ARM Unveils Next Generation Chip Design

ARM has announced its new Cortex-A8 processor at the ARM developers conference in Santa Clara, California. The new chip design is expected to be far more powerful than current chips making it ideal for demanding mobile products running multi-channel video, audio, and gaming applications. It is expected to offer speeds of over 1GHz.

The ARM Cortex-A8 processor delivers up to 2,000 DMIPS. By comparison, current ARM 9 chips typically achieve around 150-300 DMIPS.

ARM Cortex-A8 processorARM has already secured five licensees for the Cortex-A8 processor, including Freescale, Matsushita, Samsung and Texas Instruments, and future support from major EDA and Operating System vendors.

The Cortex-A8 processor is the first applications processor based on the next-generation ARMv7 architecture, and features Thumb-2 technology for greater performance, energy efficiency, and code density. It includes the first implementation of the powerful NEON signal processing extensions to accelerate media codecs such as H.264 and MP3. The Cortex-A8 solution also includes Jazelle-RCT Java acceleration technology to optimize Just In Time (JIT) and Dynamic Adaptive Compilation (DAC), and reduces memory footprint by up to three times. Additionally, the new processor features TrustZone technology for secure transactions and Digital Rights Management (DRM), and IEM capability for low power.

The Cortex-A8 processor features an advanced superscalar pipeline which can execute multiple instructions at the same time and deliver more than 2.0 DMIPS per MHz. The processor integrates a size configurable level 2 cache which works in conjunction with fast 16K or 32K level 1 caches to minimize access time and maximize throughput. The chip is expected to offer speeds 1 GHz in high-performance 90nm and 65nm processes.

The ARM Cortex-A8 processor is available for licensing now, along with the majority of the supporting technology. ARM expects products using the new design will debut in 2007-2008.

Thanks to Jon Choo for the tip.

Article Comments


The following comments are owned by whoever posted them. PalmInfocenter is not responsible for them in any way.
Please Login or register here to add your comments.

Start a new Comment Down

Palm, No. PPC, yes.

legodude522 @ 10/5/2005 3:16:23 PM # Q
I wouldn't find a 1ghz chip in a Palm. PocketPCs tend to have faster processors to support the larger Windows Mobile. I wonder how well the efficency it is over power conservation.

Palm m125 December 25, 2003 to March 24 2004 > palmOne Zire 71 March 24, 2004 to March 31, 2005. Tapwave Zodiac 1 April 18, 2005 to present.
RE: Palm, No. PPC, yes.
sr4 @ 10/5/2005 3:29:19 PM # Q
The new HTC Wizard runs on a 195MHz processor. Its not the OS (accept for multi-tasking), but Video and VOIP is going to require FIPS no matter what the OS.

Of course you can off-load to dedicated co-processors (e.g. graphics accelerators), but that will only get you so far in the end.


Power consumption
Sam H @ 10/5/2005 4:14:08 PM # Q
I wonder how well the efficency it is over power conservation.

Projected power consumption is 300mW, which is not bad.

RE: Palm, No. PPC, yes.
KultiVator @ 10/6/2005 9:58:34 AM # Q
> Of course you can off-load to dedicated co-processors
> (e.g. graphics accelerators), but that will only get
> you so far in the end.

And generally eat more power, reducing battery life. (Before flaming me... I know that some custom ASICs can increase effeciency and theoretically reduce power-consumption - but often real-world implementations DO NOT fully exploit the features and disable the custom chip-set when not needed). Oh what it would be to live in a perfect world!

Sounds like a sensible move to have a dedicated single CPU with custom extensions specifically designed for mobile devices, especially for the (nearly universal) requirement of handling popular compressed data streams, such as MP3, MPEG, DivX, etc.

Get Real. Get Beer. Get Silly. Get KultiVated.

