Advertising is the rattling of a stick inside a swill bucket - George Orwell
INTEL’S SUPREMACY in the server market might soon be coming in for a challenge, from none other than low-power mobile phone chip company Arm.
The company says it will go head-to-head with the behemoth Chipzilla, by attempting to bung its own multicore chip designs - including its latest offering, the Cortex-A9 - into servers.
Although originally designed for mobile devices, the Cortex-A9 is adaptable to servers due to its design, which allows for up to four cores on just one chip.
Ian Ferguson, Arm’s director of enterprise solutions, told Networkworld that certain un-named chipmakers had been whispering in his ear about wanting the firm’s low-power chip architecture to lower the amounts of power guzzle-age and thereby also the cost of server ownership.
But analysts are skeptical, mainly because x86 server architecture is seeing such a meteoric rise of late. Also, due to the x86’s burgeoning popularity, quite a bit of server software infrastructure is now being built specifically with X86 architecture in mind, which could make life a bit difficult for Arm.
Ferguson didn’t dismiss the difficulties, but noted that by partnering with software giants like the Vole and Adobe, as well as with other chipmakers, software support was definitely not beyond the range of possibility.
When it comes to competition between the two firms, it is actually Intel which first muscled its way into Arm’s space, and stepped on the firm’s toes with its low-power x86 Atom chip, but Ferguson is more than happy to take on the challenge.
Throwing down the gauntlet, Ferguson noted, "Intel is certainly there... but this isn't really to take out an x86 and stick in an Arm core. It's really looking to solve an increasingly significant challenge related to power, electricity and cooling".
He added that low-end to mid-range servers for Web serving or for running open sauce apps could also be a good area for Arm to focus on in the early stages, with the advantages being that servers with Arm chips could boost performance with extra cores when required, but could also give the CPU a low-powered break when the system wasn’t firing on all cylinders. This would be achieved by having Arm’s server run things like the Operating System, and other chips dealing with graphics, acceleration and networking.
Lets just hope Intel don’t mind a strong Arm challenge. µ
L’Inq
NetworkWorld
Can we stop this nonsense? IA vs non-IA battle has been fought and settled, and IA has come out as the undisputed winner. Intel itself is having trouble pushing its own Itanic line into servers. No other instruction-set architecture can challenge IA in well-established markets, let alone, the new-comer ARM. IBM tried this will Cell, Sun tried this with Niagara, and finally, both of them reverted back to good old Xeon or Opteron.

Low power? Who cares about low power, when each DIMM consumes 5 watts? Or each HDD consumes 15 watts? Have you ever heard of Amdhal's law? How in the world ARM aims to create synthesizable designs to compete with Nehalem MP?

Get real....
I'm kind of following up on Core Dude ... in that I also doubt ARM's ability to make in-roads into THE server market. Having said that, ARM does have some interesting technology and if they play their cards correctly could stand to make a nice gain.

The differentiating factor, as I see it, is that the Cell is simply not a server CPU. Niagara was also not designed to be a server chip -- it's more of a specialized communications chip (fascinating actually). Sun's Rock is a full-blown server CPU and will kick some serious Intel fanny -- performance-wise, though likely not market-wise, all while consuming insane amounts of power [the Rock runs a tad warm 8].

Virtualization would seem to be more important than low power consumption. However, low power is "green" so you at least have to pay lip service.

I'd think that any RISC based CPU is going to have an advantage over CISC design. Perhaps that over-generalizes what I'm really thinking, since you'd really want lots of cores NOT running at super high clocks. Efficiency of the instruction set architecture ... something like watt-instructions per clock. I'm sure that's already been hashed out elsewhere.

Btw, hope I didn't sound like drashek, sheesh already.
Most people commenting so far have said "no way can it compete with the Intel XYZ" missing the possibility that market dynamics can change very very rapidly.

Compact Flash is moving in on hard disks. (you don't need 1 terabyte drives for office documents, even scanned ones).

If electricity prices were to spike ARM+CF would be a good option, especially if you're the guy in the corner on the electricity-generating pedal cycle.
x86 is a fundamentally bad design. Even Intel has made much better chips but people were convinced they needed Windows & other apps that required x86. Yes, they did have an NT for non x86 machines but it couldn't run the apps natively.

But none of that matters with open source software. Open source will kill proprietary crap not because it's free but because it can work on much better hardware.

Even a home computer costs more to operate over it's lifetime than it does to buy. Most people don't realize that but server farmers do.
When it comes to bang per clock cycle its difficult to beat architectures like the PowerPC or the Sparc, But its obviously not what people are worried about because the x86 is charging through the pack like an out of control bulldozer --- despite it being essentially a crappy klugde of an architecture. Its probably something to do with cost effectiveness, how you get the whole hardware package for a decent cost, how you've got lots of choices you can make to get the price/performance/power point you're aiming for and the sheer volume of experience with the parts.

An x86 is low power -- provided its not actually doing anything. Same for any other processor. Its easy to be low power if all you're doing is hanging around waiting for keystrokes.