Windows tablets

[mczak]

**) A single HD 4000 EU can do two MADs per clock and has two texture samplers, theoretically doubling the performance of HD 3000 EU: http://www.anandtech.com/show/4830/intels-ivy-bridge-architecture-exposed/5

I don't think that's quite right. For the texture samplers that's one sampler per "EU array" instead of one for all arrays. Hence texturing rate actually goes down per EU for the GT1 variant (as it has 8 EUs instead of 6) but goes up per EU for the GT2 variant (of course, in absolute numbers, same texturing rate when comparing IVB GT1 to SNB GT1 and twice the rate when comparing IVB GT2 to SNB GT2).
I'm sceptical as far the doubled MADs are concerned. The intel slides don't mention it, just saying "extended co-issue". If co-issue worked with transcendental ops on SNB I can't tell how as it seems to be hidden from driver in that case. But these obviously had a lower execution rate than ordinary math (again, hidden from the driver) so even if it can co-issue MADs now it might not be able to issue them at the same base rate (the chip doesn't have a "true" MAD anyway). Well could be but call me not convinced yet.
I'm actually a bit surprised that intel didn't slap on another 1000 or so in the model numbers since they got a new feature set. Instead the "HD2500" moniker seems to indicate it isn't quite as fast as the HD3000 was.
(Oh and it's not a "completely new" architecture, rather just significantly different. It still shares quite a bit with its predecessors, just look at the open-source driver).

 

[dagamer]

Are Windows tablets going to use Intel or ARM?

Looks like Intel is making a big push with Ultrabooks right now at CES. Not clear if they're showing what will be Ivy Bridge products or not.

They'll probably use both, but I expect ARM tablets to have longer battery life and thinner designs. Intel won't really be able to match what an ARM tablet can do until Haswell comes out next year.

The key issue with ARM tablets we don't know about is whether the desktop app will be present. My personal feeling is that it adds far too much complexity compared to something like an iPad, however, when you have tablet designs like the ASUS Transformer Prime, it allows you to run desktop-like apps when docked with a keyboard and possibly larger monitor.

 

[sebbbi]

Are Windows tablets going to use Intel or ARM?

There will be both x86 based tablets and ARM based tablets. x86 tablets are fully backwards compatible with Windows 7 and Vista, and run all the old Windows software. Win8 also uses same driver model as its predecessors and has full support for PC USB devices. So you can connect all your existing PC USB peripherals to them (printer, scanner, digital camera, external DVD burner, Blu Ray drive, USB memory sticks/drives, etc). And you can connect them (wlan) to your home/work Windows network for file/printer sharing, etc.

Microsoft hasn't yet told much official information about the feature set in the ARM version of Win8, except that old x86 Windows software is not compatible with the ARM version. Speculation tells us that only Metro applications are available in the ARM version, and there's no traditional Windows desktop at all. ARM manufacturers have so far only shown Metro apps on their devices in trade shows. But it's better to wait for official information about the ARM version. Microsoft has been awfully quiet about it.

According to engadget's review, that slate had a 3.5 hour battery runtime. How on earth is that usable?

That's true. When stressed the battery lasts less than 4 hours. It has exactly the same hardware as the newest 11" Macbook Air has (64/128 GB Samsung SSD, 1366x768 display, Sandy Bridge i5 2467M processor, 4 GB of memory). Battery life is very similar to Macbook Air when both devices are running the same software.

Samsung Slate is very good for artists, because it has a 1024 level Wacom pressure sensor (and supports full sized Wacom pen), and is capable of running full PC version of Photoshop and other programs professional artists use. You can dock it and connect a PC keyboard and mouse to it (at hotel room, when you do some real work... much more ergonomic than writing on a small 11"/13" laptop keyboard). It's a real computer, not a glorified web browser. The battery life is very much comparable to the best 11" and 13" laptops. 7 hours in light use, and 3.5 hours in synthetic battery stress test.

