The Intel Atom Thread

[podspi]

I'm surprised there are people who disagree with Atom in the Surface 3. What advantage is there to ARM, other than it being a relatively interesting piece of tech?

WARM is why I probably will pick up a Surface 1/2 at some point, just as a collector's item.

 

[Thala]

I for myself wouldn't think twice before picking up this new Surface 3 if I wanted an all around device for media and productivity.

Now i dont need an all around device. And as i mentioned productivity beyond Office/Outlook i would not choose an Atom anyway. Besides there are literally thousands of other options if you look for a Cherrytrail Windows device.
I would have bought an ARM based (e.g. NVidia Tegra) Surface 3 on release day, now i will look elsewhere.

If you look past Geekbench you will find out a slower 2.3GHz Moorefield is still competitive with some of the latest ARM SoCs CPU-wise.

Sorry. It has been critisized on and on, that with Web Benchmarks you measure the Javascript implementation more than anything else.
As soon as you run a Benchmark, which really tests the CPU, Silvermont/Airmont falls far behind. Pretending IPC is good based on Web benchmarks is just stupid.
And then we have the famous "memory benchmarks", which measure memory performance. They are as useless and misleading as having a benchmarks for battery capacity. Having a device with huge memory bandwidth but still no IPC advantage is like having a big battery but no longer duration. It just shows the inefficiency of the system.
If we remove "Web" and "Memory" from Basemark benchmark, you have damn slow device - essentially yesterdays technology.

If you want to trade this for x86 compatibility - fine for you.
For myself it means that i will need to go back to iOS or Android if i want the latest and greatest despite i was very satisified with both RT Surface devices.

In the end it is Microsoft's loss because by limiting the options they do not do their ecosystem a favour. You only gain market share if you offer more choice, not less.

 

[Arachnotronic]

What about posting a video of you playing it with touchscreen only, I want to hear you curse the Surface 3 :biggrin:

I don't see how PC games can be played without proper controls (joysticks, pad, mouse). I bet that those that can be played with touchscreen, already have a port to or come from iOS/Android.

Yeah that's why I don't own a convertible. Big honking desktops + laptop + iPad here

 

[Sweepr]

Now i dont need an all around device. And as i mentioned productivity beyond Office/Outlook i would not choose an Atom anyway.

If you don't need an all around media/productivity device then why you chose the Surface in the first place and not an iPad or Android tablet with a much larger number of mobile apps and developer support?

Besides there are literally thousands of other options if you look for a Cherrytrail Windows device.

None other than the Surface 3 announced till now, and I doubt any of them will get the same treament in terms of build quality and features. Same could be said about Haswell-U and Windows, but the Surface Pro 3 is still a success.

Sorry. It has been critisized on and on, that with Web Benchmarks you measure the Javascript implementation more than anything else.
As soon as you run a Benchmark, which really tests the CPU, Silvermont/Airmont falls far behind.

Which benchmarks? Given the positive reactions from the Surface 3 announcement I'm inclined to believe that people would give up some performance (and bragging rights) in order to gain full support from the x86 Windows ecosystem. Let the sales speak for themselves a few months from now.

If you want to trade this for x86 compatibility - fine for you.
For myself it means that i will need to go back to iOS or Android if i want the latest and greatest despite i was very satisified with both RT Surface devices.

Seriously, what kind of applications available for Windows RT would run slow with an Airmont core and a lot better with ARM's latest and greatest stuff? We don't even know how Airmont will stack up (all we have are some Geekbench submissions from ES). The fact that Cherry Trail should be very power-efficient shouldn't be downplayed either, especially if it can provide a similar experience with better battery life.

In the end it is Microsoft's loss because by limiting the options they do not do their ecosystem a favour. You only gain market share if you offer more choice, not less.

They are doing their ecosystem a favour by closing the gap between the regular Surface and the Surface Pro in terms of software and features, that's what part of the community has been asking for years, despite your/my personal preferences.

 

[Thala]

If you don't need an all around media/productivity device then why you chose the Surface in the first place and not an iPad or Android tablet with a much larger number of mobile apps and developer support?

