News - Android 4.0 running on ASUS Eee Pad Transformer Prime

[HEXUS]

NVIDIA gives us a preview of Android's tablet future.

Read more.

[Brewster0101]

I wonder if Nvidia have any plans to release a desktop CPU to take on Intel and AMD, they clearly have the know how.

[Smudger]

Don't they use ARM chips as the basis for Tegra?

[TheAnimus]

define basis!

They use ARM as the CPU design, but now adays its really all about the video decoding or hardware acceleration which is all their own.

[Scribe]

The question for me then is how powerful will ARM cpus become in the next few years. Will it be enough to compete in the same market? Hmmm

[krazy_olie]

The question for me then is how powerful will ARM cpus become in the next few years. Will it be enough to compete in the same market? Hmmm

ARM by definition can't do x86 windows/mac os/linux so in the short term... no. No doubt if there really is a Windows for ARM someone will put it on a desktop/laptop but x86 will still dominate the market in the short term

[kingpotnoodle]

ARM is an instruction set, like MIPS, x86, x86-64, IA64, POWER, SPARC etc. It's not a CPU in itself. ARM has it's roots in the old Acorns we used at school in the 90s, originally ARM stood for Acorn RISC Machine... they since replaced the Acorn with Advanced.

ARM Holdings (the company) produce reference CPU designs using their instruction set, these are licensed along with the instruction set to companies such as NVidia in return for wads of cash. So far to easy... some companies (Qualcomm notably) only use this as the basis for their own CPU designs, so not all ARM based CPUs are created equal, rather like how Intel and AMD use the same x86-64 set in the modern CPUs but have varying success/wins.

Each of the ARM licensees also uses it's own graphics and supplementary chip technology, there is a surprisingly large difference between the competing SOCs (System On a Chip). ARM Holdings also license a GPU design, the Mali - the main user of the design is Samsung in the Exynos line. TI use PowerVR SGX, Qualcomm bought ATI's mobile business a few years back and NVidia use their own GeForce derived design.

ARM instructions are few (RISC) and they get more instructions in per clock cycle, the disadvantage is that the "shortcut" instructions for things like AVX, SIMD etc are not present, as with most processors comparing on clock speed is pretty pointless. ARM based designs usually scale down well and they are low power but when it comes to raw grunt they are left crying in the dust by even a low end Celeron, they just cry quietly without using a lot of power; that's why they are used in phones/tablets but Celeron's aren't. ARM designs could be scaled out to compete on performance but doing this probably means increasing power consumption and losing some of the advantage they traditionally had. As things stand though an ARM desktop would suck badly on performance, it would be very slow compared to Intel/AMD chips using x86-64, add to that poor software support (i.e. full Windows, not even Windows 8 is pencilled to be the exact same edition on ARM CPUs) and it's a pretty useless idea.

ARM CPUs are not some magic bullet to make computing hugely faster (other RISC architectures have spectacularly failed to conquer the desktop space, Apple dumped POWER to go to x86), they go into lower power envelopes than traditional x86 designs but Intel is pushing to change that with Atom, so the future looks rather competitive.

I don't really see a case to scale out ARM designs to compete in the desktop space, it would be a very hard sell without the software support, the ARM using companies have carved a profitable dominance in the portable device market, I doubt they will focus anywhere else, especially seeing as that portable device market is exploding whilst the PC & larger laptop market is stagnant at best.

ARM in servers *might* have some legs in hyperscale web environments which process millions/billions of tiny transactions and benefit from throwing huge numbers of cores into the equation (probably the same things that already rock on an UltraSPARC T2 processor) but whether they take off remains to be seen, it's not like UltraSPARC T2 has taken over the world or anything...

Basically in 5/10 years time I still expect to buy mostly x86-64 based servers, an Intel based desktop PC (don't expect AMD to get any better and I don't do budget for my desktops), x86-64 laptop, most likely an ARM based tablet and phone. Exactly the same as now.

I'm not a pessimist, I'm a realist.

[Monkeysweat]

b1 is 32 GB and the c1 is 64 GB, i don't believe there will be a 16 gb model

[Scribe]

b1 is 32 GB and the c1 is 64 GB, i don't believe there will be a 16 gb model

I can confirm that there wont be, ASUS is targeting the high-end market hence the name 'Prime' as opposed to the Transformer 2 etc. They will continue to sell the Transformer, perhaps at a lower cost come release of the Prime. Mind you the fact that an a1 appears reserved could mean they're not counting this out as an option for the future, likely dependant on original Transformer sales.