 

[rpg.314]

That's true. When stressed the battery lasts less than 4 hours. It has exactly the same hardware as the newest 11" Macbook Air has (64/128 GB Samsung SSD, 1366x768 display, Sandy Bridge i5 2467M processor, 4 GB of memory). Battery life is very similar to Macbook Air when both devices are running the same software.

Samsung Slate is very good for artists, because it has a 1024 level Wacom pressure sensor (and supports full sized Wacom pen), and is capable of running full PC version of Photoshop and other programs professional artists use. You can dock it and connect a PC keyboard and mouse to it (at hotel room, when you do some real work... much more ergonomic than writing on a small 11"/13" laptop keyboard). It's a real computer, not a glorified web browser. The battery life is very much comparable to the best 11" and 13" laptops. 7 hours in light use, and 3.5 hours in synthetic battery stress test.

I dunno about that.

http://www.engadget.com/2011/12/16/samsung-series-7-slate-pc-review/

3.5 hour runtime.

Macbook Air

5.5 hour runtime.

They may have the same hw, but clearly they have different batteries.

 

[sebbbi]

They may have the same hw, but clearly they have different batteries.

Macbook Air has 5400 mAh battery. Samsung Slate has 5520 mAh battery. That conclusion doesn't hold.

Are you perhaps comparing OSX scores to Windows 7 scores? Macbook Air lasts a considerably longer in OS X compared to Windows 7 install with Boot Camp (Boot Camp is the easiest way to run exactly same binary code on both devices). If you are doing the battery life test using a browser, Safari performs a lot better on OSX compared to Windows, while for example Opera performs much better on Windows (I would guess they used Internet Explorer on Samsung Slate). The differences between OS/browser configurations are pretty big, so the only way to get a fair result is to compare both using the same OS (either Bootcamp the Mac or install "hachintosh" OSX to the Slate, or at least test with the most efficient browser).

Both Apple and Samsung claim 7 hour battery lifes for their devices. I have seen battery life benchmarks for both in the 3-4 hour range, and it seems that you can even get lower if you really try (http://maccrazy.com/macbook-air-battery-life). But in real life, both have pretty good battery life, and are fantastic devices I much rather carry around than a beefier laptop.

 

[french toast]

Am i the only person who isnt overly enamoured with Ivy bridge??

The way Intel has been pluggin this 22nm FINfet you would think it was the second coming of jesus christ!

With all that advancement, we get exactly the same TDP as the parts its replacing...
Granted the cpu performance looks around %20 better and they have improved their rubbish GPU 60% which is welcome, but i feel a bit underwhelmed by Ivy bridge, maybe i was expecting miracles..

 

[Exophase]

Am i the only person who isnt overly enamoured with Ivy bridge??

The way Intel has been pluggin this 22nm FINfet you would think it was the second coming of jesus christ!

With all that advancement, we get exactly the same TDP as the parts its replacing...
Granted the cpu performance looks around %20 better and they have improved their rubbish GPU 60% which is welcome, but i feel a bit underwhelmed by Ivy bridge, maybe i was expecting miracles..

What do you mean, exactly the same TDP? The 95W parts are going down to 77W with 100MHz more clock, while offering a much faster GPU and some IPC improvements. I'd say that's pretty good - not earth shattering, but by Intel shrink standards at least on par. I am curious as to exactly what you were expecting, what this miracle was supposed to be.

There's no compelling reason for most SB owners to upgrade, but that's just how things are now, you can't really expect huge increases to single thread performance anymore.. And Intel doesn't want to put more cores on affordable CPUs, which for most of us isn't really that compelling either.

 

[french toast]

Ok you have caught me out there, im not familiar with the whole inventory, i just came on and looked at the example at the top of this page;

Sandy Bridge (Core i7-2677M, 17W)
CPU: 2/4 threads, 1.80 GHz base, 2.90 GHz turbo
GPU: HD 3000 (12 EU), 350 MHz base, 1200 MHz turbo

Ivy Bridge (Core i7-3667U, 17W)
CPU: 2/4 threads, 2.00 GHz base, 3.20 GHz turbo
GPU: HD 4000 (16 EU **), 350 MHz base, 1150 MHz turbo

My assumptions were based off this..