Because productivity on this form-factor and performance range in my case ends with office, which is available for RT. Add printer support, multi-monitor support, multi-user support (which goes down to the file-system) and other services like file services, scripting services - essentially everything that Windows offers, which is very limited with Android and iOS. I had not a single use-case where i wanted to install additional desktop apps.
Make no mistake, Windows itself and it's broad range of features made me favour the Surface - but not x86 compatibility.
Oh and of course the was the nice "feature" of beeing practically immune to viruses and trojans as there is no known virus, which runs on Windows for ARM.

Which benchmarks? Given the positive reactions from the Surface 3 announcement I'm inclined to believe that people would give up some performance (and bragging rights) in order to gain full support from the x86 Windows ecosystem. Let the sales speak for themselves a few months from now.

Which Benchmarks? The very same benchmarks you did link from Anandtech.
It does not help if you generalize. Most people were perfectly happy with their Surface RT devices and did not have the need to tap into the x86 ecosystem. You'll get positive reactions from so called "power-users" posting on the boards of course.

Seriously, what kind of applications available for Windows RT would run slow with an Airmont core and a lot better with ARM's latest and greatest stuff?

Most games for instance do not run close to 60fps for instance. There is a huge gap to close regarding performance.
Why choose a slower CPU/GPU if there are faster ones on the market for the same price?

We don't even know how Airmont will stack up (all we have are some Geekbench submissions from ES). The fact that Cherry Trail should be very power-efficient shouldn't be downplayed either, especially if it can provide a similar experience with better battery life.

Airmont is just a tick. Given, the GPU is somewhat faster but it was not the best anyway when it was released.
Regarding power, there was no evidence that an Atom tablet lasts longer than an ARM tablet...and we are talking about times when Intel had the FinFet advantage - which is essentially gone by now.

Example: My Surface 2 for instance plays video for 12-13 hours on medium brightness and that is with the much cursed Cortex A15 on 28nm. My Surface Pro 3 barely manage to reach 7 hours despite bigger battery. I dont have an 22nm Atom device for testing.

They are doing their ecosystem a favour by closing the gap between the regular Surface and the Surface Pro in terms of software and features, that's what part of the community has been asking for years, despite your/my personal preferences.

Again, you are not growing your ecosystem if you limit choice. For performance hungry users looking for a tablet, they now can look elsewhere but not in the Windows ecosystem.
The community you are talking about were most likely already part of the Microsoft ecosystem, because they already had the choice of buying and x86 tablet. As i said, the Surface 3 is now yet another Atom offering.

 

[Sweepr]

Because productivity on this form-factor and performance range in my case ends with office, which is available for RT.

Well it doesn't necessarily ends with Office for other people, hence why x86 support matters for them.

Which Benchmarks? The very same benchmarks you did link from Anandtech.
It does not help if you generalize. Most people were perfectly happy with their Surface RT devices and did not have the need to tap into the x86 ecosystem. You'll get positive reactions from so called "power-users" posting on the boards of course.

You mean the same benchmarks that gave them this conclusion about Moorefied? And we all know Cherry Trail will be faster.

The Z3580 performs very well in all of our web based benchmarks. Its scores are in the same range as devices with Apple's A8 and NVIDIA's Tegra K1 which currently hold the best scores of the devices we've tested.

Going over the Venue 8's strengths beyond its design, we find that it does quite well in our CPU tests. The Intel Atom Z3580 manages to hold its own against the competition's high end ARM SoCs, and at this point the list of applications that don't run on Intel devices consists of only a tiny handful of Android NDK apps. Its speed also doesn't come at the cost of battery life, with the Venue 8 trading blows with the competition in our battery tests.

Most games for instance do not run close to 60fps for instance. There is a huge gap to close regarding performance.
Why choose a slower CPU/GPU if there are faster ones on the market for the same price?

