Skip to comments.ARM CEO is bullish on PC chip potential -- Executive sees tablets outselling netbooks this year
Posted on 02/16/2011 9:20:43 AM PST by Ernest_at_the_Beach
As the lines between mobile phones and computers have blurred over the past three years, Intel and ARM have found themselves in more direct competition, with the former trying to enter the mobile-phone market and ARM eyeing the traditional computer business.
(Excerpt) Read more at marketwatch.com ...
I love ARM when I’m out and about with a smartphone in my pocket.
Beyond that give me Gigabytes of Ram, Terabytes of storage, more than a foot of screen size and enough processing power to render a Hollywood movie. ARM does not cut it when I’m somewhere where I can crack my laptop open.
In the next few years, ARM-based chips will be more powerful than your current laptop. OTOH, Intel is getting the Atom processor onto SoC for tablets and phones. This will be interesting.
I have a quad core i7, capable of running 8 independent threads, that combined with the increased number of instructions a risc chip needs to complete the job means that the gap is still wider than it looks.
I want an extremely efficient processor in my pocket but I don’t want my entire computing experience limited my those power constraints.
I have to disagree with your statement. I design ARM based SOCs for a living... and have been playing with a TI OMAP3730 lately running at 1Ghz. (Yeah -it’s a single core ARM + a Digital Signal processing CPU.) It has something like 1/15th the horse power of my I5 laptop running the same operating system.
The ARM processors are good at sipping power - and they ARE comparable to the ADAM processors in compute power - but they don’t even come close to top-of-the-line Intel CPUs.
Also - ARM is dependent on the merchant SOC manufacturing for speed capability, i.e. they’re a few technology nodes behind where Intel is always playing. There CPUs are only at 1-1.5Ghz right now while Intel has been at 3Ghz for years. There are architectural differences inside the CPUs themselves that limit this speed even at the same technology node. ARM IS getting more sophisticated at it - but Intel has been using leading-edge methodology for making things for fast for a couple decades now.
For all of these reasons I don’t agree! ;-)
There CPUs are only at 1-1.5Ghz right now while Intel has been at 3Ghz for years.
High clock speed is not very conducive to the low power consumption desired in most ARM installations. Actually, in all these high-powered ARM systems coming down the pipe I'm getting a bit worried about battery life.
As I said, things are getting interesting, already in testing ARM systems are up to Core 2 Duo performance. Expand from there.
Most excellent observations from a clearcut boots on the ground expert.
Maybe the hardware world is splitting into (at least) two camps - power sipping chips for mobile clients, and performance-centric chips for server, data center applications. Just as we don’t have one size fits all, maybe we don’t have one CPU architecture fits all.
Think I'll post the Anandtech observations on this...be back in a bit....
I like them cause they try to quantify just what the chips can do.
So NVIDIA is saying....we are gonna Blitz everyone else...
Wonder if they can pull it off...when they don't have their own Fabrication Facility....
The Fermi Graphics processor was late cause the Fab at TSMC couldn't deliver.
There are lots of fabs, and NVIDIA/TSMC isn't the only player. IMHO, we're getting to the point where the integer cores are supplying all we need for 99% of what they're good for. Everything else, such as video encoding and decoding, audio, gaming, UI and animation is leveraging special purpose areas of the silicon that do the job far faster than integer cores could. For example, everything I named above is accelerated by the dedicated SIMD, graphics or video decoder portions of Apple's A4 chip. The integer core is relegated to being the scheduler for the really high-speed stuff.
Looking thru the comments...page 5...2nd from the top ...lot's of detail but note this statement:
Don't believe what Nvidia is saying (that dual channel isn't required). They know its required but for some reason couldn't get it into this chip.
Another aspect to consider is memory performance. Bandwidth is an issue and the new high performance RAM requires a lot of juice.
Yeah it does, but SoC does have the advantage of putting the RAM in the same package as the CPU.
Even with embedded memory, bandwidth is not awsome. They are having trouble feeding this generation of mobile GPU. Add faster CPU’s and GPU’s and they have a major bottleneck.
Well maybe things are going better on that front than I thought:
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.