So at the notebook level anyway, we didnt get a reduction in power consumption, although we did get better performance granted.
I supose i was expecting something silly like 50% lower power for about the same performance, after all that would give Intel the most benefit.

 

[metafor]

Ok you have caught me out there, im not familiar with the whole inventory, i just came on and looked at the example at the top of this page;

Sandy Bridge (Core i7-2677M, 17W)
CPU: 2/4 threads, 1.80 GHz base, 2.90 GHz turbo
GPU: HD 3000 (12 EU), 350 MHz base, 1200 MHz turbo

Ivy Bridge (Core i7-3667U, 17W)
CPU: 2/4 threads, 2.00 GHz base, 3.20 GHz turbo
GPU: HD 4000 (16 EU **), 350 MHz base, 1150 MHz turbo

My assumptions were based off this..

So at the notebook level anyway, we didnt get a reduction in power consumption, although we did get better performance granted.
I supose i was expecting something silly like 50% lower power for about the same performance, after all that would give Intel the most benefit.

Well obviously they can scale down the clockspeed and voltage to reach much lower power levels and still provide plenty of performance. It may just be that Intel feels 17W TDP is low enough for laptops.

Until we get higher performance ~1W ARM chips going into laptops, they're not gonna change their mind on that one.

 

[Silent_Buddha]

As long as laptop owners want to run full program packages rather than the lightweight and relatively limited apps you can get for most Arm devices, I don't think Intel will be going much lower for laptops. For lower end devices like netbooks you have the Atom processors.

But speaking from experience the Atom processors while nice for lightweight useage is completely inadequate for most heavyweight applications (Photoshop, for example, which as Sebbbi pointed out is one of the killer aps for Windows slates). Hence we have slates with the ULP Intel CPUs.

It's all about a compromise between performance and power use. If you don't need the performance then go with an Atom or Arm device. If you need the performance, especially for work applications, then there's not much choice.

If I ever see a full Adobe package featuring Photoshop running well on an Arm processor I'll take back my words that it's mainly only useful for lightweight apps and portable media. But until then, Arm does not and currently cannot address the market the Intel ULP processors are designed for.

Even Apple isn't silly enough to use Arm for their MacBooks even though they are theoretically fully capable of doing so with MacOSX.

Regards,
SB

 

[metafor]

I can't say I agree. The most complaints I've heard about Atom netbooks isn't the lack of ability for heavy duty work but the unresponsive UI (mostly due to Windows being resource heavy) and the lack of video capabilities (to handle 1080p youtube, for instance). Add to that the fact that Office applications can bog down an Atom netbook due to heavy resource expectations and you can see why people are unsatisfied.

I expect none of these issues to exist come Windows 8 on ARM. I also expect that at some point, the Macbook Air will come with an ARM chip that can provide a better user experience than any Atom based netbook.

 

[french toast]

Well obviously they can scale down the clockspeed and voltage to reach much lower power levels and still provide plenty of performance. It may just be that Intel feels 17W TDP is low enough for laptops.

Until we get higher performance ~1W ARM chips going into laptops, they're not gonna change their mind on that one.

Very soon they might I just assumed that the push has too be about low power and graphics, they own the IPC side of things as it stands, they might be underestimating ARM.

Personally i was expecting something like the same ish performance for half the power or 30% lower power with at least 2x gpu.
But anyway it is obviously an improvement.

Just a thought, do you think silvermont could be a ultra low power ivybridge? obviously the performance would be nowhere near, but i mean the basic architecture, do you think it could scale that low?

 

[Silent_Buddha]

I can't say I agree. The most complaints I've heard about Atom netbooks isn't the lack of ability for heavy duty work but the unresponsive UI (mostly due to Windows being resource heavy) and the lack of video capabilities (to handle 1080p youtube, for instance). Add to that the fact that Office applications can bog down an Atom netbook due to heavy resource expectations and you can see why people are unsatisfied.