Probably because they listened to Surface users that want to do more than Office work with their $500 devices. And again, better benchmark performance doesn't necessarily translate to a more fluid experience and is not always perceptible by users. Also we don't even know how Cherry Trail performs yet, it might have slower graphics than K1/A8X but it should be able to play all mobile games just fine. It also allows users to play hundreds of Windows hardcore titles (exclusives and console ports) at lowered settings. Android/iOS hardcore game library is a joke in comparison and most games run perfectly fine with older hardware like the Apple A7 and Snapdragon 801.

Example: My Surface 2 for instance plays video for 12-13 hours on medium brightness and that is with the much cursed Cortex A15 on 28nm. My Surface Pro 3 barely manage to reach 7 hours despite bigger battery. I dont have an 22nm Atom device for testing.

There should be tests soon, but I expect Airmont to be extremely power-efficient for basic stuff like web browsing.

Again, you are not growing your ecosystem if you limit choice. For performance hungry users looking for a tablet, they now can look elsewhere but not in the Windows ecosystem.
The community you are talking about were most likely already part of the Microsoft ecosystem, because they already had the choice of buying and x86 tablet. As i said, the Surface 3 is now yet another Atom offering.

Performance-hungry users will be fine with Apple and Google's best if they really care about running tablet apps a bit faster, but then you're not getting the benefit of x86 software support which some people do value. At the end of the day it's all about choices and compromises. Microsoft's choice just shows that Surface 1 and 2 sales weren't good enough to justify keeping their ecosystem fragmented and IMHO (and quite a lot of users agree) it was the right one.

 

[Kaido]

Atom X7-Z8700, 4GB RAM, 1080p/1440p IPS screen, 64GB storage (base model) + MicroSD, reasonable battery capacity, great build quality, full x86 64-bit Windows 8.1 (upgradeable to Windows 10), $399 starting price. Come on Microsoft, you can do it (Surface 4).

Wow, nailed it! :thumbsup:

 

[Thala]

Well it doesn't necessarily ends with Office for other people, hence why x86 support matters for them.

Look Sweepr, you somehow miss my point entirely. I do not doubt that there are people for whom x86 support matters. These people have lots of choices in the Windows ecosystem.
There are many people as well, which want to do web-browsing, gaming, mailing and the causual Office stuff with their tablets, who care for performance.
These people do not have a choice anymore within the Windows eco-system. Microsoft basically bound them to competing ecosystems.
Exactly this is my Argument.

Also we don't even know how Cherry Trail performs yet, it might have slower graphics than K1/A8X but it should be able to play all mobile games just fine.

Sure, just like an AMD Desktop processor and a mid-range desktop GPU runs games just fine. Still people choose Intel desktop processors and fast desktop GPUs.
Also of course Cherrytrail is 1/2-1 year late compared to Tegra K1/A8X. They are both still 20nm/28nm planar. Technology gap is closing which inherently puts x86 at disadvantage.

There should be tests soon, but I expect Airmont to be extremely power-efficient for basic stuff like web browsing.

I am not argueing this. They might be able to compete with ARM designs regarding power in use-cases like browsing.

Microsoft's choice just shows that Surface 1 and 2 sales weren't good enough to justify keeping their ecosystem fragmented and IMHO (and quite a lot of users agree) it was the right one.

It is not just about one device, which when looked at the issue in isolation is just an abitrary choice. In the big picture Microsoft just abandoned ARM, the leading mobile architecture, for tablets.
What Microsoft needs to do is increase and widen the choice in order to grow their ecosystem, not limit it, like in this case to a single CPU architecture.
For myself i will go back to Android, where i have the choice not to use mid-range produts which feature Cherrytrail.

 

[TechFan1]

It is not just about one device, which when looked at the issue in isolation is just an abitrary choice. In the big picture Microsoft just abandoned ARM, the leading mobile architecture, for tablets.
What Microsoft needs to do is increase and widen the choice in order to grow their ecosystem, not limit it, like in this case to a single CPU architecture.
For myself i will go back to Android, where i have the choice not to use mid-range produts which feature Cherrytrail.

Microsoft didn't abandon ARM. OEMs can still make an ARM based tablet. It's just that pretty much all the manufacturer's, and Microsoft decided Atom allowed them to make a more competitive product for whatever reason.