Reply to this comment

The news here: Jazelle-RCT

cervezas @ 10/5/2005 4:06:57 PM # Q
The relevant news about the new ARM chip design isn't so much that it can support 1GHz clockspeeds, it's that it includes a powerful new execution environment for just-in-time compilation of bytecode such as that used by Java or .NET. This will help narrow the gap between the memory bloat associated with Windows Mobile or Java apps and the efficiency of apps compiled to native ARM code (like Palm OS Protein apps for Cobalt or Palm Linux).

This seems like good news for Microsoft and RIM (whose OS is written in Java) but not so good news for PalmSource and Symbian since it erodes one of their differentiation points from the other two: fast code execution with very low overhead.

It's also likely to fuel the boom in mobile Java as a platform. ACCESS seems to be more involved with Java than PalmSource was able to be and of course there's already a lot of good Java work done on Linux, so it's possible we'll see a lot more Java apps on a future Palm OS platform and fewer that are written using the its venerable C/C++ API.

David Beers
Pikesoft Mobile Computing
Software Everywhere blog

Reply to this comment

Where does this leave Intel?

Puppy @ 10/5/2005 7:26:44 PM # Q
What I find interesting is that Intel isnít mentioned as licensing the design. Some of the features sound like SIMD-which Intel already added to the ARM 9. Does this mean Intel is going to continue building off of ARM 9, and weíll get sort of a split in the ARM market?

Iíve already wondered what itís doing to the market when some PDAs have MMX and some donítÖ

RE: Where does this leave Intel?
KultiVator @ 10/6/2005 12:05:01 PM # Q
Yes Puppy. I too feel that Intel are conspicuous by their absence from the above announcement.

Maybe they are betting their future on the new mobile x86 line of processors?

I can see some sense in such a move, but it would be a big gamble.

The old HP Journada (several years now retired) was probably the
last pocket device to have any notable consumer-level success with an x86-based hardware design.

What OS would intel envisage running on such a platform? Will they really get the power consumption and performance to levels that are suitable for current users with their media-hungry habits? Will we see products badged with 'Powered By Pocket Pentium'?

All these and other riddles may never be explained.

Got A Problem? Getting Ridiculed? Get It Aired. Get It KultVated.

RE: Where does this leave Intel?
arp @ 10/6/2005 12:10:24 PM # Q
Well, this only reinforces the trend of mobile devices growing into full-fledged laptops. They need some weaker hardware fast or the PDA market is going to be swallowed up by the laptop market... :)

RE: Where does this leave Intel?
KultiVator @ 10/6/2005 12:36:33 PM # Q
I agree. Do we really need a laptop in a PDA form-factor?

I think a Palm PDA with (amongst other things) the ability to host USB devices (such as USB connected storage, WiFi Dongles, etc) makes a lot of sense... giving users a lot of what they'd like without the need for a full blown desktop/laptop OS and associated bloat. USB has the advantage of being able to be externally powered (via the use of a powered hub) - saving PDA power (or even recharging the PDA) when docked on your desk to the USB device(s).

Forget LifeDrive - how about a new thin formfactor with new-style Palm connector AND a mini-USB host connector + old-style Palm values, such as good amount of HEAP (real RAM) as well as an improved implementation of the non-volatile memory found on recent devices, such as the T5 / Treo 650?

Everybody seems to be forgetting the wisdom in a 'Horses For Courses' approach to PDA technology, where less can often be more.

RE: Where does this leave Intel?
hotpaw4 @ 10/6/2005 2:59:43 PM # Q
Intel paid for (maybe though their acquisition of DEC StrongArm) an architecture license for ARM. That means thay can design their own CPU and processor internals instead of just putting the CPU cores designed by ARM Inc. into their chips like most other companies. IIRC, Motorola (maybe now Freescale) also bought an architecture license, but hasn't done anything with it; they still just plug in off-the-shelf ARM CPU cores.

