear that this will run 9 hours on a single charge. So it’ll last you through the day. Speaking of which….I know that we aren’t billing this as a business machine, but it also happens to come loaded with MS Office 2013. The full monty. No trialware nonsense.

HP Pavilion11 x2

The most basic way I can think of describing this new Pavilion model is a little reductive, but it gets right to the point: It’s a hard-plastic-shelled version of what you saw in the O.G. ENVY x2 (that had an 11.6-inch screen as well), but this time out, it gets a lot more zip under the hood thanks to Bay Trail.

http://h20435.www2.hp.com/t5/The-Ne...-detachables-hit-Intel-s-Bay-Trail/ba-p/83005

 

[Imouto]

$600

Thanks but no, thanks.

 

[Sweepr]

There is a $350 ASUS tablet convertible with pretty similar specs (except for the screen).

Intel expects the systems to start at $199 for a clamshell device, $250 for a notebook with touch and $349 for a 2 in 1 device.

+140 design wins for Bay Trail, I bet some of them are those <$300 Bay Trail-M touch notebooks waiting Windows 8.1 release to be launched.

You could also buy a 13.3'' $600 Toshiba tablet convertible powered by the A4 1200 (1GHz dual-core Jaguar, less than a 1/3 of Z3770's MT CPU performance) and a 1366x768 screen. Thanks but no, thanks.

http://techreport.com/news/25318/temash-apu-pops-up-in-toshiba-convertible-tablet

 

[Khato]

I might have missed it, but I don't think we saw any battery life test for Bay Trail, be it light or heavy. We just got instantaneous power consumption figures.

Correct, hence why I prefaced such with a clear label of it all being speculation. There's always the possibility that Baytrail-T will have worse battery life than Haswell ULT, but it's highly doubtful considering that the SoC was purpose built for this particular market. And we already know that Haswell ULT shows better battery life efficiency under light workloads than anything currently available in the tablet world.

 

[Nothingness]

Correct, hence why I prefaced such with a clear label of it all being speculation. There's always the possibility that Baytrail-T will have worse battery life than Haswell ULT, but it's highly doubtful considering that the SoC was purpose built for this particular market. And we already know that Haswell ULT shows better battery life efficiency under light workloads than anything currently available in the tablet world.

I definitely agree.

But my concern is that if OEM go for low cost they also might choose components that are not power efficient. That's why since the beginning I'm claiming we should all wait for consumer devices tests, the Intel tablets surely was correctly built and tuned

 

[Shivansps]

I rather buy the Dell Inspiron 11 3000 with the Celeron 2955U at $379 and forget about the tablet thing.

 

[bullzz]

what I dont understand is how are baytrail tablets suddenly coming down to 349 and 299 price points when clovertrail ones was $450. I expect the chip cost to be similar. windows cud have cut the price down by say $30. and the 8 inch displays may cut the cost down by $10 or 20. not sure how they achieved such a low price

 

[Shivansps]

Im fully expecting both Microsoft and Intel doing some sacrifice here in order to get to the masses.
Just check the sub $200 tablets, $5-10USD SoC, cheap materials, and Android, BUT they need to do some work on the android, Android maybe free, but its not "plug and play", it has a cost, Windows if where free it ends to be cheaper than Android, because you only need to put Windows in there and install drivers, no IT work is needed.

If Intel can sell SoCs at very low prices about $20 or at cost, and Windows sell licenses for $5-20 a Windows tablet could end cheaper than a Android one. That may also explain the lack of interest in BT Android tablets by OEMs.

What i do not understand is the insane price diference on some brands, i think some oems are just greedy.

 

[podspi]

Yep, it looks like Microsoft and Intel have finally gotten serious after realizing that Android was going to eat their lunch.

 

[bullzz]

I dont get the part of intel sacrificing. clovertrail wud have costed $30 from estimates. at worst they would have cut down $15 for baytrail. I am expecting that to be price of chip with zero profit. but thats tiny if u see $100-150 difference between generations

 

[Shivansps]

Because OEM charges you premium for a Windows and/or tablet, and that is not always because of the license/SoC cost. And they are still doing it, how we can explain the Inspiron 11 3000 for 379 with a Haswell Celeron and them a Midland 10" Tablet with BT at $399?

 

[Khato]

While comparisons with Snapdragon 800 are now 'old news' comparatively... Notebookcheck has their review of the LG G2 up now - http://www.notebookcheck.com/Test-LG-G2-Smartphone.101923.0.html - which includes their typical power numbers. The most interesting of which is the delta between maximum idle (1.4W) and maximum load (4.8W). Now it's not clear how much of that is attributable to the SoC, so it's really only somewhat valid for comparison to other smartphones. The two most interesting of which would be the HTC One - http://www.notebookcheck.com/Test-HTC-One-Smartphone.89996.0.html - and Samsung Galaxy S4 - http://www.notebookcheck.com/Test-Samsung-Galaxy-S4-GT-I9505-Smartphone.91942.0.html - as they provide two data points for Snapdragon 600. The two data points being interesting as the stock Snapdragon 600 in the HTC One shows a delta power consumption of 2.9W while the 'boosted' version in the Samsung Galaxy S4 jumps up to 3.9W. Which is interesting as it's indicative of how quickly maximum SoC power scales for that last little bit of frequency, and that is likely the reason why the LG G2 performance falls so far short of both the MDP and Sony Xperia Z1.

Certainly looking forward to seeing similar data for the 5S now.