[Scribe]

ARM is an instruction set, like MIPS, x86, x86-64, IA64, POWER, SPARC etc. It's not a CPU in itself. ARM has it's roots in the old Acorns we used at school in the 90s, originally ARM stood for Acorn RISC Machine... they since replaced the Acorn with Advanced.

ARM Holdings (the company) produce reference CPU designs using their instruction set, these are licensed along with the instruction set to companies such as NVidia in return for wads of cash. So far to easy... some companies (Qualcomm notably) only use this as the basis for their own CPU designs, so not all ARM based CPUs are created equal, rather like how Intel and AMD use the same x86-64 set in the modern CPUs but have varying success/wins.

Each of the ARM licensees also uses it's own graphics and supplementary chip technology, there is a surprisingly large difference between the competing SOCs (System On a Chip). ARM Holdings also license a GPU design, the Mali - the main user of the design is Samsung in the Exynos line. TI use PowerVR SGX, Qualcomm bought ATI's mobile business a few years back and NVidia use their own GeForce derived design.

ARM instructions are few (RISC) and they get more instructions in per clock cycle, the disadvantage is that the "shortcut" instructions for things like AVX, SIMD etc are not present, as with most processors comparing on clock speed is pretty pointless. ARM based designs usually scale down well and they are low power but when it comes to raw grunt they are left crying in the dust by even a low end Celeron, they just cry quietly without using a lot of power; that's why they are used in phones/tablets but Celeron's aren't. ARM designs could be scaled out to compete on performance but doing this probably means increasing power consumption and losing some of the advantage they traditionally had. As things stand though an ARM desktop would suck badly on performance, it would be very slow compared to Intel/AMD chips using x86-64, add to that poor software support (i.e. full Windows, not even Windows 8 is pencilled to be the exact same edition on ARM CPUs) and it's a pretty useless idea.

ARM CPUs are not some magic bullet to make computing hugely faster (other RISC architectures have spectacularly failed to conquer the desktop space, Apple dumped POWER to go to x86), they go into lower power envelopes than traditional x86 designs but Intel is pushing to change that with Atom, so the future looks rather competitive.

I don't really see a case to scale out ARM designs to compete in the desktop space, it would be a very hard sell without the software support, the ARM using companies have carved a profitable dominance in the portable device market, I doubt they will focus anywhere else, especially seeing as that portable device market is exploding whilst the PC & larger laptop market is stagnant at best.

ARM in servers *might* have some legs in hyperscale web environments which process millions/billions of tiny transactions and benefit from throwing huge numbers of cores into the equation (probably the same things that already rock on an UltraSPARC T2 processor) but whether they take off remains to be seen, it's not like UltraSPARC T2 has taken over the world or anything...

Basically in 5/10 years time I still expect to buy mostly x86-64 based servers, an Intel based desktop PC (don't expect AMD to get any better and I don't do budget for my desktops), x86-64 laptop, most likely an ARM based tablet and phone. Exactly the same as now.

I'm not a pessimist, I'm a realist.

ARM is synonymous with both instruction set and CPU core design these days. If I'm being correct I should have said arm-based though as you say, the core never typically stands alone; with a core design name typically being an ARM Cortex-A15 and the instruction set ARMv7.

Naturally it's reasonable to expect various flavours of SoCs but the core non-graphical powerhouse remains the same, varying typically in core numbers and frequency, not unlike a typical CPU and so very comparable, likewise both Intel and AMD offer differing graphical cores in their APU offerings and so it's not unreasonable to compare even knowing gpu cores may vary and though SoC performance differs at any one time, the difference is rarely significant over say a 6 month period except in cases of SoCs that target different power requirements, something we'd likely see less of were ARM SoCs to enter the desktop market.

I think it's natural to assume no one will complain if the ARM architecture is scaled up and finds itself using more power for desktop usage. With a new 64-bit instruction set on the way with highly scalable interfacing between cores, advances to the NEON vector unit and more emphasis on the use of the graphical core for floating point calculations it's clear ARM is already heading this way and opening doors for companies to jump on the bandwagon with much less effort, making it a why not give it a punt scenario, especially for smaller firms looking to squeeze in, even if that begins by taking a few ARMv8 SoCs and running them in a multi CPU configuration.