I expect none of these issues to exist come Windows 8 on ARM. I also expect that at some point, the Macbook Air will come with an ARM chip that can provide a better user experience than any Atom based netbook.

If they could have they would have by now.

Apple has been working with ARM processors at least since 2005 (iPhone developement began), retail experience with it starting in 2007 (iPhone launch). I'm quite sure they'd like to be able to have a larger profit margin by using ARM processors than Intel processors.

It took them a comparatively short amount of time to transition from PowerPC CPU's to Intel CPU's. I have little doubt they could have and would have transitioned to ARM CPU's if it was even remotely capable of serving their customers needs in a fullblow MacOSX environment.

I have a feeling that if ARM was asked to run MacOSX, its performance would be far worse than Atom running a Windows desktop environment.

And yes, Atom does fall apart (performance wise) when asked to do anything serious involving computation. I've got an Atom netbook and slate. And while they are "good enough" for basic web browsing and media capabilities, it's almost intolerable for any serious application. And to note it is relatively fluid at the desktop and web-browsing. As well as fluid 1080p playback thanks to a video acceleration chip.

If all I wanted was lightweight applications and fluid video then it is perfect. But my needs go beyond that. Still it's almost good enough for travel. A good enough compromise between portability and useability.

I would have pulled the trigger on a SB ULP slate by now, except I can afford to wait for the IVB ULP slates. Hopefully one of the slate manufactures instead of going for the 200 mhz clockspeed increase, instead underclocks it for lower power use while still being slightly faster than the SB ULP predecessor.

Regards,
SB

 

[french toast]

Well im running windows 7 basic on a intel netbook with an n270 Atom...i have to say its the worse performing peice of hardware i have ever seen.
Talk about choppy/lag..and thats just loading web pages!, you tube videos?..just about on low res, but it sometimes even struggles with that.

So i dont have any faith in Atom do even the most simple tasks, maybe Android/symbian is what it was designed for.
It always runs out of battery with in a couple of hours as well, terrible.

As far as im concerned an A9 is the equivalent to a ultra low power core 2 duo, (clock for clock) with much better power/heat charactaristics.

 

[eastmen]

Well im running windows 7 basic on a intel netbook with an n270 Atom...i have to say its the worse performing peice of hardware i have ever seen.
Talk about choppy/lag..and thats just loading web pages!, you tube videos?..just about on low res, but it sometimes even struggles with that.

So i dont have any faith in Atom do even the most simple tasks, maybe Android/symbian is what it was designed for.
It always runs out of battery with in a couple of hours as well, terrible.

As far as im concerned an A9 is the equivalent to a ultra low power core 2 duo, (clock for clock) with much better power/heat charactaristics.

oh I wouldn't say that . The atom sucks for sure and I have an hp311 with a 1.6ghz atom so i can understand why you hate your netbook. I also have a amd bravos c-50 dual core 1ghz tablet that runs rings around that atom and seems to do the majority of tasks much faster than my touchpad running andriod . Loading games like plants vs zombies is much faster on my amd tablet and even loading web pages is faster.

Of course my touchpad lasts alot longer than my amd system . The touchpad has a 6300mah battery and the amd unit has a 3260mah battery. Of course the touchpad will normaly give me 8 + hours of hd video playback and the best i've gotten from the amd system is 4hrs

I can't help but think that in october when the windows 8 tablets come out a properly designed tablet using a 28nm verison of bobcat and a larger battery wouldn't put up comparable times to the andriod tablets while being more powerful and allowing me to use all my x86 apps

 

[french toast]

Yea my long running dream with this fusion push was exactly that goal, they havn't quite put out a chip thats as capable yet.

Yea i have heard that from a number of people, that the bobcats are much much smoother compared to atom, makes you wander whether they could have done the same thing as Intel did with the z series and push into smartphones?? the funny thing is Qualcomms Adreno was originally Amds imageon graphics, which was based off a mini radeon.indeed Adreno is an anagram of Radeon!