 

[NTMBK]

It is not just about one device, which when looked at the issue in isolation is just an abitrary choice. In the big picture Microsoft just abandoned ARM, the leading mobile architecture, for tablets.
What Microsoft needs to do is increase and widen the choice in order to grow their ecosystem, not limit it, like in this case to a single CPU architecture.
For myself i will go back to Android, where i have the choice not to use mid-range produts which feature Cherrytrail.

Microsoft are still going to allow for ARM on Windows tablets <8" in size- these are the ones which will not run "desktop" apps, and only run the new Metro apps, where compatibility with legacy x86 software is not a problem. But the 8"+ tablets will all be x86, with desktop support.

 

[Sweepr]

Look Sweepr, you somehow miss my point entirely. I do not doubt that there are people for whom x86 support matters. These people have lots of choices in the Windows ecosystem.
There are many people as well, which want to do web-browsing, gaming, mailing and the causual Office stuff with their tablets, who care for performance.
These people do not have a choice anymore within the Windows eco-system. Microsoft basically bound them to competing ecosystems.
Exactly this is my Argument.

I get your point, but as others said other OEMs can still launch ARM Windows devices if they want to and if you want more performance you can always opt for a new/used Surface Pro 2/3 (or 4 when Windows 10 arrives), not to mention the Core M convertibles coming in the next months (some of them might cost the same as a Surface 3). Once again I think Cherry Trail would get the job done fine for the tasks you mentioned, there's not a lot of demanding games for Windows RT anyway (unlike regular Windows). My personal choice would be running mobile + conventional Windows games in a single device instead of just mobile games at a higher framerate.

Also of course Cherrytrail is 1/2-1 year late compared to Tegra K1/A8X. They are both still 20nm/28nm planar. Technology gap is closing which inherently puts x86 at disadvantage.

They are probably looking past Cherry Trail with this decision. Broxton is coming out next year with a brand new CPU and graphics architecture.

What Microsoft needs to do is increase and widen the choice in order to grow their ecosystem, not limit it, like in this case to a single CPU architecture.
For myself i will go back to Android, where i have the choice not to use mid-range produts which feature Cherrytrail.

Which mid-range ARM SoC matches Cherry Trail right now?
Mediatek, Huawei and Allwiner are using mostly Cortex A53 cores and weak iGPUs that can't match 2013 stuff like Apple A7 and Qualcomm S801. Qualcomm's S615 is a complete joke, 15 FPS @ GFXBench T-Rex, that's slower than Bay Trail's iGPU (worse CPU too).
Complaining about graphics performance and then putting Cherry Trail next to crap available for mid-range ARM devices doesn't make sense.
On the CPU side things will get better in the next few months with the new Cortex A72 cores, but these SoCs are not coming till H2/2015 and looking at Moorefield benchmarks I think top-bin Cherry Trail will hold its own most of the time (and the efficiency comparisons should be interesting).

 

[liahos1]

how could anyone logically decry msft putting cherrytrail in the next surface (non pro). RT devices were utterly gimped and nearly useless when put in the broader context of Windows x86 software ecosystem. Baytrail was a surprisingly capable chip for full windows applications and cherry trail should be even better.

 

[Arachnotronic]

how could anyone logically decry msft putting cherrytrail in the next surface (non pro). RT devices were utterly gimped and nearly useless when put in the broader context of Windows x86 software ecosystem. Baytrail was a surprisingly capable chip for full windows applications and cherry trail should be even better.

I literally do not understand why anybody would think having a solid performing x86 chip in the Surface 3 is a bad thing.

Cherry Trail may not be the world's fastest mobile SoC, but it should offer pretty good performance and, more importantly, compatibility with the huge library of full Windows applications.

A Tegra X1, with all of its GPU might, isn't all that useful in a Windows device if I can't run Steam on the darn thing. It's not like mobile games in the Windows Store are all that demanding

 

[podspi]

Yeah, I'm not sure why the underlying instruction set is important?