At the end of the day ARM is clearly now aiming for the server market where to be successful it must offer a LOT of power but at better power efficiency than competitors and this is the start of the magical formula for desktop. Though not x86, ARM based Linux is supported by most popular distributions with the vast majority of applications already available (I'm running Ubuntu 11.10 on a makeshift ARM-based computer on my desk right now). Android support is pre-existing with an increase in desktop-like applications and games thanks to tablet devices. Windows 8 is definitely coming to ARM, initially applications will be missing but it's reasonable to expect Microsoft to bring over its popular titles with 3rd party developers only having to look to, depending if they have created assembly optimisations or not, either recompiling their programs once libraries have been recompiled or rewriting optimised sections (normally only a small percentage of modern code) and then recompiling.

I would say it's all going to come down to how fast ARMv8-based cores turn out to be but after seeing, in these early stages 128 of them interconnected on a high-speed low-latency cache coherent interconnect clocked at 3GHz (admittedly unknown but definitely improved per cycle core performance), the potential for scaling is impressive and performance heavy applications are focusing more and more on multi-threaded optimisation; in which case it'll all come down to price, which, in the server market considerations are also for running costs, whereas for desktops, it'll be the initial cost only, which may be the difficult hurdle if many chips are required.

[Guppy]

I've got a Asus Transformer and its pretty awesome, can't see any reason to get the Prime as I heard the original Transformer will be getting ICS too. Seems to me like its just a small hardware upgrade. Any word on a price?

[miniyazz]

The video demonstrates a blazingly-fast and smooth response that, up until now, has only been associated with Apple iOS devices.

A bold statement

[kasavien]

The video demonstrates a blazingly-fast and smooth response that, up until now, has only been associated with Apple iOS devices.

Why should an android device require such improved hardware over an ipad to achieve smooth operation? Surely if apple can write an os which runs so impressively on more limitted hardware, then android should be able to do the same without having to rely on hardware upgrades?

I can't belive i just said something good about apple :s

[crossy]

I've got a Asus Transformer and its pretty awesome, can't see any reason to get the Prime as I heard the original Transformer will be getting ICS too. Seems to me like its just a small hardware upgrade. Any word on a price?

Asus have stated that ICS will definitely appear on the "old" Transformer - see http://www.pocket-lint.com/news/4262...cream-sandwich. There's been some talk on the Transformer forums that it might be as soon as January, although February is more likely. Personally I'd prefer to wait a bit longer and have Asus make 100% sure that the update is rock solid.

The video demonstrates a blazingly-fast and smooth response that, up until now, has only been associated with Apple iOS devices. Enabling this new found feel of responsiveness are subtle transition effects built in to Android 4.0, the extra-responsive touch panel of the Transformer Prime and the incredible speed of the underlying Tegra 3 processor. It's difficult to gauge from a video but this could be the most responsive tablet we've seen yet, with content appearing to load particularly quickly.

Smoothness of ICS on Prime is all well and good, but I'm more interested to see how the "release" version (remember this is apparently an pre-production early build) performs on "lesser" hardware such as the Motorola Xoom or, of course, the original Transformer. If the production copy performs similarly on this kind of gear then there'll be a lot of happy folks about - otherwise if ICS needs hardware assists like quad-core to deliver the goods, then the iPad fans will scoff loudly - and rightly so.

[Scribe]

A bold statement

Prove me wrong!

[Scribe]

Smoothness of ICS on Prime is all well and good, but I'm more interested to see how the "release" version (remember this is apparently an pre-production early build) performs on "lesser" hardware such as the Motorola Xoom or, of course, the original Transformer. If the production copy performs similarly on this kind of gear then there'll be a lot of happy folks about - otherwise if ICS needs hardware assists like quad-core to deliver the goods, then the iPad fans will scoff loudly - and rightly so.

Well, I guess it depends how you look at it. Android has always been more flexible and thus typically difficult to optimise, I mean how do you create an efficient response when for all you know the user could have a high-end live 3d desktop underneath their UI? It's difficult to create an efficient process for all these possibilities.

Another difference is that Apple typically uses customised touch panels. Though ASUS now claims that the panel on the Transformer Prime is more response than that of the iPad, was the iPad the more responsive model previously? Potentially it already had an element of hardware assist but I'm just theorising here.

I guess what I'm saying is, it's difficult to draw a comparison, but it would indeed be nice if Android 4.0 did run faster on older devices. Certainly there has been a focus on performance, if you hop over to http://www.android.com/about/ice-cream-sandwich/ you can see that like for like on the almost 1 year old Nexus S, the web-browser in Android 4.0 performs about 35% faster on the SunSpider 9.1 benchmark and scores almost 70% higher on the javascript benchmark. Let's hope that this same optimisation effort has been applied throughout the OS.