Why did they sell it? (for a paltry $60 mill) has to go down as one of the supidist things of recent times, they could of paired it up with a low power bobcat, and made a killing. i worry about them at the moment, ARM is pushing up from the bottom, and Intel is pushing down from the top, its not looking good.

Ideally A 28nm quad core bobcat with a 320 core graphics, with a shed load of bandwith would go down a treat, however i think cortex a15/krait will be approaching that performance.

 

[Simon F]

I can't say I agree. The most complaints I've heard about Atom netbooks isn't the lack of ability for heavy duty work but ... the lack of video capabilities (to handle 1080p youtube, for instance).

I think that must have depended on the chip set/GPU the manufacturers chose. I'm sure some shipped with hardware video decode in the form of VXD, and that offloaded the CPU virtually entirely.

 

[tangey]

I think that must have depended on the chip set/GPU the manufacturers chose. I'm sure some shipped with hardware video decode in the form of VXD, and that offloaded the CPU virtually entirely.

As I recall (and not for the first time), the main issue with such solution was not the underlying hardware, it was the fact that the driver implementation for it was crap, especially on windows platforms. late on, video playback performance improved in win7 solutions, but it was a long time coming.

From the very first poulsbo chips, intel has done a really poor job in drivers for the graphics/video for their mobile/netbook SGX solutions. The ongoing issues with cedartrail (apparently delayed from Sept to Jan due to windows compliance issues with the drivers, and now only exposing DX9.x when initially DX10.1 was promised), some 4 years after poulsbo, would indicate that there is something badly going wrong in that particular dept of Intel.

A cyncial person like me might think that it suits intel's in-house graphics department to have IMG graphics/video poorly implemented.

 

[rpg.314]

If they could have they would have by now.

Apple has been working with ARM processors at least since 2005 (iPhone developement began), retail experience with it starting in 2007 (iPhone launch). I'm quite sure they'd like to be able to have a larger profit margin by using ARM processors than Intel processors.

It took them a comparatively short amount of time to transition from PowerPC CPU's to Intel CPU's. I have little doubt they could have and would have transitioned to ARM CPU's if it was even remotely capable of serving their customers needs in a fullblow MacOSX environment.

I have a feeling that if ARM was asked to run MacOSX, its performance would be far worse than Atom running a Windows desktop environment.

ARM is an ISA. Atom is a micro-architecture. That comparison does not compute.

I expect ARM64 cores would be VERY competitive on serial IPC with contemporary Atoms and do pretty well running a desktop OS. Last two years have brought us ~2x every year improvement and I think we have 2 more years of such growth left in 1W SoCs.

 

[rpg.314]

Macbook Air has 5400 mAh battery. Samsung Slate has 5520 mAh battery. That conclusion doesn't hold.

Are you perhaps comparing OSX scores to Windows 7 scores? Macbook Air lasts a considerably longer in OS X compared to Windows 7 install with Boot Camp (Boot Camp is the easiest way to run exactly same binary code on both devices). If you are doing the battery life test using a browser, Safari performs a lot better on OSX compared to Windows, while for example Opera performs much better on Windows (I would guess they used Internet Explorer on Samsung Slate). The differences between OS/browser configurations are pretty big, so the only way to get a fair result is to compare both using the same OS (either Bootcamp the Mac or install "hachintosh" OSX to the Slate, or at least test with the most efficient browser).

Both Apple and Samsung claim 7 hour battery lifes for their devices. I have seen battery life benchmarks for both in the 3-4 hour range, and it seems that you can even get lower if you really try (http://maccrazy.com/macbook-air-battery-life). But in real life, both have pretty good battery life, and are fantastic devices I much rather carry around than a beefier laptop.

Even on the Mac OS X comparison, Macbook Air lasted for ~40 min more, so may be it is Windows that is the culprit here. Let's hope Win 8 will go on a diet.

Either way, I think we need OS and process to shrink quite a bit before Windows tablets will be useful.