The only 'disadvantage' of x86 is backwards compatibility going back a long time?

I mean, if you liked the Surface RT, the Surface 2 was great because it could do everything the Surface RT could do, just faster with better battery life.

Surface 3 will be able to do everything Surface 2 did, just faster, with better battery life, and x86 compatibility. I do concede the virus issue, hopefully Spartan will have some sort of sandbox feature (esp when browsing in metro mode) that makes this less of an issue.

 

[Nothingness]

The only 'disadvantage' of x86 is backwards compatibility going back a long time?

The main disadvantage of x86 (apart from its ugliness, but as a user you don't care ) is Intel quasi-monopoly.

 

[Shivansps]

wth, how on earth could someone complain about having x86 in a WINDOWS device...

 

[podspi]

The main disadvantage of x86 (apart from its ugliness, but as a user you don't care ) is Intel quasi-monopoly.

In this space at least, Intel isn't even close to a monopoly, and I could care less about the alleged ugliness of x86, though I can't be bothered to.

 

[Sweepr]

First Surface 3 Geekbench submission out in the wild:

http://browser.primatelabs.com/geekbench3/2226529

It's faster than the top Bay Trail out there. Atom Z3795 runs at the same clocks, which means there is some IPC gain.
Most 2.4GHz Bay Trail submissions are in the 900-970 range for single-core and 3000-3150 for multi-core.

 

[liahos1]

First Surface 3 Geekbench submission out in the wild:

http://browser.primatelabs.com/geekbench3/2226529

It's faster than the top Bay Trail out there. Atom Z3795 runs at the same clocks, which means there is some IPC gain.
Most 2.4GHz Bay Trail submissions are in the 900-970 range for single-core and 3000-3150 for multi-core.

this looks promising. hopefully good battery life.

 

[North01]

First Surface 3 Geekbench submission out in the wild:

http://browser.primatelabs.com/geekbench3/2226529

It's faster than the top Bay Trail out there. Atom Z3795 runs at the same clocks, which means there is some IPC gain.
Most 2.4GHz Bay Trail submissions are in the 900-970 range for single-core and 3000-3150 for multi-core.

Surface Pro 3 (i3-4020Y): 1651 single-core / 3438 multi-core
Surface 3 (x7-Z8700): 1009 single-core / 3430 multi-core

 

[Nothingness]

First Surface 3 Geekbench submission out in the wild:

http://browser.primatelabs.com/geekbench3/2226529

It's faster than the top Bay Trail out there. Atom Z3795 runs at the same clocks, which means there is some IPC gain.
Most 2.4GHz Bay Trail submissions are in the 900-970 range for single-core and 3000-3150 for multi-core.

You're doing a small mistake: that 1009 score is for a 64-bit run, and is mostly the same as 64-bit z3795 scores. Look at this for instance: http://browser.primatelabs.com/geekbench3/compare/1203078?baseline=2226529

No IPC gain as far as I can see.

 

[Shivansps]

You're doing a small mistake: that 1009 score is for a 64-bit run, and is mostly the same as 64-bit z3795 scores. Look at this for instance: http://browser.primatelabs.com/geekbench3/compare/1203078?baseline=2226529

No IPC gain as far as I can see.

Yeah, still most of the MT results are in favor of the CT, not by a huge margin but they are.

The memory performance is what worry me, looks like to be running in SC, damn shame if its running on SC.

 

[Nothingness]

The memory performance is what worry me, looks like to be running in SC, damn shame if its running on SC.

Perhaps MS used slower memory chips than Lenovo.

 

[Shivansps]

Perhaps MS used slower memory chips than Lenovo.

What can be slower than a DC of LPDDR3 1066? that petty much looks like 1066DC vs SC 1600.

 

[Nothingness]

What can be slower than a DC of LPDDR3 1066? that petty much looks like 1066DC vs SC 1600.

DC 1066 vs SC 1600 has ~33% more BW (2x1066/1600) while the Lenovo vs Surface 3 seems to have 10% more BW (6.3 GB/s / 5.7 GB/s on triad MT).