Reply to this comment

Another nail in the coffin for wince. :)

Dr Opinion @ 10/5/2005 8:30:12 PM # Q
This is just another nail in the coffin for wince. :)

As mobile hardware increases in capcity, micro$uck will find it harder and harder to justify the cash-bleeding wince business -- micro$uck is probably already under extreme pressure internally to dump wince and move to an XP derivative. The problem is that micro$uck is an large, old, mature company, and those types of shops find it very hard to innovate and act with foresight.

Micro$uck would like to release an XP derivative on mobile devices, purely to bring their large XP developer base into the mobile space (very few XP developers would ever develop for wince, since the installed base is just too small).

However, micro$uck is unable to charge any money for wince. Once free engineering support is taken into account, wince is actually essentially given to "licensees" for free. If an XP derivative was released on mobile devices, m$ would still have to give it away for free, which would damage the entire XP price structure on the desktop. For example, Dell would insist on "free" XP pricing for tablets, then laptops, etc.

We're essentially seeing a battle inside micro$uck between desktop people (who generate revenue and have clout) and forward-thinkers who know that XP mobile is the only chance for micro$uck to compete in the mobile space.

For the time being, the desktop people are winning and holding back innovation, but look for wince to be obsoleted and the new XP Mobile to be announced withing an 18 month timeframe.

Of course, this change will cause huge resentment in the wince community. Although micro$uck desperately needs to make this change, it will set back their mobile efforts by years.

Palm Linux will corner the market for mobile devices, although we should expect to see XP Mobile as a significant niche player that will help keep Palm Linux innovating at the front of the market.

It's a sucky time to be a wince developer. :)

"People who like M$ products tend to be insecure crowd-following newbies lacking in experience and imagination."

RE: Another nail in the coffin for wince. :)
KultiVator @ 10/6/2005 10:09:42 AM # Q
I can't imagine wanting XP and the NT Kernel on any PDA of mine.

XP is way too bloated. The nicest thing with Palm devices is the lack of complexity - my 60-something parents despair of their PC when they have problems, but can both manage their Palm Tungstens without breaking a sweat. No messy driver issues for them to worry about - a simple, clean, no frills device that does what they need without drowning them in the kinds of guff Microsoft leaves too close to the surface of all their platforms!

It's true that Palm devices perhaps don't innovate in hardware design as much as many of us would like these days. But I dread the prospect of the day when an NT derivative is the only serious OS choice for heldheld device consumers. No PDA/SmartPhone/MediaPlayer owner should have to muck around with BOOT.INI or the stupendously lousy Registry!

Who knows, with development of Palm OS under a new & forward looking regime, perhaps things may end up looking a lot rosier for handheld technology in a year or so. With or without Palm the hardware vendor making it that far - their future lies in coming up with some exciting & worthy successors to the T3 and TE2 and Treo 650 in the next few months.

And I'm talking about a new design, not another plastic TE look-a-like!

And I'm talking decent media playback out of the box (MP3, MPG, DivX) in a real integrated multimedia app - not some bundled 3rd party duffer. Surely all of that BEOS knowledge must count for something? That and native excellent PDF and Flash viewer capabilities.

The Palm economy has slowed, Palm need to invigorate their stagnating pond before the frogs go hopping off in the direction of the neighbours yard.

Another short-post gone bad - arrrggghhh... Palm, you make me mad.

Get A Curry. Get Flatulant. Get KultiVated.

RE: Another nail in the coffin for wince. :)
Puppy @ 10/6/2005 11:31:48 AM # Q
KultiVator, I would LOVE NT as the core OS on a PDA. XP is much, MUCH less troubleprone than Palm's modern PDAs, not to mention it's a gazillion times more capable.

As far as being "bloated"-not every single feature in the desktop OS would be in the handheld version. The first X-Box 1 runs NT just fine with 64MB, and that's the amount most Windows Mobile devices have.

RE: Another nail in the coffin for wince. :)
KultiVator @ 10/6/2005 11:36:44 AM # Q
Not sure I agree, as...

Yes the Xbox is fine (but it's pretty much a solid state/static environment).

Win CE/Win Mob/Win Yuk on the other hand is a messy cludge of APIs layered so thick you could cut it with a chainsaw.

Any bets an XP based PDA would layer in a load of new XP PDA APIs, house a compatibility layer for Win CE/Win Mob/Win Yuk as well as a load of bloat from the current desktop OS.

Freak-Geeks aside, why would anyone want XP in their shirt pocket? Leave it on a laptop where it belongs, with a nice large screen + keyboard.

Instead, vote with your wallet for a purpose built platform, not some old crud of an OS shoe-horned to guzzle the battery-juice of the next HP/Dell PDA. That does of course mean voting by not buying anything if nothing currently hits the mark.

Love Beef Chilli? Get Mouthwash. Get Snogged. Get KultiVated

RE: Another nail in the coffin for wince. :)
Dr Opinion @ 10/6/2005 2:43:03 PM # Q
> "...I dread the prospect of the day when an NT derivative is the only serious OS choice for heldheld device consumers..."

Well, it will never be allowed to happen until Palm Linux is firmly entrenched in the market. So don't worry about lack of consumer choice.

In fact, I foresee the biggest problem will be that *all* devices will be running Linux with the Palm application stack. Ironically, I think we'll all be thankful that m$ is willing to subsidize XP Mobile as a niche player just to keep ACCESS honest... it's never good for a company to become a monopoly. :)

"People who like M$ products tend to be insecure crowd-following newbies lacking in experience and imagination."

Reply to this comment

What's the 'D' in DMIPS?

Wolfgard @ 10/6/2005 10:38:35 AM # Q
Hmm... I know MIPS, but what does the 'D' in DMIPS stands for? Digital?

pen & paper -> m515 -> Zire72 -> TH55 & Handera 330
RE: What's the 'D' in DMIPS?
KultiVator @ 10/6/2005 11:30:54 AM # Q
DMIPS = Dhrystone MIPS - a standard used in benchmarking the performance of processors by running some maths based on whole numbers (integers).

Whetstone is an alternative form a benchmark that uses more complex numbers (those with fractional parts - sometimes called floating point numbers).

Get Wordy. Get Techy. Get A Life. Get KultiVated

RE: What's the 'D' in DMIPS?
KultiVator @ 10/6/2005 12:22:35 PM # Q
I can see where you're coming from.

But the link doesn't seem to contain a correct answer for the context in which DMIP was used (e.g. ARM CPU performance).

Perhaps it's short for "Discontinue Manufacturing Ingenious PDAs"

A motto more and more manufacturers seem to follow, as they stumble onto the smartphone bandwagon.

RE: What's the 'D' in DMIPS?
SeldomVisitor @ 10/6/2005 12:42:17 PM # Q
The linked page has the words:

== "...DMIPS Dhrystone MIPS (Million Instructions Per Second)..."

Note - in the fairly-distant past I had a minor email interaction with the owner/page-designer of "" suggesting he make the pages a little less busy w.r.t. the actual acronyms and their definitions compared to all the other noise (ads et al). He appreciated the feedback but apparently didn't fully agree...

Looks like others find the pages, though full of info, hard to surf.

RE: What's the 'D' in DMIPS?
KultiVator @ 10/6/2005 12:49:17 PM # Q
Yes - so much so that I missed the definition altogether and ended up with a string of alternatives!

Thanks for coming back without flaming me for a genuine mistake!


RE: What's the 'D' in DMIPS?
hotpaw4 @ 10/6/2005 3:09:58 PM # Q
The Dhrystones benchmark has been so hacked by compiler writers that MIPS now stands for Meaningless or Marketing Indicator of Performance.

The compiler optimizers developed by several CPU manufacturers have been known to detect, remove or change parts of the Dhrystone benchmark to make their chips look faster when this benchmark is run. Take all DMIPS numbers with a grain of salt, unless comparing chips from the same vendor using the same compiler and optimization settings.

Reply to this comment
Start a New Comment Thread Top


Register Register | Login Log in